•The effects of Ni and Mg, two factors involved in the infertility of serpentine soils, were studied in the alpine plant Cerastium alpinum. Root growth of plants from adjacent serpentine and non-serpentine populations in Scandinavia, representing an eastern and western postglacial immigration lineage and the hybrid zone between them, were compared to study the adaptation of C. alpinum populations.
•Seedlings were placed in solutions with low or high concentrations of Ni and Mg in a full factorial experiment according to a randomized block design. The growth of roots was analyzed and discussed in relation to the soil content.
•The serpentine populations showed higher tolerance to Ni and Mg stress than non-serpentine populations. The degree of metal tolerance differed among the serpentine populations and was related to the effective concentrations of Ni and Mg in the soil at each site.
•The results suggest that serpentine tolerance is locally evolved in C. alpinum and that tolerance has arisen in parallel during the postglacial colonization of Scandinavia on serpentine soils with similar composition.
Ultramafic (serpentine) soils are adverse habitats for plants. The serpentine soils are characterized by low concentrations of the plant nutrients N, P, K and Ca and high, potentially toxic, concentrations of Mg and Ni (Brooks, 1987; Proctor, 1999). In addition, they are dry and exposed because of lack of organic material and their granular texture. Accordingly, the reason for the infertility of serpentine soils is a multifactorial phenomenon, sometimes referred to as the serpentine syndrome (Jenny, 1980; Kruckeberg, 1984). Although a range of both chemical and physical factors may influence plant growth in serpentine soils, high concentrations of Mg in relation to Ca, and elevated concentrations of Ni are considered to be important factors affecting plant growth and survival (Proctor, 1971; Marrs & Proctor, 1976; Johnston & Proctor, 1981; Gabbrielli & Pandolfini, 1984; Proctor & Nagy, 1992; Nagy & Proctor, 1997). High concentrations of Mg in relation to Ca may lead to diminished uptake of Ca. In addition, high concentration of Mg per se may have a toxic effect on plant growth (Proctor, 1970). Although Ni is an essential element, the required amounts are minor (Taiz & Zeiger, 1998). Elevated concentrations of Ni can per se effectively inhibit cell division at root meristems in non-tolerant plants (Robertson, 1985). In addition, Ni has a negative effect on photosynthesis, respiration and stomatal regulation of transpiration (Carlson et al., 1975). The role of Ni as a factor explaining infertility of serpentine soils is, however, unclear and contrasting results have been reported on its effect in combination with Mg (Johnston & Proctor, 1981; Gabbrielli & Pandolfini, 1984; Proctor & Baker, 1994). The plant available concentration of Ni also differs among serpentine soils (Proctor, 1971).
Some plants have the genetic resources to grow in serpentine soils and a distinctive flora has evolved on these soils (Rune, 1953; Brooks, 1987; Baker et al., 1992). In Scandinavia, the serpentine flora is characterized by several alpine plants of the family Caryophyllaceae (Rune, 1953; Rune & Westerbergh, 1992). One of the most common of these is a small perennial herb, Cerastium alpinum. This plant grows in serpentine soils in the Scandinavian Mountains, as well as in the boreal region in Finland. Outside serpentine soils, C. alpinum grows on open gravel rich land, alpine heaths, steep slopes and riverbanks (Hultén, 1956; Jonsell, 2001). C. alpinum is polyploid (2n = 9x = 72) and highly variable in morphology. It constitutes a complex with several subspecies and varieties (Hultén, 1956). These taxa have been distinguished on the basis of difference in hairiness, leaf morphology and the presence of subterranean shoots (Rune, 1953; Jonsell, 2001). C. alpinum is hermaphroditic and disperses with seeds formed through selfing or cross-fertilization, or vegetatively through runners.
Analyses of enzyme phenotypes by Nyberg Berglund & Westerbergh (2001) and Nyberg Berglund et al. (2001) suggest that C. alpinum has colonized Scandinavia through two independent immigration events resulting in an eastern and western lineage. These two genetic lineages seem to come into contact in a hybrid zone in northern Scandinavia. The C. alpinum populations on and off serpentine present an excellent opportunity to study the mode of serpentine tolerance in different genetic backgrounds.
Plant species have different abilities to evolve serpentine tolerant populations depending on their genetic resources. If all individuals of the two immigrating lineages of C. alpinum had the genetic prerequisites to colonize serpentine, serpentine and non-serpentine populations would respond to serpentine stress in a similar way. Alternatively, the immigrating populations may have had genetic variability in tolerance so that tolerant individuals would have been selected upon the colonization of each serpentine site. We have studied the response to serpentine stress in serpentine and non-serpentine populations in the two immigrating lineages, as well as in the presumed hybrid zone. Among the factors affecting plant growth in serpentine soils we have focused on Ni and Mg and investigated the individual and combined effects of these metals in nutrient solution culture experiments. The response to metal stress for a population is discussed in relation to the composition of the soil at each site. We address the following questions: Is there evidence for local adaptation in C. alpinum? If so, do the serpentine populations differ in the response to metal stress? Has the tolerance to Ni and Mg arisen in parallel in the two independent immigrating lineages?
Materials and Methods
Seed sources and soil analyses
Seeds from open pollinated C. alpinum L. plants in one serpentine (S) and one adjacent non-serpentine (N-S) population in three regions (A–C) were collected in 1999: (A) Kittelfjäll village (1 S) and Kittelfjället (2 N-S) in western Sweden, (B) Arnesfjellet-Rånafältet (3 S) in northern Norway and Björkliden (4 N-S) in northern Sweden, and (C) Tarpomapää (5 S) and Sodankylä (6 N-S) in northern Finland (Fig. 1). Region A represents the western immigrating lineage and region C the eastern. Region B represents the proposed hybrid zone between those lineages. The distances between the adjacent serpentine and non-serpentine populations range from 2 km to 50 km. Populations 1 to 4 are located in the alpine region, whereas populations 5 and 6 grow in the boreal region. Population 3 S is located close to an open-cast mine and population 6 N-S grows along a river. Large genetic distances have been found between the three population-pairs as well as between the adjacent populations within each pair (Nyberg Berglund & Westerbergh, 2001).
Ten soil samples from the rhizosphere of C. alpinum (0–15 cm) within each of the three serpentine populations and three soil samples within each of the three non-serpentine populations were collected 10–20 meters apart along a transect. Concentrations of metals and nutrients in these soil samples have been analysed by the Environmental Research Laboratory, Swedish Agricultural University, Umeå, Sweden, according to the following methods: exchangeable cations (Ca, K, Mg, Na) were extracted with 1 m ammonium acetate (CH3COONH4) at pH 7.0 and analysed with ICP/AES (Plasma 2000, Perkin Elmer, Norwalk, Connecticut, USA). Heavy metals (Cd, Co, Cr, Mn, Ni and Pb) were extracted with 0.005 m diethylenetriaminepentaacetic acid (DTPA) at pH 7.3 and analyzed with ICP/AES. These methods were used to estimate plant available concentrations. Ammonium and nitrate were extracted with 2 m KCl, and phosphate by ion exchange with 0.5 m HCl eluation. The elements were analysed with FIA (Tectator 5012, Foss Tectator, Sollentuna, Sweden). The pH was measured by pH meter-MP220 (Mettler-Toledo AB, Stockholm, Sweden) with 2 m KCl extracts as well as water.
Growth experiments in nickel and magnesium solutions
Seeds from the six populations were kept on constantly moist filter paper during a period of 2 wk. The germination rate ranged from 55% to 80% among different mother plants. There was no evident difference among populations.
Seedlings with developed cotyledons were coated with a soft plastic stripe for protection and placed in a floating hexagon of cellular plastic (diameter 11.5 cm, height 0.5 cm). A 1.8 cm incision was made at each corner and one seedling from each of the six populations was placed in one of the incisions. The distance between the seedlings was 4.5 cm. The hexagons were placed in 25 l plastic trays with a 1/10 dilution of a standard nutrient solution described in Huss-Danell (1978). The seedlings were grown in the nutrient solution for 1 wk before the start of the experiments. The nutrient solution used in the experiments was modified so that EDTA was exchanged for HBED as a Fe-chelator to avoid the formation of EDTA complexes with Ni in the experiments and Fe(NO3)3 was used instead of FeSO4 as an iron source (Chaney, 1988). At the start of the experiments the hexagons were transferred to 1.3 l cylindrical plastic pots coated with aluminium foil to protect the roots from light. The nutrient solutions were constantly aerated by air pumps (Tetra Whisper 6000, Tetra GmbH, Melle, Germany) and changed to fresh solutions at days 3 and 6. The pH was approx. 6.5 during the whole experiment. The experiments were performed in a laboratory at 20°C with 16 h light and 8 h darkness, provided by 400 W sodium vapour lamps. The light intensity at the plant level varied from 50 to 80 mol m−2 s−1 photons.
Experimental design To test the individual and combined effects of Ni and Mg among C. alpinum populations on different soil types and from three different regions we used a randomized block design of a full factorial experiment with one low (–) and one high (+) concentration of each metal. Based on a preliminary study the concentrations of Ni and Mg were set to 0 m and 15 m, and 0.2 mm and 5 mm, respectively. The experiment was arranged according to a split plot design in nine blocks. Each block consisted of four experimental units, one placed in each of the four treatments: the control (Ni−/Mg−), Ni (Ni+/Mg−), Mg (Ni−/Mg+) and Ni/Mg (Ni+/Mg+). Each experimental unit consisted of six seedlings, one seedling from each of the six populations. Offspring from the same mother plants were used in the four experimental units in each block so that a block consisted of, for each population, four offspring from the same mother plant, with one offspring in each treatment. This gave a total of 216 seedlings distributed among 36 experimental units.
Measurements Root elongation is particularly sensitive to the presence of metals (Baker & Walker, 1990) so that the tolerant plants grow longer roots than sensitive plants in metal solutions. Measurement of the longest root growth, originally suggested by Wilkins (1957), is the classical way to study heavy metal tolerance. The longest root was measured manually with a ruler on two occasions: at day 1 (the start of the experiment) and at day 10. In addition, we scanned all roots at day 1 and day 10 to quantify the formation of lateral roots and the total root growth. The number of lateral roots was manually scored from the images. The image analysis of the total root growth was performed with the computer program WinRHIZO version 2002c (Régents Instruments Inc., Quebec, Canada). Root and background analysis was based on grey level thresholds, which were manually adjusted for each image. The total root length was measured with Regent's nonstatistical method. In this method the length is measured by scanning all pixels in the image of the root skeleton. To each pixel a length, which takes into account the direction of the displacement between pixels, is added (Arsenault et al., 1995; Guay & Arsenault, 1996).
The metal tolerance of a plant population is usually measured as a proportional increase or decrease of root growth in comparison with growth in parallel controls. If plants from serpentine and non-serpentine populations differ in growth in the control solution but not in the metal solution, the proportional growth tends to indicate a higher tolerance of the population with lower growth rate compared to faster growing populations (Macnair, 1990, 1993; Bannister & Woodman, 1992). The use of the absolute growth measurements will consider these populations equally tolerant. When the growth rate does not differ among populations, the absolute and proportional growth measurements will give the same conclusions. In a situation where the growth strategy differs among populations it will be of major importance to distinguish between these two viewpoints. To study the variation in growth strategy in C. alpinum we used one-way anova to compare the root growth of the six populations in the control solution. The growth of the longest root and the formation of lateral roots did not differ significantly among populations during the experiment, whereas the populations showed a difference in total root growth (Fig. 3). As the populations differed in total root growth we compared the results based on absolute and proportional measurements for that response. The use of the latter increased the differences between serpentine and non-serpentine populations but showed the same trend in response as found with the absolute measurements that are presented below.
The data and residuals were checked for deviations from normality using the Kolmogorov-Smirnov (K-S) test, and nonnormally distributed data were transformed to improve normality. Data for which the distribution could not be stabilized with transformation, nonparametric statistics were used. The soil analysis data was explored with principal component analysis (PCA) on the correlation matrix. Differences in concentrations of soil elements among sites were analysed with one-way anova for elements with normally distributed residuals. Elements with nonnormally distributed residuals were analyzed with the nonparametric Kruskal–Wallis test. The root growth experiment was analyzed with a general linear model including block as a random factor, soil type (serpentine or non-serpentine), region (A, B and C), Mg and Ni as fixed factors and all their interactions. Separate analyses were performed on log-transformed data of the longest root and the total root growth, and square root transformed data of the lateral root formation. The analyses were done with MINITAB release 13.0 statistical software (Minitab Ltd., Coventry, UK).
Soil sample analyses
The analyses of plant available amounts of elements in field-collected soil samples showed a large variability in the concentrations of metals and nutrients within and among sites (Table 1). The soil analysis highlighted a number of typical differences between serpentine and non-serpentine soils. The serpentine soils had, in general, very low concentrations of Ca and K but much higher concentrations of Mg and Ni than the non-serpentine soils. The average concentration of Ca in relation to Mg ranged from 1.4 to 8.0 in the non-serpentine soils, whereas the ratio was not higher than 0.3 in the serpentine soils. The plant available amounts of Cd, Cr, Pb, Na and NH4 were under the detection level in most samples. The average pH (extracted with KCl) ranged from 5.4 to 6.7 in the serpentine soils and from 4.0 to 4.3 in the non-serpentine soils. All of the serpentine sites showed within site patchiness in Ni and Mg concentration. Soil samples collected within 10–20 meters could vary fourfold in Ni concentration and threefold in Mg concentration (data not shown).
Table 1. Characterization of soils from the collection sites in Scandanavia
1 (n = 10)
3 (n = 10)
5 (n = 10)
2 (n = 3)
4 (n = 3)
6 (n = 3)
Mean (95% confidence interval) for the elemental concentrations of plant available amounts (mg kg−1 dry soil) in n samples from the serpentine (1, 3 and 5) and non-serpentine (2, 4 and 6) populations. Different superscripts indicate significant differences (Fisher's PLSD test, α= 0.05, for Ca, K, Mg, Mn, Ni and PO4 and Kruskal–Wallis for the other elements). The relation between Ca and Mg is given as the Ca : Mg ratio.
The principal component analysis (PCA) of all elements over the detection level (Cd, Cr, Pb, Na and NH4 excluded) showed that the variation in soil composition is mainly explained by three principal components that together describe 77.8% of the variance. The first component, which explains most of the variation (47.6%), is described mainly by the differences in soil type. The second component, explaining 18.1% of the variation, is described by population differences (Fig. 2). Most samples from the serpentine populations cluster to the left and all non-serpentine samples to the right in the score plot. Many samples from the serpentine populations 1 S and 5 S overlap, while the samples from 3 S form a separate cluster in between the samples from 1 S and 5 S and the samples from the non-serpentine populations. Most soil samples from the non-serpentine populations 2 N-S and 4 N-S cluster together while the samples from 6 N-S are more scattered in the score plot.
The soil composition of 3 S deviated in several aspects from the soil composition at the other two serpentine sites. Population 3 S had approx. half the concentration of plant available Mg and twice the concentration of Ca. Accordingly, 3 S had the highest Ca : Mg ratio of the serpentine populations. In addition, this population had the lowest pH. Population 5 S showed twice the Ni concentration found in the other serpentine populations.
Root growth experiment
The absolute measurements of the growth of the longest root, the total root growth and the formation of lateral roots are shown in Fig. 3. The difference in root growth was analyzed with a general linear model with block, Ni, Mg, soil type and region as main factors (Table 2). There was no significant difference among blocks for the three root growth measurements. However, the main factors Ni, Mg, soil type and region had, in general, a significant effect on the growth. In addition, many of the interactions between these factors showed significant effects. The main factors and interactions did, however, affect the three growth measurements somewhat differently. The significant effects were explored with main effect and interaction plots. Some of these plots are shown below.
Table 2. General linear model anova results showing the effects of soil type (serpentine or non-serpentine), region (A, B or C), Ni and Mg and their interactions on the growth of the longest root, the total root growth and the formation of lateral roots of Cerastium alpinum
To determine whether C. alpinum is locally adapted to serpentine soils (i.e. is there a difference in metal tolerance between serpentine and non-serpentine populations or not?) we focused mainly on the main effect of soil type, and the effect of the interactions soil type × Ni, soil type × Mg and soil type × Ni × Mg.
Soil type as a main factor, had an important effect on all the three growth measurements (Table 2). The serpentine plants showed a higher longest root growth and total root growth than non-serpentine plants (Fig. 3). However, the non-serpentine plants had a higher production of lateral roots than serpentine plants.
The significant soil type × Ni interaction for the total root growth showed that the effect of Ni was less severe for serpentine than for non-serpentine plants (Fig. 4). Ni had no obvious effect on the production of lateral roots for serpentine plants, whereas the production was reduced in non-serpentine plants. The effect of the different treatments on lateral root formation is illustrated in Fig. 8.
The soil type × Mg interaction had a significant effect on all the three growth measurements. Mg did not reduce the growth of the longest root (Fig. 5) or the total root growth (data not shown) of serpentine plants, while it had a clear negative effect on these two growth measurements of non-serpentine plants. On the contrary, Mg stimulated the production of lateral roots of plants in non-serpentine soils but had no such effect on serpentine plants (Figs 5 and 8).
The significant soil type × Ni × Mg interaction for the longest root growth and the total root growth showed that Mg ameliorated the negative effect of Ni more effectively in serpentine than in non-serpentine plants (Fig. 6). These results strongly suggest that C. alpinum is locally adapted.
We are also interested in whether the serpentine populations in the three regions differ in the response to Ni and Mg. We have therefore focused on the significant interactions soil type × region × Mg and soil type × region × Ni × Mg.
The significant soil type × region × Mg interaction for the longest root growth showed that the serpentine population 3 S in region B was inhibited by Mg while the serpentine populations in the other two regions, in particular, 5 S were stimulated (Fig. 7a).
The significant soil type × region × Ni × Mg interaction for the total root growth showed that Mg reduced the negative effect of Ni in the serpentine populations 1 S (region A) and 5 S (region C), whereas Mg enhanced the negative effect in the serpentine population 3 S (region B, Fig. 7b,c). These results show that the serpentine population 3 S differed in the response to Ni and Mg stress compared to the other two serpentine populations, while 1 S and 5 S responded in a similar way.
Plant adaptation to serpentine stress may occur only in those populations experiencing this stress or may be widespread among populations within a species (Antonovics et al., 1971; Wu, 1990). Whether or not the ability to tolerate serpentine stress is restricted to populations that grow on serpentine may have implications for questions dealing with the adaptation and evolution of stress tolerance, and its genetic control. The two lineages of C. alpinum in Scandinavia provide an excellent opportunity to study the mode of serpentine tolerance in different genetic backgrounds.
We have found significant differences between serpentine and non-serpentine populations in tolerance to Ni and Mg, two important factors for the infertility on serpentine. Plants from serpentine populations showed, in general, a higher tolerance against both Ni and Mg than plants from adjacent non-serpentine populations. This suggests that serpentine tolerance is an adaptive rather than a constitutive trait in C. alpinum. Previous enzymatic studies (Nyberg Berglund & Westerbergh, 2001) have shown that serpentine populations of C. alpinum are genetically more similar to non-serpentine populations within the same geographic region than with distant serpentine populations. This suggests that serpentine tolerant populations have repeatedly evolved within the species. Our findings in the present study of a higher tolerance to elevated concentrations of Ni and Mg in serpentine populations strengthen the evidence for multiple origins of serpentine populations. The serpentine populations differed, however, in metal response.
The serpentine population 1 S in the west and 5 S in the east showed similar responses to Ni and Mg stress, while 3 S in the hybrid zone differed in response. Population 3 S showed the highest tolerance to Ni, whereas the growth was strongly reduced in the other two serpentine populations (Fig. 3). Plants from 3 S responded negatively to Mg, whereas Mg did not have a negative effect on the root growth of plants from 1 S and 5 S compared to the control. The negative effect of Ni found in 1 S and 5 S was reduced in the presence of Mg, whereas Mg enhanced the negative effect of Ni in 3 S (Figs 4, 7b,c).
Why do the serpentine populations show different degrees of tolerance to Ni and Mg stress? The answer can be found in the different compositions of the serpentine soils, so that a higher degree of tolerance to Ni and Mg has evolved in C. alpinum populations from soils that contain higher effective concentrations of these metals. The effective concentration of a metal is a result of interactions between several factors. For example, pH has a crucial role in the concentration of Ni in the soil solution. In well-drained soils with pH higher than 6.0 Ni is unlikely to reach toxic concentrations (Jenne, 1968). The toxic effect of Mg is enhanced with relatively low concentration of Ca. The toxicity of Mg depends also on the actual concentrations of both elements (Proctor, 1971).
According to the known interactions among soil factors, we may describe the plant available Ni and Mg concentrations in the three serpentine soils in the following way: although the Ni concentration in the soil samples of 5 S is twice the concentration in the soil of 3 S the highest plant available concentration of Ni is most likely to be found in population 3 S. First, pH is > 1 unit lower in 3 S than in the other two serpentine populations. This will increase the solubility of the Ni present in the soil particles. Second, the average total cation concentration is only approx. two thirds of those in 1 S and 5 S, which may also increase the relative activity of Ni in the soil solution in 3 S. The relatively high pH in 1 S and 5 S reduces the toxic effect of Ni so that the plants are not adapted to severe Ni stress. According to the high Mg concentrations found in the soil of 1 S and 5 S these populations are exposed to severe Mg stress (Table 1). The Mg toxicity is less severe in 3 S because of the lower Mg concentrations per se and higher Ca concentrations, ameliorating the toxic effect of Mg. In the growth experiments, plants from 1 S and 5 S grew as well in solutions with elevated concentrations of Mg as in the control. However, plants from 3 S were clearly inhibited by the high Mg concentration. Correlation between metal tolerance in a population and the metal concentration of the soil has been shown in several classical studies on the grass Agrostis tenuis (e.g. McNeilly & Bradshaw, 1968).
The fact that Mg ameliorates the effect of Ni in two of the serpentine populations (1 S and 5 S), where the Ni tolerance is rather low, but not in the serpentine population in the hybrid zone (3 S), where the Ni tolerance is much higher, suggests that these C. alpinum populations achieve their tolerance to serpentine somewhat differently. Magnesium has been reported to enhance the toxicity of Ni in the wild grass, Festuca rubra (Johnston & Proctor, 1981), contrary to this, it has been shown to ameliorate the toxic effect of Ni in oat (Avena sativa; Proctor & MacGowan, 1976) and Alyssum bertolonii (Brassicaceae; Gabbrielli & Pandolfini, 1984).
As the difference in effective concentration of Ni and Mg in the serpentine soils is reflected in the different degrees of tolerance in the populations there seems to be a difference in intensities of selection among serpentine sites. However, the serpentine populations 1 S and 5 S in the two independent immigrating lineages grow in soils with similar intensities of selection and showed similar response to Ni and Mg stress. This suggests that Ni and Mg tolerance have evolved in parallel in the two genetic lineages of C. alpinum during the postglacial colonization of Scandinavia. We therefore propose that C. alpinum is a case for parallel evolution of multiple adaptive traits. Parallel evolution of metal tolerance has been suggested for Silene vulgaris (Caryophyllaceae; Schat et al., 1996) and parallel evolution of other physiological traits have recently been shown in Lasthenia californica (Asteraceae; Rajakaruna et al., 2003).
Repeated evolution of metal tolerance has occurred in several grasses such as Agrostis tenuis (Nichols & McNeilly, 1982) and A. capillaris (Al-Hiyaly et al., 1988) and in dicots such as Mimulus guttatus (Macnair, 1983) and Silene vulgaris (Schat et al., 1996). These examples suggest that the evolution of ecotypes from generalist-to-specialist is a common pathway. Plant species that are found in metalliferious soils possess, in general, low frequencies of tolerant individuals in non-metalliferious populations (Gartside & McNeilly, 1974; Ingram, 1988). In some other cases the tolerance has been shown to be a constitutive trait. No significant difference in Ni tolerance between populations on and off serpentine has been reported in a couple of Caryophyllaceae plants, Silene dioica in Sweden (Westerbergh, 1994) and Cerastium fontanum in Scotland (Nagy & Proctor, 1997). In addition, the Ni hyperaccumulator Thlaspi montanum var. montanum (Brassicaceae) was shown to be constitutively adapted (Boyd & Martens, 1998).
Studies of serpentine tolerance have often involved measurements of tolerance as a response on total root mass or growth of the longest root. However, the elongation of main roots, initiation and extension of lateral roots reacts differently to various environmental conditions (Waisel & Eshel, 2002). In the present study, C. alpinum plants from non-serpentine populations produced significantly more lateral roots than Mg tolerant serpentine plants in Mg stress. The higher production of lateral roots in non-serpentine plants could be a secondary effect when the growth of the longest root is inhibited. Alternatively, the serpentine and non-serpentine populations have evolved different growth strategies. We found evidence supporting the latter hypothesis. First, negative correlation between the growth of the longest root and the production of lateral roots as expected for non-serpentine plants under the secondary effect hypothesis was not found (data not shown). Second, the non-serpentine population 4 N-S, which grows in a soil with almost twice the Mg concentration of the two other non-serpentine sites (Table 1), produced less lateral roots under Mg stress than 2 N-S and 6 N-S even though the non-serpentine populations showed similar growth of the longest root (Fig. 3). Non-serpentine plants may have evolved a growth strategy where an increased formation of lateral roots is a result of a higher concurrence and limited resources (e.g. Mg) on non-serpentine soils. The serpentine populations may have evolved a strategy, where the formation of lateral roots is down regulated in high concentrations of Mg and more resources are allocated to longer and deep growing roots, which are important in dry serpentine soils.
Serpentine tolerance is easily evolved in Cerastium. In Europe, at least 15 Cerastium species have evolved serpentine populations (Brooks, 1987). This suggests that the ability to evolve serpentine tolerance has occurred in the early stages of the evolution of Cerastium and in the Caryophyllaceae, in general, as this phenomenon is found in several related genera in the plant family such as Silene, Lychnis and Arenaria (Rune, 1953; Rune & Westerbergh, 1992). What makes Cerastium plants such successful colonizers of serpentine? Besides Ni and Mg, drought tolerance is considered to be a major determinant for serpentine adaptation and evolution. In fact, drought tolerance is suggested to be the main driving force for serpentine tolerance in Mimulus (Gardner & Macnair, 2000; Hughes et al., 2001). Tolerance to drought is a common feature in the genus Cerastium as many species grow in dry and gravel rich soils in Europe (Jonsell, 2001). As serpentine tolerance is likely to depend on an interaction of many genes it is of great importance for the survival of the plant population to maintain blocks of adapted genes from one generation to another. The fact that the majority of the Cerastium species can spread vegetatively and produce seeds through selfing could be an additional explanation for the successful colonization of serpentine. All of the serpentine Cerastium species are polyploids with chromosome numbers ranging from 2n = 36 to 2n = 144 (Jalas & Suominen, 1988). Like most polyploids they are likely to have evolved through interspecific hybridization (allopolyploidization). For several alpine polyploid species such as Draba (Brassicaceae), repeated polyploidization events have been shown to occur within relatively small regions and short time scales (Brochmann et al., 1992). Possibly, the genetic effects of allopolyploidization as well as repeated polyploidization events within a species have increased the number of available pathways for the evolution of serpentine tolerance.
The authors thank O. Rune for introducing us to the serpentine flora of Scandinavia. Thanks to J. Westerbergh for being so generous with his statistical knowledge, R. Chaney for advice concerning Fe-chelators, U. Gullberg for discussions of experimental designs and A. Saura for his interest and support in this work. This study was supported by Mid Sweden University, the J C Kempes memorial fund, the Royal Swedish Academy of Sciences (Krooks donation, J A Wahlberg memorial fund and Hierta-Retzius fund) to A-BNB and the Swedish Council for Forestry and Agricultural Research to AW.