Hybridization dynamics of invasive cattail (Typhaceae) stands in the Western Great Lakes Region of North America: a molecular analysis

Summary

1. By increasing vigour and broadening ecological tolerances, hybridization between native and introduced species may serve as a primary driver of invasiveness.

2. Cattails (Typha, Typhaceae) are clonal wetland graminoids that are known to hybridize where anthropogenic influences have resulted in distributional overlap.

3. In order to gauge the relative performance of hybrid vs. pure Typha, we characterized hybridization and clonal growth where native Typha latifolia and introduced Typha angustifolia occur together in the Western Great Lakes Region of North America.

4. Based on microsatellite markers, we documented F₁ hybrids as the most common class at five intensively sampled sites, constituting up to 90% of the genets and 99% of the ramets. Backcrosses to one or the other parent constituted 5–38% of the genets. Pure T. latifolia was rare and never constituted more than 12% of the genets.

5. F₁ hybrid genets achieved the highest mean ramet numbers at three sites, and were second in size only to T. angustifolia at two sites; however, these differences were not significant based on site-specific one-way anovas.

6. F₁ hybrids exhibited little height advantage over other Typha classes, although there was a general tendency for hybrids in relatively mixed stands to be among the tallest genets in shallow water, but among the shortest genets in deeper water.

7. Native T. latifolia was found growing at the shallowest water depths at the only site where it was sufficiently abundant to be included in statistical comparisons.

8. Synthesis. The role of hybridization in plant invasions can be difficult to confirm in the absence of molecular data, particularly for clonal species where the boundaries separating individuals are otherwise difficult to discern. Here, we used molecular markers to document the prevalence and performance of hybrid genets in five invasive Typha stands covering a broad area of the Western Great Lakes Region. We found an extremely high prevalence of F₁ hybrids within mixed Typha stands. This, coupled with the typically larger sizes of hybrid genets, suggests that hybrids are capable of outperforming other Typha spp. and that hybridization has played an influential role in the North American cattail invasion.

Introduction

Hybridization between native and introduced species is one of the primary drivers behind the evolution of invasiveness (Arnold 1997; Ellstrand & Schierenbeck 2000), often following closely on the heels of species introductions (Abbott 1992;Rhymer & Simberloff 1996; Ellstrand & Schierenbeck 2000). Hybridization may contribute to invasiveness by reducing genetic load and promoting hybrid vigour, by generating new phenotypic variants and providing the impetus for evolutionary change, by widening ecological tolerances (Arnold 1997) and even by spawning extreme phenotypes through transgressive segregation (Rieseberg, Archer & Wayne 1999). Recent theoretical modelling suggests that the rate at which hybrids spread depends not only on the relative vigour of the hybrids, but also on the degree of compatibility between the hybridizing species (Hall, Hastings & Ayres 2006). High compatibility coupled with hybrid vigour may result in greater than exponential growth of the hybrid population. Low compatibility, on the other hand, may create a lag time between the introduction of an exotic and the spread of hybrids. For example, hybridization between introduced Spartina alterniflora and native Spartina foliosa in the San Francisco Bay Area, although rare (Ayres, Strong & Baye 2003), has produced genotypes with heightened growth rates and fecundity and expanded environmental tolerances (Callaway & Josselyn 1992; Ayres, Strong & Baye 2003; Ayres et al. 2008), which now threaten to competitively exclude native S. foliosa (Daehler & Strong 1996; Ayres et al. 2004).

In the Great Lakes Region of North America, two pure and one hybrid species of cattail (Typha spp.) have come to be closely associated with degraded wetlands (e.g. Grace & Harrison 1986; Smith 1987; Tulbure, Johnston & Auger 2007), which are particularly vulnerable to the negative impacts of invasive plant species (McIntyre, Ladiges & Adams 1988; Zedler & Kercher 2004). The only unequivocal native in the region is Typha latifolia L. (broadleaf cattail), which appears in pollen and botanical records prior to the late 19th century (Shih & Finkelstein 2008). A second species, Typha angustifolia L. (narrowleaf cattail), invaded the region during the first half of the 20th century, with the rate of spread beginning to level off around 1970 (Shih & Finkelstein 2008). Until recently, T. angustifolia had been widely considered as an introduced species from Europe based on the work of Stuckey & Salamon (1987); however, Shih & Finkelstein (2008) suggest that T. angustifolia was present in North America prior to European settlement in coastal portions of the Northeast. Typha latifolia is associated with shallower water (< 15 cm) than T. angustifolia (e.g. Grace & Wetzel 1981, 1982; Waters & Shay 1992), and the two species vary in their growth capabilities as well. Typha latifolia is considered to achieve higher growth rates than T. angustifolia early in the growing season (Grace & Wetzel 1981; Weisner 1993; Selbo & Snow 2004) and under heavy shade (Hager 2004), although this advantage may only hold under low-nutrient conditions (e.g. Weisner 1993; Grace & Wetzel 1998; Hager 2004), whereas T. angustifolia achieves greater ramet densities (Selbo & Snow 2004), and exhibits rapid colonization abilities via its abundant and readily dispersed seeds (McNaughton 1966).

The spread of a third Typha species in the Great Lakes Region, T. × glauca Godron, believed to constitute a hybrid between T. latifolia and T. angustifolia, has coincided with the contact zone between T. angustifolia and T. latifolia (Chow-Fraser et al. 1998; Galatowitsch, Anderson & Ascher 1999). Compatibility between T. latifolia and T. angustifolia is considered relatively low, at least in part due to differences in flowering phenology, with T. angustifolia flowering earlier at middle latitudes (Selbo & Snow 2004). Both species are protogynous, suggesting that T. latifolia most probably serves as the maternal parent during most hybridization events (Selbo & Snow 2004). Typha × glauca is generally considered sexually sterile based on controlled crosses (Smith 1967, 1987; Dugle & Copps 1972), despite limited evidence from morphological (Fassett & Calhoun 1952), biochemical (Lee 1975) and allozyme (Sharitz et al. 1980) comparisons that reveal a range of characters exceeding what would generally be expected of a simple F₁ hybrid. Typha × glauca exhibits several characters common in invasive plants (Kolar & Lodge 2001; Bossdorf et al. 2005; Blumenthal & Hufbauer 2007), including large size (ramet heights up to 3 m) and rapid rhizomatous growth (Galatowitsch, Anderson & Ascher 1999). In terms of its ecological tolerances, T. × glauca reportedly occurs in a greater range of water depths than its parents (Waters & Shay 1990; Galatowitsch, Anderson & Ascher 1999).

Typha × glauca is considered the most invasive North American Typha species (e.g. Smith 1987; Galatowitsch, Anderson & Ascher 1999; Zedler & Kercher 2004), although some consider T. angustifolia equally invasive (e.g. Tulbure, Johnston & Auger 2007). Typha × glauca is reportedly more effective at supplanting native vegetation (Galatowitsch, Anderson & Ascher 1999; Woo & Zedler 2002; Boers, Veltman & Zedler 2007) and inhibits germination of native species once established (Frieswyk & Zedler 2006). In addition, T. × glauca competitively dominates both T. angustifolia and T. latifolia (Waters & Shay 1992; Kuehn & White 1999), with the latter generally considered the least aggressive of the three. On the other hand, recent evidence suggests that T. latifolia has been spreading as rapidly as T. angustifolia since the early 1980s (Shih & Finkelstein 2008).

Despite a general recognition of T. × glauca as the most invasive cattail species, explicit tests of the role of hybridization in the rapid spread of cattails are lacking (but see Frieswyk & Zedler 2007). Past studies of the clonal expansion of North American Typha have relied entirely on morphology and have described the expansion of whole Typha stands or discrete patches within stands (e.g. Wilcox et al. 2008) without any actual knowledge of genet structure. Unfortunately, such morphological analyses are hindered by the complex blending of characters that typically occurs in cases of ongoing hybridization (e.g. Rieseberg & Ellstrand 1993). On the other hand, the only molecular marker-based study of vegetative growth dynamics in natural Typha stands is that by Tsyusko et al. (2005) conducted in Ukraine on pure T. latifolia and T. angustifolia stands. We used a molecular marker-based approach coupled with intensive field sampling of five Western Great Lakes wetlands to test our basic hypothesis that North American Typha invasions have been driven to a large extent by hybridization between T. latifolia and T. angustifolia. Specifically, we addressed the following questions regarding the role of hybridization in the invasiveness of cattails. (1) How dominant are hybrid and backcrossed Typha genets in the sampled wetlands? (2) Is there evidence that hybrid genets are larger than either pure or backcrossed genets in terms of numbers of ramets or mean heights of ramets? (3) Is there evidence that a broadening of ecological tolerances contributes to the invasive properties of hybrids, as indicated by growth at a wide range of water depths?

Materials and methods

Site descriptions

In 2004, we conducted intensive sampling of five Typha stands in or near three US national parks occupying the Western Great Lakes region of North America (Fig. 1). Each of these stands consisted of pure or nearly pure Typha, which is the typical pattern seen during invasions. The sites sampled included one site at Indiana Dunes National Lakeshore (INDU), two at St. Croix National Scenic Riverway (SACN), and two at Voyageurs National Park (VOYA). The INDU site consisted of an 80.7-ha bog (Cowles Bog; 41°38′55″ N, 87°5′8″ W), which was historically occupied by native grasses and sedges (Lyon 1927). Since the 1930s Typha spp. have gradually become the dominant vegetation, increasing from 2.0 to 37.5 ha from 1938 to 1983 (Wilcox, Apfelbaum & Hiebert 1985), and at an even greater rate in more recent times (J. Marburger, personal observation). At SACN, the first site consisted of a diked pond adjacent to the western shore of the St. Croix River just above the St. Croix Falls hydroelectric dam (45°24′48″ N, 92°39′6″ W). Cattails are known to have colonized this site soon after its formation in 1964 and currently occupy a relatively narrow band along its western margin. The second SACN site was located outside the park boundaries along the western margin of Wolf Lake (45°36′54″ N, 92°39′54″ W), a 30-ha lake, which forms a portion of the Wolf Creek drainage c. 10-km upstream of its confluence with the St. Croix River. Cattails were observed in a relatively broad band encompassing the outer margins of Wolf Lake, where they are known to have formed dense stands since at least 1969. The two sites at VOYA were located within two large, interconnected lake systems, which together cover a total area of 1181 km². The first site, representing Namakan Reservoir, was located on the shores of Kabetogama Lake near Sphunge Island (48°26′7″ N, 92°59′43″ W). The second site, which represented Rainy Lake, was sampled at Cranberry Bay (48°35′15″ N, 93°3′21″ W). Based on aerial photography, cattails at Sphunge Island first appeared sometime between 1953 and 1972, whereas cattails were present in Cranberry Bay prior to 1948.

Figure 1.  A map of the Great Lakes Region of North America showing the locations of the three US national parks where intensive Typha sampling was conducted (Indiana Dunes National Seashore = INDU; St. Croix National Scenic Riverway = SACN; Voyageurs National Park = VOYA). Note that the two sites each at VOYA and SACN are too close to be distinguished in this figure.

Sample collection

In order to characterize Typha ramets and genets by hybrid class at each site, we systematically sampled leaf tissue from flowering cattail ramets for microsatellite genotyping. (Flowering ramets were chosen to provide data for a related study testing the reliability of flowering characteristics for species identification.) At the four lake and pond sites (excluding Cowles Bog at INDU), we designed our sampling scheme to encompass a broad range of water depths. In these cases, we sampled cattails every 2 m along three 100-m transects that followed the contours of the shoreline. The two outer transects tracked the lakeward and upland margins of each cattail stand, while the middle transect was placed at the midpoint between the two outer transects. By aligning the three transects relative to a common starting point along the shoreline, we created a sampling grid with regular spacing in a direction parallel to the shoreline and irregular spacing dictated by the width of the cattail stand in a direction perpendicular to the shoreline. At Cowles Bog, where there was no shoreline, we designed our sampling scheme to encompass areas representing three different stages in the history of the cattail invasion. We sampled cattail ramets intercepting three 100-m line-transects at 2-m intervals, one transect located in the original stand dating back to 1938, one in an area that was colonized between 1973 and 1983, and one in an area that was colonized sometime after 1983. At all sites, if no cattail ramet occurred within 0.5 m of a designated sampling point, that point was considered vacant. We sampled c. 15 cm of healthy, green leaf tissue from each ramet, which was labelled according to location and stored on ice in a Ziploc bag prior to freezing at −80 °C. In order to compare mean ramet heights and characteristic water depths among hybrid classes, each ramet was also measured for height to the tip of the male spike, and for the depth of water in which it was growing.

Determination of genet identity and hybrid class

In order to determine the genet identity and hybrid class of each individual cattail ramet, we genotyped all ramets using seven microsatellite loci. We first extracted DNA from leaf tissue using a CTAB (hexadecyltrimethylammonium bromide)-based method as described in Travis et al. (2002). We PCR-amplified a subset of the 20 microsatellite loci developed from T. angustifolia DNA by Tsyusko-Omeltchenko et al. (2003) according to the recommended protocols, including TA 3, 5, 7, 8, 13, 16 and 20, which were chosen for their high allelic diversity. Microsatellite alleles were resolved by electrophoresis on an ABI PRISM® 3130 Genetic Analyzer (Applied Biosystems, Inc., Foster City, CA, USA), and sized using Genemapper, Version 3.7.

We determined genet identities by matching alleles among ramets within each site, and calculated the probability of erroneously assigning two ramets to the same genet, which is largely a function of allele frequencies, as the sum of the squared single-locus genotype frequencies multiplied over all loci. We determined genotype frequencies for this calculation from the entire set of ramets at each site (as opposed to the reduced set of ramets consisting of just one ramet per genet). Similarly, we determined the probability for each genet that at least one of its assigned ramets was actually representative of a different genet with a matching microsatellite profile, P_sex, according to the methods of Arnaud-Haond et al. (2007a). Note that we did not consider inbreeding in generating these probabilities after observing near-zero locus-by-locus inbreeding coefficients (see Arnaud-Haond et al. 2007b, for further details). We report the latter results as the predicted number of misassignments of ramets to genets across all genets at each site, using a critical value of P_sex = 0.05.

Once we had assigned each sample to a genet based on its multi-locus genotype, we classified genets according to their hybrid status. Tsyusko et al. (2005) demonstrated the ease of distinguishing between T. latifolia and T. angustifolia based on 9–11 microsatellite loci. We distinguished between species and determined hybrid status on the basis of six microsatellite loci (including all loci listed above except TA13), which we have demonstrated to distinguish among species and to correspond with key morphological features in a related study (A. A. Snow, S. E. Travis, R. Wildová, T. Fér, P. M. Sweeney, J. E. Marburger, S. Windels, B. Kubátová & D. E. Goldberg, unpublished data). We classified genets as either pure T. angustifolia or T. latifolia if all of their alleles were diagnostic of one or the other species. We classified genets as F₁ hybrids if they possessed exactly one allele diagnostic of each species at each locus. We classified genets that exhibited a blend of hybrid and pure loci consistent with only one parental species as backcrosses, whereas we classified genets with an inconsistent pattern of parental loci as advanced-generation hybrids. Differences in the prevalence of hybrid classes were compared within sites by tallying the numbers of genets in each class and comparing them via chi-squared goodness-of-fit tests.

Genet size comparisons among hybrid classes

In order to compare genet sizes among hybrid classes, we counted the number of ramets per genet, which was a surrogate for the overall spatial area occupied by a genet. Because we had no prior knowledge of the hybrid classes existing within each site, we found it necessary to determine, a posteriori, the most appropriate model fitting procedures for inter-class comparisons of mean genet size. Optimally, we planned for a two-way analysis of variance (anova) including site as a random factor and hybrid class as a fixed factor. However, after observing that no two sites were precisely alike in their representative hybrid classes (where a particular class could only be included in statistical comparisons if it was represented by at least two genets), we were forced to run a series of separate one-way anovas by site with hybrid class as the single treatment factor. For each site, genet size data were converted to ranks in order to normalize strongly positively skewed distributions resulting from large numbers of small genets accompanied by relatively few genets that were much larger in size. We used the stats package in r, version 2.5.1, for all statistical modelling procedures.

Genet height and water depth comparisons among hybrid classes

We compared mean ramet heights and water depths among hybrid classes in order to determine whether differences in plant vigour or environmental tolerances, respectively, could help to explain the invasive spread of Typha in North America. We used as our raw data for these analyses the means of the ramet heights or water depths within genets at each site, i.e. we treated ramets as being nested within genets (Quinn & Keough 2002). As with our genet size comparisons, we were forced to run a series of separate tests by site because no two sites were precisely alike in their representative hybrid classes. For our analysis of differences in ramet heights among hybrid classes, we allowed for the possibility that water depth acts as a covariate of ramet height by employing an analysis of covariance (ancova). Note that we only used hybrid classes represented by at least three genets for this analysis because regressions plotted against just two data points are not statistically informative. For each site, we followed a standard model simplification procedure (outlined in Crawley 2007) using the stats package in r, version 2.5.1, in which we first fit a maximal ancova model to the data, which assumed that each hybrid class was represented by a different mean ramet height (i.e. y-intercept) and a unique relationship with water depth (i.e. regression slope). If this maximal model yielded a significant interaction between hybrid class and water depth, we conducted pairwise comparisons among the hybrid classes using a series of t-tests, and simplified the model by lumping classes whose y-intercepts and slopes were determined not to be different. If no significant interaction was detected, we simplified our model by dropping the interaction term and evaluating the significance of the main effect of hybrid class. If a significant main effect was detected, we likewise conducted pairwise comparisons using t-tests, and simplified the model further by lumping classes whose y-intercepts were determined not to be different. If neither a significant interaction nor main effect of hybrid class was detected, we dropped all effects from the model except for the simple linear effect of water depth on ramet height.

For our analysis of differences in water depths among hybrid classes, we used a series of one-way anovas by site, with hybrid class as the sole treatment factor. If a significant main effect of hybrid class was detected, we conducted multiple comparisons among the class means using Tukey’s HSD (family-wise α = 0.05).

Heritability of growth at specific water depths

We assessed the extent to which genets are genetically restricted to growing at specific water depths by calculating broad-sense heritabilities, H². Broad-sense heritability is measured, in the case of a clonal organism like Typha, as the proportion of overall phenotypic variance explained by variation among genets, as opposed to variation among ramets within genets, judged on the basis of a one-way anova testing for mean phenotypic differences among genets. Because broad-sense heritability is a reflection of genotypic variance, which combines with environmental variance in explaining phenotypic variance, H² necessarily declines as environmental heterogeneity increases (assuming phenotypic variance is held constant). Thus, the broad-sense heritability observed for Typha at one site under one set of environmental conditions would not necessarily be the same as that observed at another site, even if the number and frequency of representative genotypes were the same. We calculated separate H²s by site and hybrid class, subject to the following restrictions. First, because the within-genet variance estimates needed for calculating broad-sense heritabilities require the sampling of multiple ramets, we dropped all genets represented by fewer than three ramets from this analysis. Secondly, because a relatively large number of genets is required to gauge accurately mean phenotypic differences among genets, we dropped all hybrid classes represented by fewer than four genets with at least three ramets each.

Results

Due to minor site-to-site variation in ramet density and seasonal senescence, the number of Typha ramets genotyped ranged from a low of 98 at SACN Wolf Lake to a high of 150 at each of the VOYA sites (Table 1). The number of genets identified from among these ramets was also highly variable, which had a pronounced effect on the probability of erroneous matches among the ramets of different genets. In general, where relatively few genets were present we observed an elevated probability of ramet genotypes matching by chance, except at VOYA Sphunge Island where both the number of genets and match probabilities were relatively low. Erroneous match probabilities ranged from a low of 7.4674 × 10⁻⁵ at SACN St. Croix Falls to a high of 2.7768 × 10⁻² at VOYA Cranberry Bay (Table 1). The level of allelic diversity observed at each site created the expectation of three misassignments at INDU and two misassignments at VOYA Cranberry Bay (no misassignments were expected for the other three sites), which would have led us to underestimate the number of genets by < 10%. If we assume that any undetected misassignments were randomly distributed among hybrid classes, any effects on our other tests, e.g. comparisons of genet size, should have been even less pronounced. Note that no evidence of polyploidy was revealed by our genotyping activities.

Table 1.   Summary genet-matching statistics for five Western Great Lakes wetland sites where Typha genets were genotyped

Site	Sample size (N)	Prob. matching	Expected number of misassignments	Number of genets
INDU	146	2.24 × 10^−2	3	37
SACN St. Croix Falls	123	7.47 × 10^−5	0	64
SACN Wolf Lake	98	5.39 × 10^−4	0	44
VOYA Cranberry Bay	150	2.78 × 10^−2	2	21
VOYA Sphunge Island	150	1.17 × 10^−3	0	32

Hybrid class prevalence

Although the total number of genets varied considerably among sites (Table 1), all sites were dominated by F₁ hybrids between T. latifolia and T. angustifolia, both in terms of numbers of genets and numbers of ramets (Table 2). The prevalence of hybrid classes other than F₁ hybrids was highly variable among sites. The INDU site contained a small proportion of genets representing backcrosses to both parents (c. 5% of each), but no pure genets of either species, whereas the two SACN sites contained relatively large proportions of pure T. angustifolia genets (14–30%) and backcrosses to T. angustifolia (16–25%). The VOYA sites were the only sites containing pure T. latifolia genets, although T. latifolia was rare even at these sites (5–12%), and most of the genets representing backcrosses at VOYA were to T. latifolia (5–16%). Genets representing advanced-generation hybrids were observed only at the SACN sites, and were rare even there, constituting no more than 3% of each population.

Table 2.   Numbers of genets (G) and ramets (N) of each of six hybrid classes of Typha sampled from five Western Great Lakes wetland sites

Site	T. angustifolia		T. latifolia		F1 hybrids		Advanced-generation hybrids		Backcross to T. angustifolia		Backcross to T. latifolia
	G	N	G	N	G	N	G	N	G	N	G	N
INDU	0	0	0	0	33	142	0	0	2	2	2	2
SACN St. Croix Falls	19	45	0	0	21	42	2	2	17	24	9	10
SACN Wolf Lake	6	21	0	0	28	67	1	1	7	7	2	2
VOYA Cranberry Bay	0	0	1	1	19	148	0	0	0	0	1	1
VOYA Sphunge Island	4	13	4	6	18	124	0	0	1	1	5	6

Comparing among sites, F₁ hybrids constituted between 31% and 90% of all genets (mean ± SD = 66 ± 25%), and between 34% and 99% of all ramets (mean ± SD = 76 ± 27%). At INDU and VOYA Cranberry Bay, F₁ hybrids were significantly more abundant than all other classes combined in terms of both genets (INDU: χ²₁ = 22.73, P < 0.001; VOYA Cranberry Bay: χ²₁ = 13.76, P < 0.001) and ramets (INDU: χ²₁ = 130.44, P < 0.001; VOYA Cranberry Bay: χ²₁ = 142.11, P < 0.001). At SACN Wolf Lake and VOYA Sphunge Island, F₁ hybrids were the single most abundant class in terms of genets (SACN Wolf Lake: χ²₄ = 55.32, P < 0.001; VOYA Sphunge Island: χ²₄ = 27.69, P < 0.001), and were more abundant than all other classes combined in terms of ramets (SACN Wolf Lake: χ²₁ = 13.22, P < 0.001; VOYA Sphunge Island: χ²₁ = 64.03, P < 0.001). Although F₁ hybrids were the most abundant class at SACN St. Croix Falls in terms of genets, their abundance significantly exceeded only the advanced-generation hybrid class (χ²₄ = 18.47, P < 0.001); in terms of ramets, F₁ hybrids were significantly more abundant than all classes except T. angustifolia (χ²₄ = 58.67, P < 0.001).

Genet size comparisons among hybrid classes

Although F₁ hybrids exhibited the largest mean genet sizes at three of five sites and were second in size only to T. angustifolia at the remaining two sites (Fig. 2), no significant differences among hybrid classes in mean numbers of ramets per genet were detectable based on one-way anovas. Model results were as follows: INDU F_2,34 = 2.302, P = 0.116; SACN St. Croix Falls F_4,63 = 1.736, P = 0.153; SACN Wolf Lake F_3,39 = 1.824, P = 0.159; VOYA Sphunge Island F_3,27 = 1.728, P = 0.185. Combining P-values from these four independent tests (no statistical comparisons were possible at VOYA Cranberry Bay, where only F₁ hybrids were represented by multiple genets) according to the methods of Fisher (1954), a marginally significant overall effect of hybrid class on genet size was observed (χ²₈ = 15.125, P = 0.057). Note that the single largest genet was represented by an F₁ hybrid at four of five sites, with ramet numbers reaching as high as 46 per genet (out of 150 ramets at VOYA Cranberry Bay).

Figure 2.  Typha genet sizes by hybrid class at five Western Great Lakes wetland sites. Size is considered as number of ramets/genet (+1 SD). INDU = Indiana Dunes National Lakeshore; SACN SCF = St. Croix National Scenic Riverway, St. Croix Falls; SACN WL = St. Croix National Scenic Riverway, Wolf Lake; VOYA CB = Voyageurs National Park, Cranberry Bay; VOYA SI = Voyageurs National Park, Sphunge Island.

Genet height and water depth comparisons among hybrid classes

Water depth had a significant effect on ramet height at all sites, although the nature of the relationship varied among sites and at times among hybrid classes within sites, which rendered the main effect of hybrid class on ramet height difficult to interpret. At INDU and VOYA Cranberry Bay, where the effect of water depth on ramet height could be evaluated only for F₁ hybrids due to their extreme prevalence, constituting c. 90% of the genets (Table 1), ramets’ heights increased with water depth (Table 3, Fig. 3; INDU: R² = 0.313, ramet height (cm) = 236.87 + 7.10 × water depth; VOYA Cranberry Bay: R² = 0.637, ramet height (cm) = 142.61 + 1.49 × water depth). At all remaining sites, where F₁ hybrids were less prevalent, constituting an average of 50% of the genets, F₁ hybrids, as well as any other classes responding similarly to water depth, decreased in height with increasing water depth (Fig. 3; SACN St. Croix Falls: R² = 0.612, ramet height (cm) = 170.21 − 0.84 × water depth; SACN Wolf Lake: R² = 0.358, ramet height (cm) = 187.72 − 0.80 × water depth; VOYA Sphunge Island: R² = 0.670, ramet height (cm) = 307.79 − 1.56 × water depth). At SACN St. Croix Falls, backcrosses to T. latifolia were statistically indistinguishable from F₁ hybrids in terms of both ramet height and response to water depth, whereas at SACN Wolf Lake, all three representative classes were statistically indistinguishable, which included F₁ hybrids, T. angustifolia, and backcrosses to T. angustifolia. Only at VOYA Sphunge Island did F₁ hybrids respond to water depth in a way that was different from all other hybrid classes. Although a significant main effect of water depth on ramet height was observed for the three sites constituting relatively mixed stands (Table 3), there were significant differences in ramet heights among hybrid classes at only two of these sites. At both SACN St. Croix Falls and VOYA Sphunge Island, a second group of hybrid classes emerged, apart from the group that included the F₁ hybrids, which showed very little response to increasing water depth (Fig. 3). At SACN St. Croix Falls, this group included T. angustifolia and backcrosses to T. angustifolia (R² = 0.122, ramet height (cm) = 155.78 + 0.37 × water depth) whereas at VOYA Sphunge island, this group included T. angustifolia, T. latifolia, and backcrosses to T. latifolia (R² = 0.031, ramet height (cm) = 173.88 + 0.36 × water depth). This led to a significant interaction between hybrid class and water depth for the latter two sites, as well as a significant overall main effect of hybrid class on ramet height (Table 3). While the main effect of hybrid class on ramet height was difficult to interpret where it interacted significantly with water depth, the general trend was for F₁ hybrids to perform best in shallow water compared to other hybrid classes, and poorest in deep water (Fig. 3).

Table 3.   Final model results of one-way ancovas comparing ramet heights among hybrid classes by site, treating water depth as a covariate. The models for INDU and VOYA Cranberry Bay are presented as simple linear regressions of ramet height on water depth, because F₁ hybrids were the only class with a sufficient number of genets available for analysis. The model for SACN Wolf Lake was reduced to a simple linear regression after it was determined that there were no significant main or interactive effects of hybrid class on ramet height

Site	Source	d.f.	F	P-value
INDU (F1 hybrids)	Water Depth	1, 31	14.127	0.0007
SACN St. Croix Falls	Hybrid Class	1, 62	4.942	0.0299
	Water Depth	1, 62	10.353	0.0021
	Hybrid Class × Water Depth	1, 62	32.733	< 0.0001
SACN Wolf Lake	Water Depth	1, 38	21.159	< 0.0001
VOYA Cranberry Bay (F1 hybrids)	Water Depth	1, 17	29.822	< 0.0001
VOYA Sphunge Island	Hybrid Class	1, 27	22.926	< 0.0001
	Water Depth	1, 27	4.575	0.0416
	Hybrid Class × Water Depth	1, 27	9.786	0.0042

Figure 3.  The relationship between ramet height and water depth of Typha genets at five Western Great Lakes wetland sites. Proceeding from top left to bottom right, the graphs represent INDU, SACN St. Croix Falls, SACN Wolf Lake, VOYA Cranberry Bay and VOYA Sphunge Island. Where one line is drawn, either only F₁ hybrids were available in sufficient numbers to evaluate the relationship between ramet height and water depth (INDU and VOYA Cranberry Bay), or all hybrid classes were similar in both mean ramet heights and their responses to changes in water depth (SACN Wolf Lake). Where two lines are drawn, two statistically distinct groups of hybrid classes emerged, each responding differently to changes in water depth.

We found a significant effect of hybrid class on the mean water depth at which genets were growing at only one site, VOYA Sphunge Island (F_3,27 = 6.661, P = 0.002). Multiple comparisons among hybrid class means at this site revealed only one significant difference: T. latifolia was growing at a significantly shallower depth than F₁ hybrids (P = 0.002). Note that VOYA Sphunge Island was the only site where T. latifolia genets were available in sufficient numbers for inclusion in the anova. At this site, T. latifolia was found growing in the shallowest water of any hybrid class: 12.50 vs. 30.30 cm for the next shallowest class.

Heritability of growth at specific water depths

Broad-sense heritability estimates were almost entirely restricted to F₁ hybrids because of low genet numbers in other classes, and are thus the only estimates reported. These estimates were generally very low, indicating the ability of F₁ hybrid genets to spread their ramets over a wide range of water depths. Broad-sense heritability of growth at specific water depths ranged from 0.016 at VOYA Cranberry Bay to 0.248 at SACN Wolf Lake. The INDU site represented a departure from this trend, with an H² of 0.596. Mean H² at all sites other than INDU was just 0.111 ± 0.109 (SD).

Discussion

F1 hybrids were unequivocally the most dominant class at four of the five sites surveyed, constituting 56–90% of all genets, and 68–99% of all ramets (Table 2). The sole exception was represented by SACN St. Croix Falls, where F₁ hybrids and T. angustifolia were roughly equal in dominance, each constituting c. 30% of the genets and 35% of the ramets. Likely as a result of the pronounced variability in genet size representing the F₁ hybrid and T. angustifolia classes, combined with the low genet numbers in the T. latifolia and advanced-generation hybrid classes, statistical comparisons failed to show significant differences in genet sizes within any one site. Nevertheless, F₁ hybrids had the largest mean genet sizes at three of five sites, whereas average T. angustifolia genet sizes were slightly larger at the remaining two sites (Fig. 2). Four sites (excluding SACN St. Croix Falls) were characterized by a few, i.e. 2–5, highly dominant genets that each comprised 10–30% of the ramets at their respective sites. These dominant genets were exclusively F₁ hybrids at three of four sites. At the fourth site, SACN Wolf Lake, the single largest genet was an F₁ hybrid, with a second, slightly less dominant genet represented by T. angustifolia. It is also noteworthy that many small F₁ hybrid genets, consisting of fewer than three ramets, were present at all sites, perhaps reflecting recent seedling recruitment. Again with the exception of SACN St. Croix Falls, which represented a more even species mixture, an average of 68% of small genets were F₁ hybrids across all sites. The remaining classes were in stark contrast to F₁ hybrids (and to a lesser extent T. angustifolia), with the largest genets at each site, combined over classes, averaging fewer than two ramets. In fact, only one genet in 55 of these underrepresented classes consisted of more than three ramets. Overall, the wide range in genet sizes of F₁ hybrids, including many small genets in addition to some highly dominant genets, suggests that they are competitively superior in terms of both vegetative growth and seedling recruitment, with T. angustifolia nearly as or equally dominant under limited circumstances. This idea is in keeping with the findings of other researchers who have documented the competitive superiority of T. × glauca (e.g. Smith 1987; Galatowitsch, Anderson & Ascher 1999; Zedler & Kercher 2004).

One of the morphological features that could make T. ×glauca competitively superior to other Great Lakes Typha spp. is its greater ramet height, which could limit light and nutrient availability relative to other species. We found relatively little height advantage for F₁ hybrids, after allowing for the possibility of covariation between ramet heights and water depth for all hybrid classes considered. An intriguing pattern did emerge in the relationship between ramet height and water depth, but will require further investigation to confirm. At the two sites where F₁ hybrids constituted c. 90% of the genets, and were thus almost entirely free from competition with other Typha spp., they increased in height with increasing water depth. At all other sites, where F₁ hybrids constituted an average of 50% of the genets, and where competition may have been more intense, F₁ hybrids tended to be at their tallest in shallow water but to decline in height with increasing water depth, a trend that was reversed in some other competing classes. These results suggest that F₁ hybrids may perform best in deep water in the absence of competition, but may enjoy their greatest competitive advantage in relatively shallow water, perhaps explaining why they have been relatively more effective at supplanting T. latifolia, which has been reported to prefer shallower water than T. angustifolia (e.g. Grace & Wetzel 1981, 1982; Waters & Shay 1992).

Not surprisingly, given the range of water depths at which we observed F₁ hybrid genets, they often exhibited broad-sense heritabilities of growth at specific water depths that were quite low, as low as 0.0156. This observation is consistent with clonal amplitudes that encompass a wide range of water depths. These low heritability values do not necessarily mean that there are no genetic constraints on the ability of hybrids to grow at different depths, just that they are capable of a highly plastic response to water depth variation. One notable exception to the generally low heritability values that we observed occurred at INDU, where H² reached as high as 0.60. However, INDU also exhibited an unusually small range of water depths, just 0–10 cm (Fig. 3), and this lack of environmental heterogeneity may have been largely responsible for the high heritability estimate observed.

Backcrosses, either to T. latifolia or T. angustifolia, were present at all of our sites, collectively making up 5–41% of the genets. This observation, considered in the context of earlier studies of Typha hybridization, raises several important questions. First, why has evidence of backcrossing not been found previously in Typha, and secondly, what are the consequences of backcrossing for the continued evolution of invasiveness within the genus? Despite several recent molecular marker-based studies (e.g. Kuehn, Minor & White 1999; Tsyusko et al. 2005), which have failed to reveal evidence of backcrossing, earlier researchers noted what appeared to be molecular or morphological intermediates between F₁ hybrids and their respective parents (Fassett & Calhoun 1952; Lee 1975; Sharitz et al. 1980). These conflicting accounts suggest that backcrossing is a relatively rare event, but that the progeny of such unions are sufficiently fit, at least in our study, to retain a foothold in predominantly F₁ hybrid populations. A potential consequence of backcrossing is the rapid evolution of increasingly aggressive genotypes (Ellstrand & Schierenbeck 2000), as has been seen in invasive Spartina (Ayres et al. 2008). It may be too soon to gauge fully the competitive abilities of backcrossed Typha; however, at the time of our study, backcrossed genets were typically among the smallest observed, consisting of a mean of just 1.2 ramets, whereas we observed F₁ hybrid genets consisting of up to 46 ramets. These results suggest that the vigour seen in F₁ hybrids is primarily a result of heterozygote superiority, or heterosis, which would tend to decline in backcrosses and advanced-generation hybrids, reducing their relative vigour.

Buggs (2007, see also Currat et al. 2008) suggested that introgression of neutral markers will typically occur from a native species to an invading species as the invading species advances. This would suggest that the proportion of apparent backcrosses to each parental species that we observed in Typha could be used to determine the relative age of an invading T. angustifolia population. More recent invasions would necessarily exhibit more backcrosses to T. latifolia, whereas older invasions should exhibit more backcrosses to T. angustifolia. If we apply this rule to our data, we would conclude that T. angustifolia reached north-western Wisconsin first among our sites, followed by north-western Indiana, and then north-eastern Minnesota. Based on the findings of Shih & Finkelstein (2008), which document the gradual westward spread of T. angustifolia from the region of New England, INDU (Indiana) should represent the first of our sites to be colonized. On the other hand, SACN (Wisconsin) would appear to be the first site reached by T. angustifolia based on both the pattern of introgression and the number of genets per unit area that we observed. These apparently conflicting results suggest that the general westward progression of T. angustifolia has varied on a local or regional scale. Cowles Bog, in particular, may have been buffered from invasion by its lack of connectedness with flowing water, potentially explaining its later invasion than SACN.

The relative ubiquity of backcrossed genets was observed despite the absence of pure genets of the corresponding parental species at several sites. For example, we observed backcrosses to T. latifolia but no pure T. latifolia at both SACN sites. In all likelihood, more exhaustive sampling at these sites would have turned up pure individuals of the corresponding parent. It is also possible that backcrossed seed and/or pure pollen had dispersed into our study sites, thereby accounting for the presence of backcrosses in the absence of their pure parent(s). Finally, it is possible that T. latifolia was present formerly but has since been competitively excluded. Typha latifolia is an inferior competitor to both T. angustifolia and T. × glauca at high nutrient levels (e.g. Weisner 1993; Grace & Wetzel 1998; Hager 2004). We only observed T. latifolia at two sites, both at VOYA, and never observed more than two ramets per genet, which may further attest to the poor competitive abilities of this species. In fact, we suggest that T. latifolia may be in danger of extirpation from many areas where it has historically been present. Although recent work by Shih & Finkelstein (2008) seems to indicate that T. latifolia is currently spreading as rapidly as T. angustifolia and T. × glauca, this conclusion was based on a morphological analysis, which can be a notoriously unreliable method of species identification in cases of ongoing hybridization (e.g. Rieseberg & Ellstrand 1993). Unfortunately, hybrid assimilation is not an unfamiliar process where a common species has invaded the range of a native congener (e.g. Rhymer & Simberloff 1996; Perry, Lodge & Feder 2002). The preference of T. latifolia for shallow water, as shown here and by others (e.g. Grace & Wetzel 1981, 1982; Waters & Shay 1992), may represent the sole refuge for T. latifolia growing in mixed stands.

In conclusion, we suggest that hybridization has been an extremely important driver of invasiveness in North American cattail. Having examined relative genet sizes by hybrid class at five sites in the Western Great Lakes region that were chosen randomly in terms of the wetland types they represented, we observed a strikingly consistent pattern of dominance by T. × glauca. One of the critical questions that remains to be answered is whether the spread of T. × glauca is dependent on anthropogenic forms of disturbance such as altered hydrology and elevated nutrient levels (e.g. Wilcox, Apfelbaum & Hiebert 1985; Woo & Zedler 2002), or whether hybrid vigour alone would be sufficient for the evolution of cattail invasiveness. Thus, future research should aim to apply a common garden or mesocosm approach to the elucidation of the relative roles of constant vs. variable water levels, high vs. low nutrient levels, and low vs. high levels of competition with native wetland taxa, in the spread of T. latifolia, T. angustifolia, T. × glauca and its respective backcrosses.

Acknowledgements

We thank T. Armstrong, J. Fox, S. Lehmann, B. Sikes, W. Smith, C. Trembath and K. Wessner for field and laboratory assistance, R. Mercklein for field support at St. Croix National Scenic Riverway and D. Mason for granting permission to reproduce his Cowles Bog vegetation maps. This work was supported by a Park-Oriented Biological Support grant to the USGS National Wetlands Research Center and the National Park Service Midwest Region. The use of trade names is for descriptive purposes only and does not imply endorsement by the US Government.