1Herbivory and litter decomposition are key controllers of ecosystem carbon and nutrient cycling. We hypothesized that foliar defences of plant species against vertebrate herbivores would reduce leaf digestibility and would subsequently, through ‘afterlife effects’, reduce litter decomposability.
2We tested this hypothesis by screening 32 subarctic plant species, belonging to eight types in terms of life form and nutrient economy strategy, for (1) leaf digestibility in cow rumen juice; (2) biochemical and structural traits that might explain variation in digestibility; and (3) litter mass loss during simultaneous incubation in an outdoor subarctic litter bed.
3Interspecific variation in green-leaf digestibility corresponded significantly with that in litter decomposability; this relationship was strongly driven by overall variation among the eight plant types (r = 0·92). The same relationship was not detectable within plant types in taxonomic relatedness tests.
4Several biochemical and structural parameters (phenol-to-N ratio, lignin-to-N ratio) explained a significant part of the variation in leaf digestibility, but again only between and not within plant types.
5Our results provide further support for the role played by foliar defence in the link between plant and soil via the decomposition pathway. They are also a new example of the potential control of plant functional types over carbon and nutrient dynamics in ecosystems.
We sampled species within a 25 km radius around Abisko, north Sweden (68°21′ N, 18°49′ E); the exception was Picea obovata, which was sampled 80 km east of Abisko. We sampled species in their typical lower-altitude (300–600 m) subarctic terrestrial ecosystems (Appendix 1). These included: (1) heathland (drier tundra) with ericaceous dwarf shrub species, Calamagrostis lapponica grass and forbs, some with 4 m tall Mountain Birch (Betula pubescens ssp. tortuosa) trees; (2) wet (4 m tall) woodland with Mountain Birch, Willow (Salix spp.), Grey Alder (Alnus incana) and perennial forbs; (3) ruderal sites with grasses, forbs and Willow; (4) coniferous forest patches (Pinus sylvestris, Picea abies×obovata); (5) mires with sedges (Eriophorum spp., Carex spp.); and (6) forb-rich subalpine meadows. For the digestibility study we selected 32 vascular plant species that are common in the area and which range widely in ecology and taxonomy. They were a subset of a larger species set employed for nutrient resorption and decomposition studies (Quested et al. 2003). The species groupings in this paper (‘plant types’, see Appendix 1) are a combination of life forms corresponding with the broadly defined ‘functional types’ described by Chapin (1996), and more strictly defined functional types in terms of N uptake and N conservation strategies in the above ecosystems (Quested et al. 2003).
green leaves: sampling, digestibility, chemistry and structure
All leaves were collected between 10 July and 9 August 1999. For digestibility and lignin analyses we cut off fresh, green, undamaged leaves with their petioles (if any) from healthy mature plants. Usually leaves from at least five plants were pooled, but from only two trees in Picea abies×obovata and Pinus sylvestris. The leaves were stored air-dried and dark until the analyses.
Digestibility of green leaves was analysed at the Institute of Rural Studies, Aberystwyth, Wales, as in vitro cow's rumen liquor digestibility, based on the method of Tilley & Terry (1963). Three 0·5 g subsamples of each leaf population were ground and incubated with buffered cow's rumen liquor for 48 h, then with acid-pepsin solution for a further 48 h. Dry-matter digestibility (digested mass per unit initial mass, g kg−1) was calculated using the dry mass of the residue.
For lignin concentration, three (in a few cases two) 2 g subsamples of each leaf population were ground and processed at the Institute of Rural Studies with the Van Soest assay as described by Allen et al. 1989) and discussed by Palm & Rowland (1997). It is important to note that this assay does not distinguish between different types of lignin in terms of the composition and amount of polymerization of its building blocks (Kögel-Knabner 2002). Some lignin types are more resistant to microbial degradation than others, and this difference has been linked to plant taxa or types (Kögel-Knabner 2002). The same may apply to other acid-insoluble compounds such as cutins and suberins, which are also included in the lignin fraction in this assay. Thus our lignin data can only be viewed as a crude indication of structural leaf defence.
For total phenol concentration and foliar N concentration, four separate subsamples were ground for most species, while for some species a bulk sample was ground from which four subsamples were taken for analysis. Ground subsamples (≈2·5 mg) of plant material were weighed to the nearest µg into tin elemental analysis cups. Mass-based N concentrations were determined using an isotope mass ratio mass spectrometer (Tracermass Europa Scientific Ltd, Crewe, UK). We used a combination of methanol extraction and the Folin–Ciocalteau assay to determine mass-based total soluble phenol concentrations of leaves (Waterman & Mole 1994; Palm & Rowland 1997). This method has the disadvantage that it does not differentiate between phenolic compounds involved in defence and non-defence phenols such as anthocyanins involved in autumn coloration (Stadler Martin & Martin 1982; Harborne 1997). Also, it (mostly) misses an important group of defence phenolics, the condensed tannins (Kögel-Knabner 2002), as well as more herbivore-specific non-phenolic toxins. On the other hand, total soluble phenol concentrations as measured here indicate overall chemical defence (McKey et al. 1978) and inhibition of net N mineralization rates in soil (Palm & Rowland 1997), and are probably robust enough to be used in broad comparisons among large species assemblages.
Specific leaf area (fresh lamina area over lamina dry mass, SLA) tends to position species on an axis of foliar investments in assimilation-related vs defence-related chemistry and structure (see below). We used 10 leaves per species, where possible from several different plants. We measured leaf area with an Area Meter (Delta-T, Burwell, Cambridge, UK) and leaf dry mass after 72 h at 60 °C.
litter: sampling and decomposition study
During September 1999 we collected fresh leaf litter of the same 32 species and sites as for green leaves. We collected freshly senesced, undecomposed leaf litter from mature plants. We made bulk collections of litter from several (in most cases many) plants. In species that shed their leaf litter we either collected this from the ground or caught dead leaves that dropped after gently shaking plants. In species that retain dead leaves on the plant (e.g. Arctostaphylos alpina, Empetrum nigrum) or in herbaceous plants that die back completely above-ground (forbs, monocotyledons), we cut off leaves that were subjectively judged to be dead but still undecomposed. Petioles and the rachides of compound leaves were generally shed as an integral component of the leaf litter and were collected and processed as such. In Equisetum sylvaticum the entire shoot functions as the green photosynthetic unit analogous to leaves and also senesces as a unit; we collected such units as leaf litter. Further sampling details are given by Quested et al. (2003).
The litter collections were cleaned, then air-dried and stored in paper bags in a laboratory at 20 °C until needed. Most litters reached equilibrium moisture content, usually between 3 and 6% (as a percentage of oven-dry mass at 40 °C; see below) within 4 days. The moisture content for each collection was determined from a subsample at the time of weighing of the litter bag samples. We dried these subsamples at 40 °C to use the material for subsequent phenol assays (see above). Litter subsamples were taken at random from each collection and 1·0 ± 0·1 g air-dried material was weighed to the nearest mg and sealed into a nylon litter bag with 0·3 mm mesh, which allowed exchange of micro-organisms and small soil invertebrates. Filled litter bags were stored air-dry in the laboratory at 20 °C until incubation. Further details about litter processing and additional treatments to evaluate possible methodological artefacts are given by Quested et al. (2003).
After remoistening, the litter bag samples were incubated in outdoor litter beds sensu Cornelissen (1996), fenced off against vertebrate herbivores in a nursery at the Abisko Research Station. These litter beds were exposed to the natural macroclimate (for climate and weather data see Quested et al. 2003). Because of its slightly lower, sheltered location compared to the surrounding birch heath-woodland, it was (during the experimental period) almost continuously covered by snow from November to May. The main litter bed consisted of two 100 mm deep, rectangular wooden frames, sunk into the ground. A layer of grit stones (particle sizes 10–20 mm) of local origin constituted a free-draining foundation on top of the original soil profile. Each frame was subdivided into 0·7 × 0·7 m squares with wooden sides, each pair of squares serving as a plot hosting one of five replicate litter bags of each species. (An additional five litter bags per species were buried simultaneously for long-term incubation.)
This litter bed was filled with the incubation medium, a loose 100 mm layer of thoroughly mixed litter taken from nearby typical birch heath-woodland (details in Quested et al. 2003) on 20 September 1999. On 2 October 1999 all litter bags were positioned flat, without overlapping one another, ≈3–4 cm below the surface. After compaction of the litter medium, incubation depth was soon reduced to 2 cm. This is deeper than the natural depth of most fresh litter, and will probably amplify differences among species in potential decomposition rate (unpublished results). We did not attempt to obtain absolute decomposition rates directly representative of the species’ natural litter environments. The litter bags were harvested after 1 year's incubation on 18 September 2000.
We used mean species values for digestibility, chemistry, SLA and litter mass loss as observations in data analyses.
Six of the 32 species were not included in the 1999/2000 decomposition experiment, but their decomposability was determined in a similar earlier experiment after 1 year's incubation in the same litter bed, in 1998/99, along with many other vascular species (details in Quested et al. 2003). To be able to use the data for these six species, we transformed decomposability data from the 1998/99 experiments to those for the 1999/2000 experiments, using a 30-species subset common to both experiments. Incubations in the 1998/99 experiment (X) and the 1999/2000 experiment (Y) revealed a strong correspondence of mass losses in linear regression (R2 = 0·81) and the equation (Y = 0·994X) was used to transform data for the six missing species to 1999/2000 values.
For N concentration ([N]) and specific leaf area (SLA), which may be relatively sensitive to allometric/ontogenetic factors, collections from the same leaf populations were made in July and again in August for a subset of the species, and regressed against one another. We used the regression equations to transform other August-collected [N] and SLA data to July values.
Transformations from August (X) to July (Y) values followed Y = 0·83X (N = 8, R2 = 0·68) for foliar N and Y = 0·944X (N = 18, R2 = 0·81) for SLA. Using these transformations, we could calculate July-based [N] and SLA for all species.
We carried out separate analyses of the hypothesized relationships both within and between plant types. Within types, we employed taxonomic relatedness tests as described by Kelly & Beerling (1995), based on Felsenstein (1985). This method investigates the occurrence of a relationship at each fork of a taxonomic tree, thus differentiating between relatively recent evolutionary divergences (e.g. within genera and families) and more ancient ones (e.g. within subclasses and classes). We used the taxonomic classification of Cronquist (1981). First we recorded the sign (positive or negative) of a given hypothesized relationship, e.g. positive between leaf digestibility and litter decomposability. Each taxonomic contrast consisted of a set of two or more lower taxa belonging to the same higher taxon (e.g. three species within the same genus), and we recorded whether for this contrast the slope of the relationship showed a positive or negative trend. The average trait value of the lower taxa could subsequently be used for a similar comparison of the next level up (e.g. two genera within a family). We continued stepwise up to the between-classes level where relevant. We used this procedure only within each of the eight plant types, so as not to confound the analyses between plant types. Combining the results of the contrasts at lower (up to families within orders) or higher taxonomic levels (between orders and up), or all together, we employed χ2 tests to investigate whether the relationship was significantly more often positive or negative (as hypothesized) than expected by chance. For detailed examples with continuous traits, see Cornelissen, Castro Díez & Hunt (1996).
We used the final clustering for each plant type after all the taxonomic contrasts to obtain the overall ‘mean’ trait value for that type. (These values were hardly different from simple means among species within types: data not shown.) We then calculated Pearson's product-moment correlation across the eight plant types to test the correspondence between green-leaf digestibility and leaf litter decomposability, as well as between several biochemical or structural parameters and digestibility of green leaves. Biochemical parameters were ln-transformed prior to analysis in order to approximate normality of frequency distributions.
There was clear correspondence of interspecific variation in green-leaf digestibility and that in leaf litter decomposability (Fig. 1). However, this overall pattern (r = 0·73, N = 32, P < 0·001) was strongly driven by differences between (Table 1) as opposed to within plant types (Table 2). Hemiparasites, N2-fixing forbs and (to a lesser extent) other forbs tended to combine high leaf digestibility with high decomposability. Woody evergreens, fern allies and monocotyledonous groups showed relatively low values for both, although graminoids had lower litter decomposability than expected, based on leaf digestibility (Fig. 1). Woody deciduous species were intermediate for both traits. There was a non-significant overall positive trend (P < 0·1) for the same relationship across higher taxa in taxonomic contrasts, but this pattern disappeared when lower and higher taxa were considered together (Table 2).
Table 1. Pearson's correlation (r) between (a) green leaf digestibility and leaf litter decomposability at the plant-type level; (b) each of several biochemical and/or morphological traits of green leaves and their digestibility
Each data point was the mean for a plant type after stepwise taxonomic clustering of taxa. N = 8 for all relationships. NS, not significant (P > 0·05).
Digestibility vs decomposability
Ln([phenol]/[N]) vs digestibility
Ln([lignin]/[N]) vs digestibility
Specific leaf area vs digestibility
Ln([N]) vs digestibility
Table 2. Results of taxonomic relatedness analyses for relation between (a) green leaf digestibility and leaf litter decomposability; (b) each of several biochemical and/or morphological traits of green leaves and their digestibility
Taxonomic contrasts were taken only within plant types. For each taxonomic contrast for each relationship, the sign of the regression slope was recorded. The proportion of all contrasts for which the slope sign (+/– in parentheses) corresponded with the hypothesis is indicated separately for lower taxonomic levels (species within genera; genera within families; families within orders) and for higher levels (orders within subclasses; subclasses within classes; between classes), and subsequently for all taxonomic contrasts. Taxonomic classification follows Cronquist (1981). In the raw data, phenol and phenol/N data were missing for Vaccinium myrtillus and SLA for Equisetum sylvaticum. None of the proportions differed significantly from that expected by chance (0·5) in χ2 tests.
Digestibility vs decomposability (+)
Ln([phenol]/[N]) vs digestibility (–)
Ln([lignin]/[N]) vs digestibility (–)
Specific leaf area vs digestibility (+)
Ln([N)]vs digestibility (+)
Ln[phenol]vs digestibility (–)
Ln[lignin]vs digestibility (–)
The compound traits phenol-to-N ratio and lignin-to-N ratio (both ln-transformed) correlated significantly and negatively with green-leaf digestibility among the eight plant types (Table 1). Among the other traits, only SLA showed a non-significant positive trend. These results contrasted with those for the entire unweighted set of 32 species, where foliar lignin-to-N ratio (ln transformed r = −0·60, N = 32, P < 0·001) and SLA (r = 0·58, N = 31, P < 0·001) were the best predictors of digestibility, followed by phenol-to-N ratio (ln-transformed r = −0·48, N = 31, P < 0·01) and the simple biochemical (ln-transformed) traits ([phenols]: r = −0·38, N = 31, P < 0·05; [lignin]: r = −0·48, N = 32, P < 0·01; [N]: r = 0·38, N = 32, P < 0·05). The difference in pattern across all unweighted species from that among plant types emphasizes the importance of testing these relationships also within plant types, where taxonomic contrasts revealed no patterns for any of the same biochemical or structural traits, at lower and/or higher levels of taxonomy (Table 2).
In support of our hypothesis, our results revealed clear correspondence between the digestibility of green leaves and the decomposability of the leaf litters among 32 diverse vascular plant species from a subarctic region. Interspecific variation in leaf litter decomposability related to ‘afterlife effects’ of foliar defences, has now been demonstrated for the first time in the context of vertebrate herbivores. This complements previous evidence for the same relationship in the context of invertebrate herbivory, particularly insects and molluscs. Taken together, these results suggest that, in general, plants with more heavily defended leaves will be eaten less by herbivores, which may cause greater input into soil of recalcitrant litter that will subsequently be decomposed more slowly. This would potentially result in slower nutrient and carbon cycling.
However, to interpret this relationship between green-leaf digestibility and litter decomposability in an evolutionary rather than a merely regional context, it was necessary to dissect it at lower and higher levels of functional and taxonomic aggregation. Such analyses revealed that the overall (unweighted) relationship was driven strongly by variation among the eight plant types distinguished here, while it was not apparent within types. This supports the view that classification of plant species based on morphology and functionality, as opposed to higher taxonomy, is a useful concept for investigating carbon- and nutrient-cycling processes (Chapin, Autumn & Pugnaire 1993; Díaz & Cabido 1997; Lavorel et al. 1997; Grime 2001). At the same time, there was great variation among species within certain plant types, particularly among forbs, both for leaf digestibility and litter decomposability. This could indicate that for a functional classification to be informative, a relatively large species set is a prerequisite, and 32 species may well be close to the minimum required. Alternatively, it might highlight that a priori classifications are partly subjective and need to be judged critically for their merit in a particular context. Our plant types were based pragmatically on a mixture of ‘classical’ life forms as commonly used in arctic ecological research (Chapin, Autumn & Pugnaire 1993) and nutrient economy strategies (particularly N2-fixers and hemiparasites; Tilman et al. 2001; Quested et al. 2003). An alternative might have been to classify strictly and explicitly according to nutrient economy, with central roles for mycorrhizal association type and leaf life span (Cornelissen et al. 2001), but this would have resulted in groups too small to be useful for this exercise.
Among the chemical and morphological leaf parameters measured, the compound traits phenol/N ratio and lignin/N ratio were the strongest predictors of interspecific variation in leaf digestibility. Both phenols (as contributors to chemical defence) and lignin (as a main contributor to physical structure and defence) are significant determinants of leaf digestibility and litter decomposition (McKey et al. 1978; Swift, Heal & Anderson 1979; Melillo, Aber & Muratore 1982; Palm & Rowland 1997), although we emphasize again that lignin and phenol as measured in our study can be only crude indicators of recalcitrance to microbial degradation (see Methods). Low nutritional quality, as represented by foliar N concentration, might be seen as a further form of antiherbivore defence (Augner 1995), but green-leaf N concentration alone had poor if any explanatory value for interspecific variation in leaf digestibility. However, it does make an important contribution to predicting leaf digestibility as a component of phenol-to-N or lignin-to-N ratio. The explanatory power of [N] alone in relation to litter decomposability was still weaker (r = 0·36; N = 30; ln-transformed [N]) and non-significant, which probably resulted from interspecific variation in N resorption from senescing leaves (Aerts 1996; Quested et al. 2003). Indeed, N concentrations of green leaves vs leaf litter were less correlated across the 32 species (r = 0·54; N = 32; ln-transformed data) than concentrations of lignin in green leaves vs leaf litter (r = 0·75; N = 29; ln-transformed data), lignin being a recalcitrant compound which is presumably not broken down and resorbed at all during senescence.
Specific leaf area combines structural (and defence) investments with investments in photosynthesis-related chemistry into one trait (Reich, Walters & Ellsworth 1992; Garnier & Laurent 1994; Westoby 1998; Cornelissen et al. 1999), high N usually corresponding with high SLA, and high lignin with low SLA (Reich, Walters & Ellsworth 1992; Van Arendonk & Poorter 1994). Indeed, the correlation of SLA with leaf digestibility matched that of lignin/N ratio with digestibility for the unweighted species set, although among plant types it was only marginally significant. Correspondingly, green-leaf SLA was also a reasonable correlate of litter decomposability in this study (r = 0·55, N = 31, P < 0·01), as was (ln-transformed) green-leaf lignin/N ratio (r = 0·52, N = 32, P < 0·01).
The fact that we did not find any single close predictor of leaf digestibility suggests that: (1) we may have missed other important biochemical compounds in relation to microbial decomposition (see Methods), such as non-phenolic toxins and condensed tannins (the latter have been found to reduce digestibility as much as hydrolysable tannins, McKey et al. 1978); (2) our biochemical assays to determine total lignin and soluble phenolics did not differentiate sufficiently between recalcitrant vs more degradable forms (see Methods); and/or (3) there are multiple controls over digestibility involving complex interactions between several chemical or structural tissue quality parameters.
While this study has provided support for our hypothesis, further tests are needed before we can draw broad generalizations about the association of leaf defences against vertebrate herbivores with resistance of litter to decomposition. For instance: (1) are species decomposability rankings in litter bed studies sufficiently robust to overcome variability caused by methodological and environmental factors as well as incubation duration? The preliminary answer seems affirmative, based on previous findings (Cornelissen 1996; Cornelissen et al. 1999; Quested et al. 2003), but see Berg & Ekbohm (1991) for an example of interaction between species mass loss ranking and incubation time. (2) To what extent do condensed tannins and specific non-phenolic toxins (not measured here) play a role in the relationship, as well as (3) foliar defences other than digestibility inhibitors, for instance chemical repellents or external structures such as spines or leaf accessibility as determined by whole-plant architecture? (4) To what extent do the patterns depend on the vertebrate species from which digestive juice is used (e.g. different large mammals, rodents, birds)? There is evidence that both supports and contradicts the dependence of plant digestibility rankings on the identity of the herbivore species (Bryant et al. 1991). The Welsh cows from which rumen juice was used in our study had not previously been exposed to subarctic plants, thus providing an independent digestibility test. However, this rumen juice could have been relatively efficient at digesting cellulose-rich grasses, given that grasses are an important component of their diet. Indeed, the grasses in our study showed a better digestibility than expected from the regression against decomposability. (5) Can we extrapolate data from a subarctic flora to floras in other biomes (Cornelissen et al. 1999)? The correspondence between rankings of leaf resistance to herbivory and litter resistance to decomposition among plant species and types should be investigated in other floras and ecosystems, and with other herbivores.
With these cautionary notes, our new findings provide further support for the key role played by foliar defences in the link between plant and soil via the decomposition pathway. They are also a new example of the potential control of plants over carbon and nutrient dynamics in ecosystems.
Gary Easton ground the samples for lignin and digestibility analysis. We thank Anne Témesvary and several other colleagues at Abisko Research Station for their help and hospitality. This work was funded by the Nordic Council of Ministers Nordic Arctic Research Program (NARP), the Swedish Academy of Sciences (KVA) and an EU Framework IV grant to T.V. Callaghan within the BASIS project.
Study species, plant type, family and ecosystem of sampling