• Here we tested two possible nonexclusive explanations for the maintenance of a hybrid swarm between Senecio jacobaea and Senecio aquaticus; first, that genotype-by-environment interactions involving water and nutrient clines are involved in hybrid fitness, and second, heterosis in early hybrid generations may provide an initial hybrid advantage that contributes to hybrid persistence.
• In three climate chamber studies, fitness and root growth were measured for parental species and natural and artificial F1 hybrids, in order to determine whether hybrids occur in habitats where they are more fit than parental species.
• Natural hybrids, which are generally back-crossed to S. jacobaea, always equaled S. jacobaea in growth characteristics. Maternal effects played a role in the fitness of F1 hybrids, with offspring from S. jacobaea mothers exhibiting higher fitness than those from S. aquaticus mothers, and compared with parental species and natural hybrids.
• Natural hybrids are not distributed in zones where they are most fit with respect to nutrient and water regimes. Superior fitness of early generation hybrids may contribute to hybrid swarm stability.
It is now widely accepted that hybridization can contribute to evolutionary processes (Arnold, 1997; Rieseberg, 1997; Rieseberg & Carney, 1998). From introgression (Rieseberg et al., 2000), to speciation (Rieseberg, 1997), to adaptive radiation (Seehausen, 2004), hybridization has been credited with diversification of innumerable traits, as well as species. Yet not all hybridization events have long-term consequences; some natural hybrids, for example, may be ephemeral, or sterile, such that no future genetic contribution can occur via the hybrid lineage. The fate of hybrids in natural hybrid zones, and thus their evolutionary potential, is thought to be dependent on a number of factors, the most important of which may be the fitness of hybrids in relation to parental species: ‘if hybrids were uniformly less fit than the parental species, the role of hybridization in adaptive evolution would be minimal …’ (Rieseberg & Carney, 1998).
There are two classes of models explaining the stability of hybrid zones in nature. Environmentally independent models propose that hybrids demonstrate some fitness level in relation to parental species, which is fixed regardless of environment. Such models are epitomized by the tension zone model proposed by Barton & Hewitt (1985), which relies on intrinsic fitness inferiority of hybrids. Barton & Hewitt (1985) propose that maintenance of hybrid swarms results from an equilibrium between continuous dispersal of hybrids into hybrid zones, and subsequent negative selection against such hybrids.
Environmentally dependent models (Moore, 1977; Harrison, 1986, 1990; Howard, 1986) of hybrid zone stability involve ‘genotype-by-environment’ interactions, according to which variable hybrid genotypes interact with environmental gradients to produce zones where hybrid fitness differs in relation to parental fitness. Many empirical examples from plant literature confirm that at least in plants, hybrids indeed often vary in fitness in relation to parental species across environmental gradients (Arnold & Hodges, 1995; Arnold, 1997; Campbell & Waser, 2001; Johnston et al., 2001). Furthermore, superior hybrid fitness in environments intermediate, or unique in comparison, to parental environments may contribute to the maintenance of hybrid populations (Arnold & Hodges, 1995a,b; Arnold, 1997). For example, in a study of F1 reciprocal crosses between Ipomopsis aggregata and I. tenuituba, Campbell & Waser (2001) found an interaction between maternal species and clinal location of planting, such that hybrids from I. tenuituba mothers were more fit than other plant groups in the hybrid habitat. Similarly, Wang et al. (1997) showed local adaptation of two subspecies of Artemisia tridentata and their hybrids to their respective environments. Other authors have pinpointed examples of single environmental factors that mediate hybrid genotype–environment interactions, such as light intensity (Iris fulva ×Iris hexagona; Arnold, 1997).
One of the shortcomings of many hybridization studies is that experimentation is limited only to early generation artificial hybrids, or to natural hybrids that have a complex crossing history. For instance, it is an agricultural paradigm that F1 hybrids generally exhibit heterosis over parental species, and that such heterosis can be conserved in early generation back- and intercrosses. Still, many researchers only use F1, F2, and BC generations for studies of hybrid fitness (e.g. Burke et al., 1998; Campbell & Waser, 2001; Johnston et al., 2001; Parris, 2001), which may result in misleading conclusions. Conversely, studies only using natural hybrids, which are often highly back- or intercrossed, may fail to encompass the processes that occur early in hybrid zone formation, or population dynamics affected by early (i.e. F1, F2, and BC) hybrid generations occurring in stable populations (Carney et al., 2000).
Here, we select two Senecio (Senecioneae; Asteraceae) species of which hybridization reports are common and widely distributed, in order to test whether hybrid vigor and genotype-by-environment interactions may lead to hybrid superiority and persistence in nature. We include parental species, F1 hybrids, and natural hybrids in an attempt to understand the processes leading to the maintenance of a natural hybrid swarm.
The Senecio genus is extremely diverse, containing at least 1500 species distributed throughout the world. Hybridization is thought to be common within the genus (Lowe & Abbott, 2000; Kirk et al., 2004), and hybridization processes may have partially contributed to historical diversification. Senecio jacobaea L. and S. aquaticus Hill. are typically characterized as drought and flood resistant species, respectively. While S. aquaticus and S. jacobaea are closely related (Pelser et al., 2003), field observation and literature (Chater & Walters, 1976; Weeda & Deursen, 1994) indicate that these species occupy distinct ecological niches. Molecular analysis of our study populations confirms that natural hybrids are generally backcrossed to S. jacobaea (Kirk et al., 2004).
We present the results of three climate room experiments, which are intended to explore differences between S. aquaticus and S. jacobaea and hybrids with respect to growth across a variety of environmental conditions. We test the following hypotheses:
1Genotype-by-environment interactions determine parental and hybrid fitness, such that each group is more fit than the others in endemic water and nutrient conditions.
2Performance of F1 hybrids differs depending on parental maternal species. If this is true, we expect that hybrids from S. jacobaea mothers are more fit, because we observe that natural hybrids are highly backcrossed to S. jacobaea.
3Performance of F1 hybrids is greater than that of natural hybrids. If this is true, F1 heterosis may prolong hybrid breakdown, or contribute to the success of natural hybrid swarms.
4Root elongation ability is a key adaptation for adaptation of S. jacobaea to drought, and may contribute to the fitness of natural hybrids in drying environments.
Materials and Methods
Senecio jacobaea L. and S. aquaticus Hill. are monocarpic, biennial to perennial, self-incompatible species native to Western Eurasia. Senecio jacobaea colonizes areas subject to high disturbance such as roadside ditches and sand dunes, and S. aquaticus is an established resident of flood-prone marsh areas. Little overlap seems to occur between the local distribution of these species (Weeda & Deursen, 1994).
Viable hybrids between S. jacobaea and S. aquaticus have been reported from a number of locations including the UK (Stace, 1975), Germany (Christian Düring, personal communication), and the Netherlands (Kirk et al., 2004). In this investigation, we study natural S. jacobaea×S. aquaticus hybrids from the Zwanenwater reserve (the Netherlands). Composed mostly of sand dunes, the Zwanenwater reserve contains a small lake around which a hybrid population exists (Fig. 1). Senecio jacobaea are abundant in the dunes surrounding the lake, while S. aquaticus occurs infrequently at the lake fringe. Alleged hybrids, first observed at the Zwanenwater in 1979 (Ruud van der Meijden, personal communication) can be found in a narrow zone spanning a bank at the edge the lake, which appears to be intermediate to parental sites with regards to soil organic content and humidity.
Seeds of S. jacobaea, S. aquaticus, and natural hybrids were collected from plants in the field during 2001 and 2002. Putative hybrids were identified in the field based on leaf lobe and flower morphology, and were later confirmed to be hybrids based on diagnostic amplified fragment length polymorphism (AFLP) markers (Kirk et al., 2004).
F1 hybrids were produced by collecting second year rosettes of parental plants, exhibiting the development of flowering stems, from the field. To minimize chances that introgressive genes were present in experimental parents, Senecio aquaticus individuals were collected from a marshy agricultural grassland c. 500 m from the hybrid zone (Fig. 1), and S. jacobaea individuals were collected from dunes located c. 300 m from the hybrid zone. Plants from both species were placed in a glasshouse, allowed to flower, and were crossed in pairs of S. jacobaea×S. aquaticus by rubbing flowerheads together. Seeds were harvested from both parental plants.
Determination of soil water and organic content from field sites
Soil samples were obtained from soil surrounding the roots of hybrid and parental plants collected during August 2001. Ten soil samples were analyzed from each plant group (hybrids, S. jacobaea, and S. aquaticus). Soil samples were weighed, dried in an oven, and weighed again to determine soil moisture content. Samples were then ignited at 500°C for a period of 4 h to burn off organic matter, and were reweighed to determine organic content. Percent water and organic content were calculated based on weight.
Climate room experiments
In all cases, seeds were germinated in Petri dishes on moist filter paper (light 16 h, temperature 20°C, relative humidity 100%). Approximately 1 wk after germination, equal sized seedling were selected and placed under experimental conditions (light 16 h, temperature 20°C: 15°C, rh 70%).
Experiment 1 Experiment 1 was designed to test whether the fitness of natural hybrids is greater than parental species in intermediate nutrient and water conditions. Seeds from four different S. aquaticus maternal plants, four different S. jacobaea maternal plants, and seven different natural hybrid maternal plants were used in the experiment. We transplanted six equal sized seedlings from each maternal plant into separate 10 cm × 10 cm × 10 cm pots, three of which contained potting soil and three of which contained dune sand, so that one offspring from each maternal plant was subjected to each treatment. All plants were given normal water for 1 wk to allow for the establishment of the seedlings. Following this week, dry, wet, and intermediate soil conditions were established. The six treatments thus included all combinations of soil types and water treatments. Dry treatments were given small amounts of water from the bottom of the pots three times during the course of the experiment in order to keep the plants alive while maintaining near drought conditions. Intermediate groups were kept in 1 cm of water, and wet group pots were submerged in water to a depth of 1/2 cm from the top of the pot.
Experimental conditions were maintained for 6 wk until plants were harvested. Harvested plants were measured individually for total fresh and dry biomass.
Experiment 2 The purpose of experiment 2 was to test whether root length and water table depth plays a role in the relative fitness of parental species and natural hybrids. We randomly selected seeds from four different S. aquaticus maternal plants, four different S. jacobaea maternal plants, and seven different natural hybrid maternal plants for experimental use. One equal sized seedling from each maternal plant was transplanted into each of four experimental column (15 cm diameter) lengths: 20 cm, 45 cm, 75 cm, and 100 cm. The columns contained soil composed of 50% dune sand and 50% potting soil. All columns were given sufficient water at the beginning of the experiment to allow for seedling establishment. Therefore at the beginning of the experiment, soil throughout the total length of the column was moist. Experimental conditions were established 2 wk after seedlings were transplanted to columns. Twenty cm columns were placed in bins in which the water table was varied from 1 cm above the soil surface to 3 cm below the soil surface. Plants in 45-cm columns were watered regularly from the top. Plants in 75- and 100-cm columns were given no further water throughout the course of the experiment. Because evaporation occurred from the soil surface, plants in 75- and 100-cm columns experienced an increasing gradient from dry to moist with column depth during the course of the experiment. Plants were harvested 8 wk after the establishment of experimental conditions, when growth curves began to level out. Roots were carefully washed, and root length was measured by extending roots to full length on a table. Harvested plants were measured individually for total fresh and dry biomass.
Experiment 3 Experiment 3 was designed to test whether artificial F1 hybrids are more fit than natural hybrids, whether maternal effects play a role in the fitness of F1 hybrids, and whether root length is plastic or fixed in S. jacobaea and hybrids. Small clonal plants (tissue culture) were used rather than seedlings in this experiment. We selected five S. aquaticus genotypes, five S. jacobaea genotypes, and five natural hybrid genotypes for experimental use. We also included genotypes from F1 producing crosses, from both parental plants in the reciprocal cross, such that genotypes originated from five crosses, and 10 parental plants (five S. aquaticus mothers, and five S. jacobaea mothers). F1 genotypes were unrelated to parental genotypes used in this experiment. One equal sized clone from each genotype was transplanted into each of six experimental columns (1 m length, 15 cm diameter), yielding a total of 150 experimental plants. One plant from each genotype was thus subjected to each of six experimental treatments.
The experiment was established to test a combination of two nutrient and three water treatments. We used sieved dune sand to fill all columns. In half the columns, the dune sand was mixed with ‘Osmocote’ slow release fertilizer (N : P : K = 15 : 11 : 13 + 2MgO) at a concentration of 1.3 g l−1 sand to provide a nutrient rich medium. After establishment of seedlings, columns were partially submerged in water of three different depths: 5 cm, 50 cm, and 100 cm.
All columns were given sufficient water at the beginning of the experiment to allow for seedling establishment. Therefore at the beginning of the experiment, soil throughout the total length of the column was moist. Experimental conditions were established 2 wk after seedlings were transplanted to columns.
Eight weeks after establishing experimental conditions, we emptied two ‘dry’ columns in which seedlings had died, in order to observe the depth at which moisture was available to plants growing in dry treatments. We observed that moisture was available at a depth of 45 cm, and we therefore allowed the remaining 148 experimental plants to grow for a further 2 wk, after draining the 5 cm water from the ‘dry’ treatment tubes.
At harvesting, f. wt and d. wt of above- and below-ground plant parts (shoots and roots) were measured. Roots were carefully washed, and root length was measured by extending roots to full length on a table.
All analyses were carried out using SPSS. Data were tested for normality, and data from Experiment 3 was square-root transformed to achieve normality. Dry mass was used as an estimator of fitness, as dry mass is highly correlated with seed production in monocarpic plants (Klinkhamer & De Jong, 1987).
Data were analyzed using mixed model two-way anovas to determine whether dependent variables (d. wt and root length) differed according to plant group (random factor), treatment (fixed factor), or an interaction between plant group and treatment. Plant group was treated as a random factor because the groups tested here represent only a few of a large number of classes (i.e. there are a range of hybrid classes, and we only sampled two of them here; see Green & Tukey, 1960, for a more detailed explanation). In experiments 1 and 3, treatment was broken up according to soil type and water regime for multifactoral analysis. In Experiment 3, we first tested for differences between F1 hybrids from differing maternal species, and based on the results, defined offspring from different parental species as separate groups for analysis of the total data.
In cases where species was a significant factor, we further analyzed data by separating data into species and using one-way anovas followed by Tukey multiple comparisons to identify differences between treatments per species.
Determination of soil water and organic content from field sites
Water content of soil collected from the root zone of all three plant groups differed significantly, while organic content of S. aquaticus soil differed from that of the other two plant groups (Fig. 2). Hybrids are found on sites intermediate to parental species sites in soil water content, and similar to S. jacobaea sites in soil organic content. Senecio aquaticus is found on moist organic soil, while S. jacobaea occupies dry sites with little organic content.
Overall, all plant groups attained maximal dry mass when grown in organic soil with sufficient but not excess water (medium organic/nutrient treatment), except F1 hybrids in Experiment 3, which performed best in wet, nutrient rich soil. Except under extreme drought (Experiment 2), S. aquaticus outperformed or equaled S. jacobaea, and we never observed a significant difference between natural hybrids and S. jacobaea under any treatment.
Experiment 1 anova demonstrates that plant d. wt was affected by an interaction between treatment, soil type, and plant group (Table 1). Except in sand, S. aquaticus performed best in all treatments. In sand, S. aquaticus increased in d. wt with increasing soil moisture, while in potting soil, d. wt was highest in the medium water treatment. Conversely, S. jacobaea and hybrids were always negatively affected by flooding by comparison with the intermediate water treatments (Fig. 3), although in sand the negative effect is not significant. Although a significant three–way interaction occurs, natural hybrids never outperform either parental species in any of the treatments.
Table 1. anova results experiment 1: the effects of soil medium (S), water treatment (W), and plant group (P) on total d. wt
Total d. wt
S × W
S × P
W × P
S × W × P
8.497 × 10−2
Experiment 2 Plant group and column length had an interactive effect on both total d. wt and root length (Table 2). Again, we noted no significant differences in d. wt or root characteristics between S. jacobaea and hybrids. Senecio aquaticus attained higher d. wt than S. jacobaea and hybrids when grown in conditions subject to frequent flooding, but performed worse in all other treatments (Fig. 4a). Senecio jacobaea and hybrids decreased in d. wt as column length increased, performing best when water availability was high (20- and 45-cm columns), and significantly worse when water availability was low (75- and 100-cm columns). Plant group–environmental interactions never led to circumstances in which natural hybrids performed better than parents.
Table 2. anova results Experiment 2: the effects of column length (C) and plant group (P) on total d. wt and root length
Total d. wt
C × P
C × P
Root length was affected by a significant interaction between plant species (random factor) and column length (fixed factor) (Table 2). Root length of S. aquaticus remained constant across column lengths, while root length of S. jacobaea and hybrids increased with increasing column length. As a result, root length of S. aquaticus was lower than that of both hybrids and S. jacobaea at column lengths of 45 cm, 75 cm, and 100 cm.
Experiment 3 Four plants died during the experiment, and we excluded them from our analysis.
Maternal effects interacted significantly with nutrient and water treatments to determine the fitness of F1 hybrids (Table 3), such that offspring from S. jacobaea mothers perform generally better than those from S. aquaticus mothers (Fig. 5).
Table 3. anova results testing for maternal effects in F1 hybrids: interactions between nutrient (N) and water (W) treatments and maternal species (M) on total d. wt
Total d. wt
N × W
N × M
2.303 × 10−2
W × M
2.062 × 10−2
N × W × M
6.757 × 10−4
We found significant interactions between both plant group and nutrient treatment, and plant group and water treatment (Table 4). Natural hybrids performed equally well as S. jacobaea in all conditions (Fig. 5a), supporting results presented from Experiments 1 and 2. Artificial hybrids from S. jacobaea mothers exhibited hybrid vigor, always performing better than either parent (although differences were not always significant). F1 offspring from S. aquaticus mothers performed better than parental species only in nutrient rich treatments.
Table 4. anova results Experiment 3: interactions between nutrient (N) and water (W) treatments and plant group (P) on total d. wt and root length
Total d. wt
N × W
N × P
W × P
7.058 × 10−2
N × W × P
7.405 × 10−3
4.044 × 10−2
N × W
N × P
W × P
N × W × P
While species was not a significant factor in the overall anova (Table 4), there was some evidence for species–by–environment interactions in root length (Fig. 5b). In sand, S. jacobaea and both hybrid classes exhibited more plasticity in root length than did S. aquaticus (Fig. 5b); as in Experiment 2, root length of S. aquaticus did not differ across differing treatments, while root length was significantly different between dry and wet treatments for S. jacobaea and hybrids (Fig. 5b). However, in the nutrient treatment, S. aquaticus was able to extend its roots equally well to S. jacobaea, suggesting that given sufficient nutrients, and a weak drying gradient (compared with Experiment 2), S. aquaticus can indeed extend roots to lengths reaching at least 1 m. Both natural and artificial hybrids have significantly longer roots in dry conditions than in wet, in both sand and nutrient treatments, and thus exhibit plastic root elongation responses to drying that are similar to S. jacobaea.
While laboratory and field studies confirm that natural hybrids often vary in fitness in relation to parental species across environmental gradients (Arnold, 1997; Rieseberg & Carney, 1998; Campbell & Waser, 2001), and that natural hybrids can in fact be relatively more fit than parental species in certain environments (Arnold & Hodges, 1995), we did not find that this is the case here. Contrary to the expectations of environmentally dependent models of hybrid zone stability, we did not find that hybrids, either artificial or natural, vary in fitness with relation to parental species, or that natural hybrids are distributed in zones where fitness is greatest.
By contrast to environmentally dependent models, Barton & Hewitt (1985) propose that maintenance of hybrid swarms results from an equilibrium between continuous dispersal of hybrids into hybrid zones, and subsequent negative selection against such hybrids. While passive processes such as dispersal and neutral selection could certainly play a role in the persistence of natural S. jacobaea×S. aquaticus hybrids in the Zwanenwater, we did not find that natural hybrids are less fit than parental species, as expected in a tension zone. Senecio aquaticus is relatively rare in the immediate area of the hybrid zone, while S. jacobaea is fairly abundant. Repeated hybridization events followed by back-crossing to relatively abundant S. jacobaea could be an explanation for the observed distribution, and genetic composition (high levels of back-crossing; Kirk et al., 2004) in the hybrid population.
Our data, however, are not completely compatible with environmentally independent models such as that of Barton & Hewitt (1985), which assume that most hybrids have a fixed fitness level relative to parents. We have found that different hybrid generations differ in fitness, with early generations performing much better than natural hybrids, which have presumably undergone many generations of intercrossing and back-crossing (Kirk et al., 2004). F1 hybrids possess a combination of adaptations, including flooding and drought resistance, which are normally found exclusively in either one or the other parent. Such adaptations, combined with superior fitness of F1 hybrids (from S. jacobaea mothers) in all environments, may allow early generation hybrids to expand their range to the habitats of both parents. Unfortunately for the hybrids, which cannot reproduce clonally, back-crossing and intercrossing must occur, reducing the fitness of subsequent generations, such that heterotic F1s cannot dominate the population. Our findings do not completely agree with any current models that explain hybrid zone stability, and we propose that heterosis followed by a decrease in hybrid fitness could contribute to the stability of hybrid swarms.
That maternal effects have a significant role in the fitness of hybrids may also have notable consequences for hybrid swarm dynamics. Not much is known about the evolutionary genetics of, and mechanisms behind maternal effects, although it is thought that such effects can be evolutionarily adaptive and subject to natural selection (i.e. Thiede, 1998; Fox et al., 1999). In the case described here, experimental plants were tissue cultured and selected for size equality at the beginning of the experiment, which rules out possible maternal effects on seed size, germination, and early growth, factors which are often influenced by maternal environment (Platenkamp & Shaw, 1993). Thus, maternal benefits to offspring might have been conferred by non-nuclear inheritance of genetic material from cytoplasm and organelles of maternal plants (Roach & Wulff, 1987). Alternatively, fitness of offspring might be epigenetically regulated (e.g. Jaenisch & Bird, 2003) by the mother plant, such that mother plants control nuclear gene expression in the offspring through cues passed via the cytoplasm.
Maternal effects on the fitness of hybrids might also partially explain back-crossing of natural hybrids in the Zwanenwater reserve to S. jacobaea (Kirk et al., 2004). If S. jacobaea mothers produce offspring that are more fit than those produced by S. aquaticus mothers, back-crosses to S. jacobaea might have a competitive advantage, and/or greater reproductive success than other hybrid classes. At least one other study (Campbell & Waser, 2001) showed that fitness of F1 hybrids can be mediated by parental species. To test whether natural Senecio hybrids indeed originate mostly from S. jacobaea mothers, it would be interesting to examine the origin of cytoplasmic genes in natural hybrids.
Distribution of parental species S. jacobaea and S. aquaticus is well accounted for by adaptation of root responses to local water and nutrient regimes in the respective environments of these species. Differences between root elongation ability of S. jacobaea and S. aquaticus in response to low water tables are striking. When soil moisture is available at greater depths and nutrients are lacking, S. jacobaea uses an avoidance strategy in order to reach such water sources that are unavailable to S. aquaticus, and therefore maintains a significantly higher d. wt than S. aquaticus when water is scarce. In all other growth situations we have shown that S. aquaticus is equally or more fit than S. jacobaea. As expected, S. aquaticus performs best (both relative to other species and to itself) in conditions similar those in which this species grows in the field (medium to wet soil with high organic content).
Unlike S. aquaticus, S. jacobaea and natural hybrids suffer reductions in d. wt when subjected to soil saturation. Furthermore, leaf yellowing of S. jacobaea and natural hybrids becomes apparent after several weeks of growth in wet soils (data not shown). It is therefore clear that S. jacobaea and natural hybrids are susceptible to anoxia related stress, which may explain why these groups are excluded from flood susceptible environments.
To our knowledge, this is the first instance in which superiority of F1 hybrids over natural hybrids has been considered as a potential contributing factor for hybrid swarm stability. It is often assumed that hybrid swarms that exhibit low fitness or hybrid breakdown reinforce reproductive isolation between parental species (Barton & Hewitt, 1985), and thus have little evolutionary potential (Rieseberg & Carney, 1998). If the expansion and/or persistence of hybrid populations can be facilitated by the success of early generation hybrids, negative selection against less fit later generation hybrids may be delayed. The prevalence of back-crossing in such populations, as is the case here (Kirk et al., 2004), might provide the opportunity for introgression to occur even if the average fitness of natural hybrids is low. A previous study (Kirk et al., 2004) has shown evidence for introgression of chemical expression specific to S. aquaticus into S. jacobaea in the Zwanenwater area, which may suggest that hybrids don't have to be superior to make an evolutionary contribution.
The authors thank Karin van Veen, Henk Nell, Margriet Peet, Jung van der Meulen and Coen Bruin for technical assistance, and two anonymous reviewers for useful comments on the manuscript. Heather Kirk thanks the Natural Sciences and Engineering Research Council of Canada (NSERC) for financially supporting her research.