• Field reciprocal transplantations of two metallicolous populations (Mpops) and two nonmetallicolous populations (NMpops) of Thlaspi caerulescens were performed here to determine the pattern of local adaptation and to assess the cost of adaptation of Mpops to a metalliferous environment (Menv). The role of herbivores as an important selective pressure in the nonmetalliferous environment (NMenv) was also examined.
• Growth, survival, fitness, life cycle and herbivore consumption were monitored for each transplant for 2 yr.
• Local adaptation of Mpops to their own environment was clearly demonstrated, as Mpops consistently outperformed NMpops in Menv. In NMenv, no advantage of NMpops over Mpops was detected. However, the fitness of Mpops was generally lower in NMenv than in Menv. Herbivore consumption appeared to be a significant selective pressure for Mpops in NMenv.
• An imbalance of selective forces between Menv and NMenv probably explains the greater local adaptation of Mpops. Therefore, colonization of NMenv by Mpops appears possible. Although Mpops were able to survive and reproduce in NMenv, they nevertheless expressed a cost attributable in part to their higher susceptibility to herbivores.
Plant species that colonize contrasting environments can develop genetically divergent populations, particularly if gene flow between populations is limited. Genetic differentiation between populations may be the result of natural selection or/and founder effect (Lefèbvre & Vernet, 1990; Linhart & Grant, 1996; Jakobsson & Dinnetz, 2005). To determine whether genetic differentiation is the result of natural selection, populations must be reciprocally transplanted. Superior performance of native plants compared with nonnative ones provides evidence for local adaptation (Kawecki & Ebert, 2004).
Pseudometallophytes (i.e. species that colonize both metalliferous and nonmetalliferous environments) represent a highly relevant model with which to study local adaptation (Linhart & Grant, 1996). Selection in the metalliferous environment is very intense and leads to the evolution of locally adapted metal-tolerant populations, referred to as the metallicolous ecotype, within a few years (Antonovics et al., 1971; Macnair, 1987; Al-Hiyaly et al., 1988). Evolutionary scenarios of metal tolerance often assume that metallicolous plants are disadvantaged on nonmetalliferous soils, thus explaining their absence on nonpolluted soil (McNeilly, 1968; Cook et al., 1972; Hickey & McNeilly, 1975). This is the so-called ‘cost of tolerance’, for which two hypotheses have been formulated. First, the trade-off hypothesis proposes that resources needed for tolerance processes are diverted away from other traits such as growth rate, seed production or competitive ability (Harper et al., 1997a). Secondly, the metal requirement hypothesis postulates that tolerance mechanisms such as efficient metal sequestration lead to metal deficiency when metal-tolerant plants grow on nonmetalliferous soil (Baker & Walker, 1990; Harper et al., 1997b, 1998). Several studies using genetic lines of Mimulus guttatus selected for contrasting metal tolerance failed to demonstrate any correlation between high tolerance and reduced fitness (Macnair & Watkins, 1983; Harper et al., 1997a,b, 1998). In addition to the cost of metal tolerance, another nonexclusive hypothesis can be developed to explain the failure of metal-tolerant genotypes to spread in nonmetalliferous soil. Indeed, metallicolous plants are adapted not only to high metal contamination, but also to a series of environmental factors specific to the metalliferous environment such as water and nutrient deficiency, high exposition to the sun, low competition and low herbivore pressure (Antonovics et al., 1971; Hickey & McNeilly, 1975; Wright et al., 2006; Noret et al., 2007). Therefore, the absence of metallicolous plants on nonmetalliferous soil would be explained by a general cost of adaptation to the metalliferous environment. This adaptation cost was demonstrated in several pseudometallophyte species by field experiments on normal soil (Cook et al., 1972; Hickey & McNeilly, 1975) and by reciprocal transplantations (Sambatti & Rice, 2006; Wright et al., 2006).
The pseudometallophyte and metal hyperaccumulator Thlaspi caerulescens represents a good model with which to assess this cost. Despite constitutive metal tolerance and accumulation (Meerts & Van Isacker, 1997; Escarréet al., 2000; Assunção et al., 2003), there is some genetic divergence for several traits between the metallicolous ecotype (which develop on calamine metalliferous soils) and the nonmetallicolous ecotype (growing on nonmetalliferous soils) of T. caerulescens. First, a higher zinc (Zn) accumulation capacity and lower metal tolerance were consistently found in nonmetallicolous populations compared with metallicolous populations (Meerts & Van Isacker, 1997; Escarréet al., 2000; Assunção et al., 2003; Dechamps et al., 2007; Jiménez et al., 2007). Secondly, metallicolous populations constitutively produced lower concentrations of defensive compounds (i.e. glucosinolates) against herbivores than nonmetallicolous populations (Noret et al., 2007). Lastly, genetic differences for reproductive traits between metallicolous and nonmetallicolous populations were shown. A higher inbreeding coefficient and pollen:ovule ratio were systematically found in nonmetallicolous populations compared with metallicolous ones, suggesting a higher self-fertility in nonmetallicolous populations (Dubois et al., 2003). Nonmetallicolous populations also tended to delay flowering until the second year, whereas most metallicolous individuals flowered from the first year (Dechamps et al., 2007). Metallicolous plants also showed great plasticity in terms of life cycle, with an annual strategy on nonmetalliferous soil and a short perennial strategy on contaminated soil (Dechamps et al., 2007). All these genetic differences between metallicolous and nonmetallicolous populations suggest that T. caerulescens probably evolved not only under contrasting heavy metal pressures, but also under a series of environmental factors specific to metalliferous and nonmetalliferous environments. In metallicolous populations, the higher metal tolerance, lower metal accumulation capacity, lower level of chemical defence and the particular reproduction strategy are all genetic features that may contribute to the expression of a cost when these plants are transplanted to a nonmetalliferous environment. Field reciprocal transplantations of metallicolous and nonmetallicolous populations are needed to take into account the impact of all environmental components on plant fitness.
In this study, a field reciprocal transplant experiment was conducted on two ecotypes of T. caerulescens. Two metallicolous populations from east Belgium and two nonmetallicolous populations from the Grand Duchy of Luxemburg were used. The following questions were addressed.
• What is the pattern of local adaptation? Does the performance of the metallicolous ecotype in the nonmetalliferous environment reveal a cost of adaptation to the metalliferous environment?
• Does herbivore consumption similarly affect the two ecotypes?
• Are the genetic differences in life cycles observed in the glasshouse maintained in the field?
• Are populations within ecotype locally adapted to their own sites?
Materials and Methods
For each ecotype (the metallicolous ecotype (Mtype) and the nonmetallicolous ecotype (NMtype)) of Thlaspi caerulescens J. & C. Presl (Brassicaceae), two populations were randomly selected. There were two metallicolous populations (Mpops: Prayon (popPR) and Angleur (popANG)) and two nonmetallicolous populations (NMpops: Winseler (popWIN) and Wilwerwiltz (popWIL)) located at two sites (Msites: sitePR and siteANG) in a calamine metalliferous environment (Menv) and at two sites (NMsites: siteWIN and siteWIL) in a nonmetalliferous environment (NMenv). The location and habitat of the four sites are summarized in Table 1. The metalliferous environment is located in the Liège province of Belgium. SitePR has been contaminated for c. 150 yr by dust from a lead-zinc-cadmium (Pb-Zn-Cd) smelter whereas siteANG is a slag heap contaminated with Zn and Pb (waste from a Zn smelter). The nonmetalliferous environment is located in Luxemburg and consists of steep road banks, exposed to the south for siteWIL, and to the north for siteWIN.
Table 1. Characteristics of the two metalliferous (Msites) and the two nonmetalliferous (NMsites) study sites
Metalliferous environment (Menv)
Nonmetalliferous environment (NMenv)
Some data are means (SE). Values sharing an identical superscript letter are not significantly different (P > 0.05, Fisher's post hoc test).
Mean gastropod density (number of gastropods per trap per site)
Seeds were randomly collected from > 20 plants in each population. Seeds were sown in September 2004 in seed trays filled with compost and placed in a glasshouse. Four weeks after emergence (at the four-leaf stage), seedlings were transplanted into the four field sites studied.
In each of the four sites, seedlings from the four populations were planted out in a block design, with 12 blocks (50 × 50 cm) per site. Each block was subdivided into four subplots (25 × 25 cm) for each of the four origins. A planting guide was used to make holes in a 3 × 3 grid within each subplot, which thus received nine seedlings. Before planting out, resident vegetation was mown to limit competition, and roots of seedlings were relieved of compost balls. Seedlings were watered once just after planting out.
Field data collection
Vegetative growth was assessed from the number of leaves after 6 months. Survival and reproductive parameters were monitored for each plant over its lifetime. As reproductive parameters, we collected the following data: presence or lack of flowering (1–0) and, in flowering plants, the number of fertile fruits and the life cycle (flowering in year 1 or/and year 2). The net reproductive rate (R0) was used as a fitness estimate as in Dechamps et al. (2007): R0 = Σ lxmx (lx, probability of surviving from birth to reproductive age x; mx, fecundity at age x). We used the number of fertile fruits as a fecundity estimator. Flowering axes were harvested before complete maturation of fruits to avoid seed dispersal. Herbivore damage was visually estimated on each transplant from photographs taken 4 months after planting out according to an ordinal nonproportional scale (Noret et al., 2007). Each transplant was assigned to one of four damage classes according to its level of consumption (class 1, no visible damage; class 2, leaf area removed < 5%; class 3, leaf area removed between 5 and 25%; class 4, leaf area removed between 26 and 100%). At the end of the experiment, some blocks were excluded from the analysis because of disturbance by animals and/or mowing of flowering axes. The number of remaining blocks at the end of the experiment was 11 at sitePR, nine at siteANG, seven at siteWIN and 11 at siteWIL.
Characterization of soils
Twelve soil samples were collected at each site, adjacent to the 12 blocks (depth 0–10 cm). The proportion of stones (i.e. particles > 2 mm) was determined by weighing. pH was measured with a glass electrode on a water-saturated soil sample. Mineral elements were extracted with 1 N ammonium acetate-EDTA (pH 4.65) for 30 min (10 g dry soil in 50 ml) (Cottenie et al., 1982). The supernatant was filtered and analysed by inductively coupled plasma-optical emission spectrometers (ICP-OES) (Vista MPX, Varian Inc., Palo Alto, CA, USA) for the following elements: Zn, Cd, Pb, copper (Cu), magnesium (Mg), calcium (Ca), potassium (K) and phosphorus (P).
The data corresponding to the number of leaves, the number of fertile fruits per flowering plant and the fitness were analysed by ANOVA after logarithmic transformation (using statistica 7; Statsoft, 2005). The factors tested were (1) environment (Menv and NMenv), (2) site (sitePR, siteANG, siteWIN and siteWIL), (3) ecotype (Mtype and NMtype), and (4) population (popPR, popANG, popWIN and popWIL). Sites were nested within environments and populations were nested within ecotypes. Environment and ecotype were considered as fixed factors whereas site (environment) and population (ecotype) were treated as random factors (Kawecki & Ebert, 2004). We could not test block effect because, within some of the blocks, no plants or only one plant survived. Fisher's post hoc test was used to compare means. Differences in the proportion of different life cycle classes were compared with χ2 tests. Survival curves were analysed with a specific generalized linear regression (a parametric log-linear hazard function) using the genmod procedure of SAS (1999) (Egli & Schmid, 2001). The ordinal herbivory data formed a contingency table, which was analysed with a log-linear model using the catmod procedure of SAS (1999). The impact of herbivory level on plant fitness was analysed by linear regression (statistica 7; Statsoft, 2005). For the genmod and catmod procedures, the classification factors were the same as those used in the ANOVA but all factors were considered as fixed.
Local adaptation was explored at two levels: environment and site within environment. To demonstrate local adaptation, we used the ‘local vs foreign’ criterion (Kawecki & Ebert, 2004). Local adaptation is shown if the local ecotype (or population) outperforms the foreign ecotype (or population) in its home environment (or site). Results were also analysed using the ‘home vs away’ criterion, which compares fitness of ecotypes (or populations) across environments (or sites). Each ecotype (or population) is expected to show higher fitness in its own environment (or site) (at home) than in others (away). This last criterion has the disadvantage of confounding the effect of divergent selection with intrinsic differences in habitat quality, but is still informative.
Compared with soils in NMenv, soils in Menv were characterized by higher concentrations of metals (Zn, Cd, Pb and Cu) and a lower percentage of stones (Table 1). Large variation was also found between Msites (Table 1): soil from siteANG contained more Zn, but less Cd and Cu, than soil from sitePR. Compared with soil from sitePR, soil from siteANG was also characterized by higher levels of Mg, Ca and P. Within both Msites, large spatial variation was observed for Zn, Cd and Pb concentrations (sitePR, Zn = 1000–45 000 mg kg−1, Cd = 30–200 mg kg−1 and Pb = 600–4000 mg kg−1; siteANG, Zn = 10 000–55 000 mg kg−1, Cd = 10–50 mg kg−1, and Pb = 250–5000 mg kg−1). In contrast to the Msites, the two NMsites did not differ for soil parameters from each other (Table 1). It should be noted that siteANG had consistently more P and higher pH than all other sites.
Transplants survived significantly better in NMenv (after 1 yr, 30% vs 10% in Menv; Fig. 1, Table 2). There was a significant environment × ecotype interaction (Table 2), indicating that the survival of the two ecotypes (Mtype vs NMtype) differed in the two environment types (Menv vs NMenv). Specifically, the average mortality rate of NMpops was much higher than that of Mpops in Menv (from Nov 2004 to May 2005: –50% for NMpops vs –20% for Mpops; Fig. 1). In contrast, at NMsites, no obvious difference was observed in the survival curves of the different populations (Fig. 1), except for popANG transplanted to siteWIN, which tended to have a higher mortality than the three other populations (Fig. 1c). The population factor nested in ecotype, and the other interactions, were not significant (Table 2). All plants were dead 18 months after the beginning of the experiment.
Table 2. Statistical analyses of number of leaves after 6 months, fitness, survival and herbivory for Thlaspi caerulescens
Number of leaves
Number of leaves and fitness were analysed by ANOVAs. Survival was analysed with a specific generalized linear regression. A log-linear model was used for herbivory data. ***, P < 0.001; **, P < 0.01; *, P < 0.05; ns, not significant.
Environment: metalliferous (Menv) and nonmetalliferous (NMenv).
Site: two replicate sites for each environment (Menv, sitePR and siteANG; NMenv, siteWIN and siteWIL).
Ecotype: metallicolous (Mtype) and nonmetallicolous (NMtype)
Population: two replicate populations for each ecotype (Mtype, popPR and popANG; NMtype, popWIN and popWIL).
For ANOVAs, the nested factors site and population were considered to be random. The error terms for the sources were as follows (a, b, c and d are variable factors of linear combinations).
Environment: a × MSsite (env) + b × MSenv × pop (ecotype) – c × MSsite (env) × pop (ecotype)– d × MSresidual.
Ecotype: a × MSpop (ecotype) + b × MSsite (env) × ecotype– c × MSsite (env) × pop (ecotype)– d × MSresidual.
Environment × ecotype: a × MSenv × pop (ecotype) + b × MSsite (env) × ecotype– c × MSsite (env) × pop (ecotype)– d × MSresidual.
Environment × population (ecotype): a × MSsite (env) × pop (ecotype) + b × MSresidual.
Site (environment) × ecotype: a × MSsite (env) × pop (ecotype) + b × MSresidual.
The significant environment × ecotype interaction (Table 2) was attributable to the inhibition of leaf production of NMpops growing in Menv (Fig. 2a). For Mpops, the variation in leaf production between sites within environment type was higher than the variation found between environments. Two extreme sets of data were found in the same environment: the highest leaf production was at siteWIL (14 leaves for popPR and popANG) and the lowest leaf production was at siteWIN (seven leaves for popPR and nine for popANG) (Fig. 2a). This explains the marginally significant site (environment) × ecotype interaction (F2,926 = 6.48, P = 0.052; Table 2).
The ‘local vs foreign’ criterion The nonsignificant environment × ecotype interaction indicates that the native ecotype did not systematically outperform the foreign ecotype. This was particularly true at NMsites, where there was a large variation of fitness between populations within ecotype (Fig. 2b). This accounts for the significant site (environment) × population (ecotype) interaction (Table 2). At both NMsites, popWIL significantly outperformed the other populations (Fig. 2b). By contrast, in Menv both Mpops significantly outperformed NMpops (Fig. 2b). Moreover, at Msites, there was no variation of fitness between populations within ecotype (Fig. 2b).
The ‘home vs away’ criterion Mpops had a much higher fitness in Menv than in NMenv (10 vs 0.5), except for popANG transplanted to siteWIL (popANG in siteWIL: 10.8). One-way ANOVAs for each Mpop showed a significant effect of environment only for popPR (F1,340 = 20.9, P < 0.001). Both Mpops had higher fitness at siteANG compared with sitePR (15 vs 5; Fig. 2b), but this difference was not significant because of the large variation observed among blocks at siteANG. Both NMpops had much lower fitness in Menv than in NMenv (0.2 vs 4 for popWIN; 0.2 vs 25 for popWIL). One-way ANOVAs for each NMpop showed a significant effect of environment on fitness (popWIN: F1,340 = 22.4, P < 0.001; popWIL: F1,340 = 72.3, P < 0.001). Finally, within NMenv, the fitness of both NMpops was higher at siteWIL compared with siteWIN (popWIN: F1,160 = 3.9, P < 0.05; popWIL: F1,160 = 4.7, P < 0.05).
The distribution of flowering plants in the different life cycle classes was as follows: 53% flowered only in the first year (2005), 40% flowered only in the second year (2006) and 7% flowered in both years (2005 and 2006). This distribution was influenced by environment (Menv vs NMenv), ecotype (Mtype vs NMtype) and population within ecotype (popPR vs popANG and popWIN vs popWIL). The proportion of plants that delayed their reproduction to the next year was significantly higher in NMenv compared with Menv (20% vs 60%; χ2 = 33.4, P < 0.001). This trend was observed regardless of the origin of transplants (Fig. 3). Differences in life cycle between ecotypes were only found in NMenv, where NMtype had significantly more plants that reproduced only in the second year (70% for NMtype vs 50% for Mtype; χ2 = 8.9, P < 0.01). In Menv, no difference was observed between ecotypes (data not shown). Populations within ecotype showed variation in life cycle. For Mpops, more plants of popANG flowered in the first year (60% in year 1) compared with popPR (50% in year 1) (χ2 = 4.1, P < 0.05; Fig. 3). In NMpops, popWIL mainly flowered in the first year (65% in year 1) whereas a majority of plants of popWIN flowered in the second year (40% in year 1 vs 58% in year 2) (χ2 = 41.5, P < 0.001; Fig. 3). It is worth noting that not all plants that had initiated flowering produced seeds, because some of them died before fructification (Fig. 3).
There was significantly more herbivore damage in NMenv than in Menv (Table 2, Fig. 4). The proportion of plants with no herbivore damage was c. 50% in Menv and c. 25% in NMenv (Fig. 4). Herbivore consumption also diverged significantly between sites within environments (Table 2), with more herbivore damage for siteWIN in NMenv, and for siteANG in Menv (Fig. 4). At each site, Mpops were consumed to a significantly greater extent than NMpops (Fig. 4, Table 2). Moreover, the significant environment × ecotype interaction is attributable to the fact that the difference in herbivory on Mpops was higher in NMenv than in Menv (Fig. 4, Table 2). Finally, the significant population (ecotype) × environment interaction was explained by the higher herbivore consumption on popANG compared with popPR in NMenv (Fig. 4).
As there is little herbivory at Msites, and as NM plants are not consumed much by herbivores, the correlation between fitness and herbivory was only tested for Mpops in NMenv (i.e. 2 Mpops × 2 NMsites = 4 correlations). A significant negative relation between herbivory level and plant fitness was found for popANG at siteWIL (R2 = 0.10, β = –0.36, P < 0.05, n = 34). For popANG at siteWIN and popPR at siteWIN and siteWIL, there were not enough flowering plants to establish any relations (n = 4, 5 and 7, respectively).
What is the pattern of local adaptation? Does the performance of Mtype in NMenv reveal a cost of adaptation to Menv?
Local adaptation of Mtype to its home environment was clearly demonstrated by the ‘local vs foreign’ criterion. Both Mpops systematically outperformed both NMpops in Menv. Their superiority was expressed in survival, vegetative growth and fitness. This result, although not surprising, provides the first conclusive demonstration in natural conditions that Mpops of T. caerulescens represent a distinct ecotype. This finding corroborates earlier results obtained in controlled conditions (Meerts & Van Isacker, 1997; Escarréet al., 2000; Assunção et al., 2003; Dechamps et al., 2007; Jiménez et al., 2007). Thus, while T. caerulescens certainly shows species constitutive tolerance to several heavy metals, it is clear that the capacity to colonize soils with extremely high concentrations of Pb, Zn and Cd, such as those found at calamine metalliferous sites, requires an enhanced degree of tolerance.
In NMenv, the survival, growth and fitness of Mtype were not lower than those of NMtype. These data failed to demonstrate local adaptation of NMtype to its home environment according to the ‘local vs foreign’ criterion. These results confirm previous observations obtained in controlled conditions where NMpops did not outperform Mpops in noncontaminated soil (Dechamps et al., 2007). However, on the basis of the ‘home vs away’ criterion, Mpops generally expressed lower fitness in NMenv than in Menv. Even if the ‘home vs away’ criterion is sensitive to variation in overall habitat quality across sites (Kawecki & Ebert, 2004), the reduced performance of Mpops in NMenv may actually reveal a cost of adaptation to Menv. The possible role of herbivores as an important selective pressure in NMenv is discussed in the next section. Our results are consistent with those of other transplantation experiments showing this nonreciprocity of local adaptation (e.g. coastal vs inland sites: Nagy & Rice, 1997; high- vs low-altitude sites: Gauthier et al., 1998). In these studies, local adaptation was always stronger in the more stressful environment. The greater local adaptation of Mtype to its environment is probably attributable to an imbalance of selective forces between Menv and NMenv. From an evolutionary perspective, our results suggest that, at this time, colonization of NMenv by Mtype is more probable than colonization of Menv by NMtype. Reciprocal transplant experiments of other pseudometallophyte species also found contrasting patterns of selection in metalliferous and nonmetalliferous environments. Sambatti & Rice (2006) showed that local adaptation of Helianthus exilis was driven by differential survivorship at serpentine sites, whereas selection mainly acted on reproduction at nonserpentine sites. Jurjavcic et al. (2002) demonstrated local adaptation in Vulpia microstachys only for the population from the most polluted and dry site.
It is worth noting that, in the present study, only a part of the cost of adaptation to Menv was tested. Indeed, selection on germination and on early juvenile survival was not measured in our experimental design, and this may have underestimated the cost of adaptation to Menv.
Does herbivore consumption similarly affect the two ecotypes?
The higher herbivory pressure recorded in NMenv compared with Menv, and the higher herbivore damage on Mtype compared with NMtype, were consistent with the results of Noret et al. (2007), which showed that Mpops produce constitutively lower concentrations of defensive compounds than NMpops. A higher palatability of Mpops compared with NMpops was found when both ecotypes were transplanted to NMenv, where gastropod density is higher. Noret et al. (2007) proposed that metallicolous populations have lost efficient defence mechanisms against herbivores because they evolved in an environment with low herbivore pressure (i.e. Menv). Among Mpops, popANG was consumed to the greatest extent by herbivores. In this last population, linear regression on the 34 flowering plants at siteWIL showed that herbivory had a negative effect on plant fitness. This result highlights the potential for herbivores to exert selection on T. caerulescens in NMenv. The negative impact of herbivory on plant fitness has been described in detail elsewhere (e.g. Marquis, 1992; Strauss, 1997). Experiments taking into account other life cycle stages not considered in the present study (e.g. seedling survival and pollen and ovule quality) would permit better characterization of the impact of herbivores on fitness.
Are the genetic differences in life cycles observed in the glasshouse maintained in the field?
Variation in plant life cycle was influenced by both environment and genetic differences between ecotypes and populations. Environmental effects on life cycle were obvious, as the Menv and NMenv can be characterized by one life cycle class. In Menv, most plants flowered from the first year, while in NMenv all populations tended to delay flowering until the second year. In Menv, NMtype suffered from high mortality from the first year, probably as a result of metal toxicity, so that all NM genotypes programmed to flower in year 2 disappeared. Therefore, the high plasticity of life cycle was particularly marked for Mpops, for which no effect of the environment on survival was detected. Such life cycle plasticity of Mpops was also found to be expressed in controlled conditions (Dechamps et al., 2007) where an annual strategy was favoured on normal soil and a short perennial strategy on Zn-contaminated soil. However, in the present field experiment, life cycle plasticity was different as it was the biennial strategy that was favoured on nonmetalliferous soil instead of the annual strategy observed in controlled conditions. The life cycle plasticity of Mpops probably contributes to their ability to colonize heterogeneous habitats. In this study, the heterogeneity of Menv was demonstrated for metal pollution. Personal observations in Belgian metalliferous sites support the idea that spatial heterogeneity affects the life strategy of T. caerulescens: annuals (small flowering rosettes with low fruit production) are found in sunny open zones, while most perennials (large rosettes with high fruit production) grow in the shaded conditions of the underbrush. Interestingly, Dubois (2005) found the same pattern of life cycle variation at Msites in southern France. These observations therefore suggest that life cycle plasticity on metalliferous soils may allow plants to maximize their fitness. Such life cycle plasticity was recently demonstrated in Mimulus guttatus, where a perennial cycle was favoured in a permanently wet habitat and an annual cycle in a intermittently wet habitat (van Kleunen, 2007). Interestingly, while delayed flowering appeared to be a plastic response in Mpops, it seemed to be genetically fixed in NMpops, for which this delay was also found to be expressed in controlled conditions (Dechamps et al., 2007). The environmental constraints prevailing in NMenv probably favour the selection of a biennial life cycle. The road banks on which plants develop are subjected to a disturbance regime as a consequence of intermittent microscrees, which result from the frequent mowing of the steep road banks. This disturbance regime creates intermittent gaps known to favour biennial species (Silvertown & Charlesworth, 2001).
Are populations within ecotype locally adapted to their own sites?
In Menv, no specific adaptation of either population to their home site was detected. Furthermore, higher fitness (not significant) was found at siteANG, which appears to be more favourable for T. caerulescens. Better nutrient status (higher Mg, P and Ca), higher pH and lower Cd concentration may explain why siteANG appears to be more favourable than sitePR. Even the NMpop of siteWIL survived longer at siteANG, suggesting that effective metal toxicity is reduced by the higher soil base status at this site (Simon, 1978). In NMenv, popWIL outperformed popWIN at its home site (siteWIL). However, popWIL also outperformed popWIN at siteWIN. These results are inconclusive concerning site-specific adaptation, and demonstrate a great disparity between the fitnesses of two populations that are ecologically and geographically close to each other (fitness of popWIN < fitness of popWIL). SiteWIN might be considered as ecologically marginal for T. caerulescens. Indeed, it is exposed to the north, while most NMpops in Luxemburg express marked preference for south-facing slopes (Molitor et al., 2005). Moreover, because we collected seed directly in the field (e.g. Nagy & Rice, 1997; Sambatti & Rice, 2006), we cannot exclude a negative effect of maternal environment on seed quality. These maternal effects could interact with genotype and affect fitness (Rossiter, 1996). However, no difference in seed weight was found between the two NMpops (data not shown). Stochastic effects might also explain the systematic inferior performance of popWIN. Indeed, NMpops in Luxemburg probably constitute a metapopulation where the founder effect and genetic drift frequently occur (Molitor et al., 2005). The low population density of T. caerulescens at siteWIN may have favoured inbreeding. Galloway & Fenster (2000) were also unable to demonstrate a specific-site adaptation in a metapopulation of an annual legume.
Our study has demonstrated that the local adaptations of metallicolous and nonmetallicolous ecotypes of T. caerulescens to their own environments are not equal. Selection in the metalliferous environment is so strong and specific that colonization by a foreign genotype would be extremely difficult. In contrast, we found that the nonmetallicolous ecotype did not show fitness superiority compared with the metallicolous ecotype in the nonmetalliferous environment. Therefore, from an evolutionary perspective, colonization of the nonmetalliferous environment by the metallicolous ecotype is certainly more probable than the colonization of the metalliferous environment by the nonmetallicolous ecotype. The plasticity of the life cycle probably plays a crucial role in the ability of the metallicolous ecotype to colonize contrasting environments. However, the metallicolous ecotype showed decreased fitness in the nonmetalliferous environment, which may arguably represent a cost of adaptation to the metalliferous environment. The high herbivore density in the nonmetalliferous environment seems to play an essential role in the expression of this cost. Indeed, metallicolous plants are known to contain lower concentrations of defensive compounds than nonmetallicolous plants, and are for this reason consumed to a greater extent than nonmetallicolous plants when they are transplanted to a nonmetalliferous environment. In this study, we have demonstrated for the first time that herbivory on metallicolous plants reduces their fitness and therefore represents an essential selective pressure for T. caerulescens in a nonmetalliferous environment.
This work was supported by the Fonds de la Recherche Fondamentale Collective (Belgium) (Project FRFC 2.4565.02). NN is a research fellow of the Fonds National de la Recherche Scientifique of Belgium (‘Aspirant du FNRS’).