Environmental control and intersite variations of phenolics in Betula nana in tundra ecosystems

Authors


Summary

  •  Secondary metabolites make leaves unpalatable for herbivores and influence decomposition. Site-specific differences are presented in phenolics and nitrogen in Betula nana leaves from dwarf shrub tundra at Abisko, northern Sweden, and from tussock tundra at Toolik Lake, Alaska, subjected to a decade of warming, fertilization and shading.
  •  Nitrogen and a number of phenolics, including condensed and hydrolysable tannins, flavonoids, phenolic glucosides and chlorogenic acids, were analysed in B. nana leaves.
  •  Phenolic concentrations showed marked between-site differences (e.g. condensed tannins were 50% higher at Abisko than at Toolik); responses to the environmental manipulations were more pronounced at Toolik compared with Abisko. Warming increased condensed tannins and decreased hydrolysable tannins at Toolik, but had no effect at Abisko, whereas fertilization and shading generally decreased concentrations.
  • Betula invests less carbon in phenolics at Toolik than at Abisko and shows a greater response to environmental changes by investing more carbon in growth and less to phenolic production. Hence, the Toolik population has a lower herbivore-defense level, which declines further if nutrient availability increases. By contrast, under warmer conditions, the increase in bulk phenolics and decrease in leaf palatability are greater at Toolik than at Abisko.

Introduction

Carbon (C) based secondary compounds, such as phenolics and terpenes, constitute a considerable part of the dry weight and C content of plant tissues. For instance, tannin concentrations in woody plants often exceed 10% of the d. wt (Roth & Lindroth, 1994; Gebauer et al., 1998; Heyworth et al., 1998). This suggests that the allocation of C to synthesis of C-based secondary compounds (CBSC) diverts part of the C from being used for plant growth.

It has been suggested that the secondary compounds in plants act as grazer deterrent substances that have evolved as a response against herbivory (Feeny, 1976; Rhoades & Cates, 1976; Berryman, 1988), but see Clausen et al. (1992) and Ayres et al. (1997). Their effects on the herbivores can be toxic or, as for tannins, they can act through their ability to bind proteins and prevent the herbivore from assimilating nitrogen (N) from the forage (Feeny, 1976; Rhoades & Cates, 1976). In the last case, their action is strongly dosage-dependent, which may explain their high concentrations in plant tissue and implies that the concentrations should vary with grazing pressure. However, it has also been suggested (Bryant et al., 1983) that the varying levels are reflections of other metabolic functions in the plants that control C allocation to growth, for example the ability of plants to respond by increased growth at different C to nutrient (principally N) levels. If other resources than fixed C limit plant production, growth is kept back resulting in high C levels and the excess C is shunted to synthesis of secondary substances (Bryant et al., 1983). A similar mechanism could explain observed increase in CBSCs of defoliated or grazed plants, given that defoliation reduces the nutrient status of the plant proportionally more than the C status and thereby reduces the potential for growth (Tuomi et al., 1984). The hypotheses above are based on the plants’ responses to changed relationship between incorporation of C through photosynthesis and subsequently allocation to growth. A ‘family’ of related hypotheses has been built up, which all basically relate the responses in CBSC to changes in the C sink to source strength (Peñuelas & Estiarte, 1998).

We analysed a number of CBSCs in dwarf birch (Betula nana) leaves collected within two distant tundra regions in northern Sweden (Abisko) and Alaska (Toolik Lake). The Toolik Lake material was collected at two neighbouring sites with contrasting soil types. At the time of sampling, the Abisko site and one of the Toolik Lake sites had been subjected to experimental manipulation of temperature, nutrient levels and light for 9 and 10 yr, respectively.

As the plant material taxonomically belongs to the same species we, firstly, did not expect any major difference in the range of substances present in the Abisko and Toolik Lake plant material. Secondly, based on the various source-to-sink hypotheses, we expected to find treatment effects that could be related to a change of C source to sink strength in the plants of the manipulated plots. Hence, we expected that shading and fertilizer addition should decrease CBSCs because of a reduced photosynthetic C incorporation after shading and because of increased trade-off of C to growth after fertilization. The warming effect could either increase or decrease CBSCs depending on whether temperature affected the synthesis only or if it also increased growth. Thirdly, we knew from previous studies that the plants from Toolik Lake respond strongly to nutrient amendment by increased biomass production (Chapin et al., 1995), while the Abisko plants show little or no response (Graglia et al., 1997). Hence, we expected a stronger response to the manipulations in the plants from Toolik Lake than in the Abisko plants, particularly after fertilizer addition. Finally, due to known differences in soil pH and nutrient regimes between the two sites at Toolik Lake (Gough et al., 2000; IK Schmidt, unpublished), we expected large differences also in the quantities of CBSCs between the sites.

Our aim was to examine if the site-specific C allocation patterns to growth in response to the environmental manipulations were mirrored in the allocation of CBSC. That is, we wanted to get more detailed information on how the C was used in the plants than we had from our previous measurements of primary production or biomass formation. Also, several of our manipulations closely simulate predicted environmental changes in the Arctic, and they are unique because of the long period they have been in place. We anticipate therefore that the responses we observed could serve as a reliable basis for evaluations of climate change impact on forage quality. Further, changes in plant CBSCs caused by environmental changes are likely to affect the rate of decomposition of organic material (Gallet & Lebreton, 1995; Verhoeven & Toth, 1995; Nilsson et al., 1998), and eventually ecosystem processes in general (Wardle & Lavelle, 1997). As such, we also anticipate that our results will be valuable for future studies with a more explicit aim at investigating the impact of climatic change on herbivory and decomposition processes.

Materials and Methods

Site descriptions

The study was conducted in the summer of 1997. The site at Abisko, Sweden, was a subarctic dwarf shrub heath dominated by ericoids. Betula nana L. forms part of the up to 40 cm tall overstorey, and makes up approximately 10% of the above-ground biomass. The experimental site is located near Abisko Scientific Research Station (68°21′ N, 18°49′ E) at 450 m above sea level. Havström et al. (1993), Jonasson et al. (1993), Michelsen et al. (1996) and Graglia et al. (1997) have described the site in detail.

The environmentally manipulated site at Toolik Lake, Alaska, was an acidic tussock tundra equally dominated by graminoids, evergreens and deciduous shrubs and mosses described by Shaver & Chapin (1991), Chapin et al. (1995) and by Chapin & Shaver (1996). The other site at Toolik Lake was a nonacidic tussock tundra with a pH of 7.5, NH4+-N content of 0.36 g m−2 and a PO4-P content of 0.13 g m−2. In comparison, the acidic tundra had a pH of 5.0 and a NH4+-N and PO4-P content of 0.16 g m−2 and 0.08 g m−2, respectively (Gough et al., 2000; IK Schmidt, unpublished) and with different composition of subdominant species (Walker et al., 1989). Both sites are located at the arctic Long-term Ecological Research (LTER) site in the northern foothills of Brooks Range of arctic Alaska, 760 m above sea level (68°38′ N, 149°34′ W).

The Abisko experiment

The experimental set-up at the Abisko site was initiated in 1989, that is nine growing seasons before the sampling. The experiment is still in progress and consists of eight treatments replicated across six blocks, hence, making up a total of 48 plots. Each block contains the following treatments: (1) control; (2) shading; (3) low temperature enhancement; (4) high temperature enhancement; (5) fertilizer addition; (6) fertilizer addition combined with shading; (7) fertilizer addition combined with low temperature enhancement; and (8) fertilizer addition combined with high temperature enhancement. The treatments within each block were randomized.

Each year from early June until the end of August or early September, the temperature and light conditions are manipulated by transparent 0.05 mm polyethylene dome-shaped greenhouses and hessian shading screens, respectively, with an approximate height of 50 cm and a 1.2 m × 1.2 m surface area. Two types of greenhouses are used. One type has a 5–10 cm gap above the ground on two sides (hereafter called t1) that increases the growing season mean air temperature by about 2.5°C. The other type has the plastic fixed more tightly to the ground (called t2), which increases the growing season mean air temperature by about 3.5–4.0°C (Havström et al., 1993; Michelsen et al., 1996). Both types have an open top of 40 × 40 cm. The temperature enhancement increases the soil temperature at about 4 cm depths with 0.4–0.6°C (t1) and 1.2–1.8°C (t2) (Michelsen et al., 1996). The plastic reduces irradiance by about 10% in the PAR range, while the shading cloth reduces irradiance by approximately 60%. This is similar to the reduction in global radiation at ground level below the open canopy in the birch forest of the region. Fertilizer was applied in June 1989 at a rate of 4.9 (N), 1.3 (P) and 6.0 (K) g m−2 in 1989 and from 1990 to 1997 at 10.0, 2.6, and 9.0 g m−2, respectively, except in 1993, when no fertilizer was applied.

The Toolik Lake experiment

The Toolik Lake site was established in 1988 and had been in place for 10 growing seasons before sampling. The experiment is still in progress and consists of six treatments replicated across four blocks, hence, making up a total of 24 plots. Each block contains the following treatments: (1) control; (2) shading; (3) temperature increase; (4) fertilizer addition; (5) fertilizer addition combined with shading; and (6) fertilizer addition combined with temperature increase. The treatments within each block were randomized.

Fertilizer was added each spring just after snowmelt, as NH4NO3 (10 g m−2) and P2O5 (5 g m−2). The Toolik Lake type of greenhouses are built of transparent 0.15-mm plastic sheeting stretched over an A-shaped wooden frame that ranges in height from 10 cm at the side to 1 m in the centre. The greenhouses increase the growing season mean air temperature by 3.5°C and the soil temperature by 2.2°C at 10 cm depths (Chapin et al., 1995). To obtain a shading effect, the wooden frames are covered with cloth, thereby reducing the incoming radiation by 50%. The area covered by one frame is 12 m2, and as at the Abisko site, greenhouses and shading screens are placed each spring just after snowmelt and removed at the end of the growing season in the autumn.

Sampling, preparation and chemical analyses

We collected leaf material from Betula nana plants on 20 June, 20 July and 28 August at Abisko, on 21 July, 28 July and 6 August in the acidic tundra at Toolik Lake, and on 29 July in four randomly chosen nonmanipulated plots in the nonacidic tundra. From each plot in each of the blocks, 10 randomly chosen stems were sampled, making up a total of 150–300 leaves. The leaves of both short and long shoots were separated from the stems, pooled immediately after sampling and dried at room temperature. Drying at room temperature does not affect the birch phenolics (Keinänen & Julkunen-Tiitto, 1998). Once dried, the plant material was ground with pestle and mortar and stored at −18°C until extraction.

Before analysis, a 100-mg subsample of the leaf material was suspended in 10 ml of cold methanol for 20 min. Subsequently, samples were homogenized with an Ultra-Turrax homogeniser for 2 × 3 min. The homogenates were filtered, and the residue washed with 15 ml methanol (Julkunen-Tiitto et al., 1996). The extracts were evaporated using a vacuum evaporator. The samples were redissolved in 15 ml methanol-water (1 : 1) for high performance liquid chromatography (HPLC, HP1050 series). In addition, 22 samples, including at least one sample from controls, warmed, fertilized or shaded plots from each site were analysed for triterpenoids with a gas chromatograph-mass selective detector (GC-MS, HP5971) as trimethylsilyl (TMS) derivatives.

Condensed tannins were analysed colourimetrically using the vanillin:HCl test (Julkunen-Tiitto, 1985) and the acid butanol assay (Porter et al., 1986). Other phenolic compounds: flavonoids (myricetin derivatives: myricetin 3-galactoside and myricetin 3-glucoside and quercetin derivatives: quercetin 3-glucuronide, quercetin 3-glucoside and quercetin 3-arabinoside), phenolic glucosides (salidroside and betuloside), chlorogenic acid and ‘hydrolysable tannins’ (including gallic acid, mono-, di-, tri-, tetra-, and pentagalloylglucose) were analysed with HPLC according to methods described previously by Julkunen-Tiitto (1989) and Julkunen-Tiitto et al. (1996). To separate the phenolics, we used a 60 mm × 4.6 mm internal diameter column filled with HP Hypersil ODS II (3 µm) particles as a stationary phase. In addition, leaf N was analysed with a LECO FP428 N analyser.

Statistical analyses

Inter-site comparisons of differences in plant phenols in birch leaves of control plots between Abisko and Toolik Lake were tested separately for each compound by repeated measures one-way ANOVAs using MANOVA (Table 1). For each compound, intersite differences between the acidic and nonacidic plots at Toolik Lake were tested by one-way ANOVAs on samples collected 28 and 29 July (Table 2).

Table 1. Range of means (mg g−1) across the growing season and one-way repeated measurement ANOVAs testing for differences between Abisko and Toolik Lake (acidic site) controls on the concentration of analysed secondary metabolites and N in leaves of Betula nana
ComponentAbiskoToolikdf F P
  • *

    indicates that data have been log transformed before ANOVA.

Condensed tannin, Vanillin test101.0–120.176.9–90.51 8.60.030
Condensed tannin, Butanol test139.5–152.089.1–103.1117.80.006
Nitrogen* 15.0–21.422.5–27.0122.90.003
Condensed tannin/N*   7.1–9.3 3.4–4.5125.40.002
Hydrolysable tannin   1.5–4.5 1.6–2.91 0.80.410
Myricetin derivatives* 11.9–18.0 7.3–8.61 8.90.024
Quercetin derivatives 14.0–19.526.3–31.3129.40.002
Chlorogenic acid 17.3–23.614.7–24.31 1.40.280
Salidroside   2.2–2.9 0.3–0.4120.30.004
Betuloside   1.4–1.7 0.0–0.5118.00.005
Table 2. Concentration (mg g−1) ±SE of analysed secondary metabolites and N in leaves of Betula nana and one-way ANOVAs testing for differences between sites harvested on 28 and 29 July in two nonmanipulated sites at Toolik Lake
ComponentAcidicNonacidicdf F P
Condensed tannin, Vanillin test 90.3 ± 5.274.9 ± 16.61 0.60.48
Condensed tannin, Butanol test103.1 ± 9.689.4 ± 28.21 0.20.70
Nitrogen 23.0 ± 1.425.2 ± 2.41 0.50.50
Condensed tannin/N 4.5 ± 0.4 4.0 ± 1.71 0.00.83
Hydrolysable tannin 2.9 ± 0.7 3.4 ± 1.11 0.10.73
Myricetin derivatives 7.3 ± 0.5 5.3 ± 1.81 0.90.39
Quercetin derivatives 26.3 ± 3.631.4 ± 0.91 2.50.18
Chlorogenic acid 23.5 ± 1.621.3 ± 5.41 0.10.79
Salidroside 0.2 ± 0.2 1.9 ± 0.4113.20.02
Betuloside 0.3 ± 0.3 0.4 ± 0.41 0.00.90

Treatment effects on each compound (Figs 1–5) were analysed separately for each site with multiway repeated measures ANOVAs. We used repeated measures ANOVAs in order to evaluate across season effects. For both Abisko and Toolik Lake we used four-way repeated measures ANOVAs with shading, fertilizer addition, temperature manipulations and block as main effects, and with the following interactions: fertilizer addition × shading, and fertilizer addition × temperature manipulation. Degrees of freedom (d.f.) for these effects were 1, 1, 2, 5, 1 and 2 for Abisko data (error d.f. = 35), and 1, 1, 1, 3, 1 and 1 for Toolik Lake data (error d.f. = 15), respectively. The possibility of interaction between shading and temperature was not tested in the chosen design. It is biologically less meaningful than the other interactions, because shading counteracts warming by greenhouses. The block factor was included in order to separate effects of possible heterogeneity among the blocks from treatment effects. However, significant block effects generally did not occur and are not shown in the Results section. In some cases data were log transformed to meet assumptions of homogeneity of variance in ANOVAs. All statistical analyses were performed with the General Linear Models (GLM) procedure in SAS (SAS Institute, 1997) and summarized using Type II sums of squares.

Figure 1.

Concentrations of condensed tannin in leaves of Betula nana (mg g−1; means ± SE from Abisko (n = 6) and Toolik Lake (n = 4) measured by the Vanillin test (a, b) and the Butanol test (c, d), in response to experimental manipulations. At Abisko the treatments are: control (c), shading (s), low temperature enhancement (t1), high temperature enhancement (t2), fertilizer addition (f), fertilizer addition combined with shading (fs), fertilizer addition combined with low temperature enhancement (ft1), and fertilizer addition combined with high temperature enhancement (ft2). At Toolik Lake the treatments are: control (c), shading (s), temperature enhancement (t), fertilizer addition (f), fertilizer addition combined with shading (fs), and fertilizer addition combined with temperature enhancement (ft). Significance levels of *, ** and *** in the repeated measures ANOVAs indicate P≤ 0.05, P≤ 0.01 and P≤ 0.001, respectively.

Figure 2.

Concentration of flavonoids: myricetin derivatives (a, b) and quercetin derivatives (c, d) in leaves of Betula nana from Abisko and Toolik Lake (mg g−1; means ± SE; n = 6 in Abisko and n = 4 in Toolik Lake). Treatment labels and significance levels are as in Fig. 1. 

Figure 3.

Concentration of the phenolic glucosides salidroside (a, b) and betuloside (c, d) in leaves of Betula nana from Abisko and Toolik Lake (mg g−1; means ± SE; n = 6 in Abisko and n = 4 in Toolik Lake). Treatment labels and significance levels are as in Fig. 1. 

Figure 4.

Concentration of chlorogenic acid (a, b) and gallic acid derivatives (hydrolysable tannins) (c, d) in leaves of Betula nana from Abisko and Toolik Lake (mg g−1; means ± SE; n = 6 in Abisko and n = 4 in Toolik Lake). Treatment labels and significance levels are as in Fig. 1. 

Figure 5.

Concentration of N (a, b) and condensed tannins : N ratio (c, d) in leaves of Betula nana from Abisko and Toolik Lake (means ± SE; n = 6 in Abisko and n = 4 in Toolik Lake). The amounts of condensed tannin are based on values from the butanol test. Treatment labels and significance levels are as in Fig. 1.

Results

Site effects

The mass spectra of the GC-MS analysis showed an overall similarity in substances present in the leaf tissues from Abisko and Toolik Lake. However, the Abisko material contained an unknown triterpenoid similar to papyrific acid, which was absent from the Toolik Lake material.

The concentrations of condensed tannins, the two phenolic glucosides and one of the flavonoid groups (myricetin derivatives) were significantly higher in tissue from control plots at Abisko than at Toolik Lake (Table 1). For example, the concentrations of salidroside and betuloside were four and six times higher in Abisko. In contrast, the concentration of quercetin derivatives, another flavonoid group, was significantly higher at Toolik Lake ranging between 26.3 and 31.1 mg g−1 across the sampling period but only between 14.0 and 19.5 mg g−1 at Abisko. Also the leaf concentration of N was considerably higher at Toolik Lake than at Abisko (Table 1). Concentrations of chlorogenic acids and hydrolysable tannins were not significantly different between the two sites (Table 1).

In comparison to these pronounced regional differences, the concentrations in tissues from the acidic and nonacidic sites at Toolik Lake were very similar, except for higher concentration of salidroside in the leaves from the nonacidic site (Table 2). Hence, it appears that differences in soil chemistry did not exert any strong influence on the quantities of measured constituents.

Seasonal variation

In general, the concentration of the various secondary components and N decreased across the season at both sites, except for N at Toolik Lake which in most cases was lowest at the second harvest on 28 July (Figs 1, 2, 3, 4, 5). This could possibly be a short-term variation during the three weeks sampling period at Toolik Lake compared to the truly seasonal variation measured at Abisko.

Treatment effects

Fertilization

Fertilizer addition generally reduced the concentrations of condensed tannins and flavonoids at both sites, although not always significantly so (Table 3). However, there were no effects on phenolic glucosides and hydrolysable tannins (except for interactions with temperature and shading), and the level of chlorogenic acids decreased at Toolik Lake. The nitrogen concentration increased by 16% at Abisko and 24% at Toolik Lake, whereas the condensed tannin : N ratio decreased by 26% at Abisko and 42% at Toolik Lake (Table 3). In general, the effects were much stronger in the leaf tissue at Toolik Lake than at Abisko (Table 3).

Table 3. Relative main factor effects of fertilization, temperature enhancement and shading on seasonal mean concentrations of secondary metabolites and nitrogen in leaves of Betula nana at Abisko and Toolik Lake in percent of values in nonfertilized, nonwarmed and nonshaded plots, respectively. Significance levels of * , ** and *** indicates P ≤ 0.05, P ≤ 0.01 and P ≤ 0.001, respectively. Significant main factor effects are included only in cases where there were no significant interactions between factors. For details on main effects and interactions, see Figs 1, 2, 3, 4, 5
CompoundFertilizationTemperatureShading
AbiskoToolikAbiskoToolikAbiskoToolik
Condensed tannins (Vanillin test) 88** 74***108149** 90 69**
Condensed tannins (Butanol test) 88** 69109159** 86 66
Nitrogen116***124*** 94 74**110*124**
Condensed tannins/nitrogen 74*** 58***116210** 77** 54*
Hydrolysable tannins130133 96 70* 97 85
Myricetin derivatives 87 81 96104104 71***
Quercetin derivatives 80*** 87 99 71 92*101
Chlorogenic acid 95 74***112131 73*** 71***
Salidroside100138 99104 96 88
Betuloside 91 71 95 15110257

Shading

A similar picture appeared after shading, which increased the N concentration by 10% at Abisko and 24% at Toolik (Table 3), but decreased, or did not affect, the concentrations of the CBSC compounds (Figs 1, 2, 3, 4) and, consequently, decreased the condensed tannin : N ratio. Also, as after the fertilizer addition, the leaf tissue from Toolik Lake showed a higher degree of sensitivity to the shading than the tissue from Abisko (Table 3). It is striking that shading generally did not cause significant changes in any of the substances that did not respond to fertilizer addition. However, shading caused a lower magnitude of responses than the fertilizer addition and in some cases, for example for condensed tannins and myricetin, the decrease was pronounced only at Toolik Lake (Table 3). Also, fertilizer addition and shading interacted strongly in some cases, for example in one of the assays of condensed tannins (the butanol-HCl-test on Toolik Lake material), and reduced the level more than expected from the single treatments (Fig. 1). Furthermore, fertilizer addition and shading interacted and increased the level of hydrolysable tannins at Abisko (Fig. 4).

Temperature enhancement

The temperature treatment had no effect on any of the measured compounds at Abisko, except that warming interacted with fertilizer addition and increased the concentration of hydrolysable tannin (Fig. 4). Nor did the temperature change the N concentration or the condensed tannin : N ratio significantly. In contrast, the temperature enhancement applied at Toolik Lake increased the concentration of condensed tannins by about 50% (Table 3), whereas hydrolysable tannins decreased (Fig. 4). The main effect of warming in one of the flavonoid groups, the myricetin derivatives, was due primarily to an interaction effect in the combined fertilized/temperature treatment, in which the concentration was lower than expected from single treatments. Further, as an effect of the increase of condensed tannin and decrease of N, the condensed tannin : N ratio increased by 110% (Table 3).

Discussion

Between site differences in CBSCs

Although the analyses were not designed to screen for qualitative differences in CBSCs, we found some differences between the two Betula nana populations, for instance in the unexpected difference in occurrence of the triterpenoid in the Abisko leaf material, which was not present in the Toolik Lake material.

The quantitative differences were pronounced, as the control material from Abisko in all cases except one (quercetin derivatives) had significantly higher levels of CBSCs or, in two cases, did not differ from the Toolik Lake material. Hence, the dwarf birches at Abisko invest a much higher proportion of C as CBSCs in the leaves than the birches at Toolik Lake. In fact, by assuming that the molecular proportion of C is similar in corresponding groups of CBSCs between the sites, the C investment in the analysed compounds is in the order of 50% higher at Abisko than at Toolik Lake. Furthermore, although the concentrations declined somewhat through the growing season, the CBSCs made up well above 150 mg g−1 dry mass of the leaves at all sampling occasions, suggesting that this investment could represent a considerable C cost. Basically, the differences in levels between Abisko and Toolik Lake coincide with earlier observations that at least condensed tannins often increase as N decreases in the leaves (Peñuelas & Estiarte, 1998).

The great differences suggest that the control of the investment also differs between the sites. This could be for instance a genetic control (the Alaskan species is often considered a different subspecies B. nana ssp. exilis;Hultén, 1968), a function of different environmental conditions, or an evolutionary response to differences in herbivore pressure. Our data do not permit any conclusion about the reasons. However, the great similarity in CBSC concentrations between the two local populations at the Toolik Lake site strongly suggests that at least differences in soil chemistry reflected as, for example, soil pH are of low importance at these sites. Regardless of the control of the investments, the effect of the differences in CBSCs levels between the regions could be a lower palatability for herbivores and lower decomposition rates in the Abisko population than in the two populations at Toolik Lake.

Treatment responses

The generally lower levels of CBSC at Toolik Lake became even more pronounced across the fertilizer addition and shading treatments as the levels of most substances were reduced further. For instance, the concentration of condensed tannins, that is the quantitatively most abundant component, was reduced to about 50% of the control level in the combined fertilizer and shading treatment at Toolik Lake. By comparison, it was reduced by less than 25% at Abisko, and mainly as a response to fertilizer addition. Similar strong treatment effects were found for almost all substances in the leaves from Toolik Lake contrasting with much lower responses in the Abisko population.

The pronounced reduction of C allocation to condensed tannins at Toolik Lake after fertilizer addition coincides with a very strong growth response. Indeed after 9 yr of fertilizer addition, the B. nana there had increased its biomass 2- to 3-fold compared with the controls. Hence, the fertilizer addition obviously increased the allocation of C to growth at the expense of allocation to formation of tannins (Herms & Mattson, 1992).

The temperature treatment had much more limited effects than the fertilizer addition at both sites, although warming increased the concentrations of, for instance, condensed tannins at Toolik Lake. However, temperature treatment alone did not result in any significant growth response at Toolik Lake (Chapin et al., 1995). Hence, it appears that the ‘extra’ photosynthetic C gain, which is likely to occur as temperature increases, could not be diverted to growth probably because nutrients were in short supply. As a consequence, part of the ‘surplus’ seems to have been incorporated into condensed tannins. This agrees with previous observations of high correlation between tannin concentration in north Scandinavian tundra plants and temperature sums over an 8-yr period (Jonasson et al., 1986). In contrast, shading is likely to reduce available C and results in both a reduced growth and a reduction of CBSC levels, which also was obvious, particularly in the combined treatments of shading and fertilizer addition. It appears therefore that the partitioning of C between growth and synthesis of CBSC in general follows the changes in sink strength imposed by the treatments.

The relationship between C allocation to growth and formation of CBSCs is further strengthened by the low responses to all treatments at Abisko, where we have been unable to record any changed growth in B. nana after any of the treatments (Graglia et al., 1997). The small, but significant decline of CBSCs after fertilizer addition does not contradict this because our sampling for biomass effects is less sensitive than the analyses of leaf chemistry. It is possible therefore, that B. nana has responded by increased growth to the fertilizer additions but that the response is too small to be picked up in the analyses of biomass responses.

However, the detailed responses did not always coincide across large groups of CBSCs. For instance, it is striking that the condensed tannins increased with warming at Toolik Lake (Fig. 1), but that the quercetin derivatives decreased strongly (Fig. 2). It is known that enhanced UV-B radiation induces increased concentration of flavonoid compounds in some Betula species (Lavola et al., 1997) and in the ericaceous dwarf shrub Cassiope tetragona (Björn et al., 1997). Hence, the decrease of flavonoids in the temperature treatment could be due to a change in the UV : PAR ratio, possibly mediated by the plastic used for the greenhouses, which reduces the UV radiation, particularly in the UV-B spectrum by about 15% (data not shown). Further, it is notable that hydrolysable tannins declined in plants after all single treatments, but that they increased when both shading and temperature enhancement were combined with fertilizer addition. A similar response was not found for any other substance. Differences in responses between hydrolysable tannins and, for example condensed tannins have, however, been reported previously and could be a function of different pathways for their synthesis (Koricheva et al., 1998; Peñuelas & Estiarte, 1998; Haukioja et al., 1999).

Conclusions

We conclude that the previously-observed higher growth response to the treatments in the Toolik Lake population than in the Abisko population also persisted in the response of the main CBSC. This implies that the Toolik Lake population for unknown reason has a much higher flexibility in its C allocation than the population at Abisko, and that the population at Toolik Lake gave priority to growth rather than to accumulation of CBSC when environmental conditions were favourable. This occurred, for example when growth-limiting nutrient availability increased.

Given that CBSC in general has a grazer deterrent effect, our results suggest that the defence level in B. nana leaves is much lower in birches from Toolik Lake than from Abisko. Further, the Toolik Lake population also had higher leaf N concentration, which together with the lower CBSC levels indicates a general higher tissue quality than at Abisko. The lower general defence level and the preference of C allocation to growth implies a risk if nutrient levels increase, because the grazer deterrent effect will decline particularly in strongly responding populations, as observed in the birches from Toolik Lake. However, in a warmer environment the defence level will increase and approach that in the less flexible population at Abisko.

Acknowledgements

We are much indebted to Prof. T. V. Callaghan and the staff at the Abisko Scientific Research Station for logistic support and to the technical staff at the Department of Biology, University of Joensuu and technician K. Heinsen at the Department of Plant Ecology, University of Copenhagen for help with the chemical analyses. E. G., S. J. and A. M. wish to thank the Danish Natural Science Research Council (grant no. 9501046) and the Swedish Environmental Protection Board (grant no. 127402) for financial support.

Ancillary