1 While herbivory is considered a major selective force operating in mature vegetation, much less is known about how herbivores affect the expression of defence at the seedling stage. In this study we quantified several chemical and physical properties, usually considered to affect seedling palatability, of 14 Western Australian Hakea species. We also determined whether seedling defences were related to other plant life-history attributes.
2Of all of the attributes measured, chemical defence (phenolic content of shoots) exhibited the most significant correlations with other traits (six of a possible eight). We also detected an apparent trade-off between chemical (phenolics) and physical (spinescence) defence.
3Seed mass, relative growth rate and specific leaf area are traits commonly held to have the most predictive power in simplifying our understanding of plant ecological and evolutionary processes across species boundaries. Within this genus, however, these attributes were poorly correlated with other traits.
4We conclude that more extensive studies of seedling defence attributes are needed for a fuller understanding of what governs the evolution of plant life-history traits.
In order to make generalizations, ecologists have attempted to link the distribution of plant species along environmental gradients to their physiological and morphological attributes (Grime 1979; Grime et al. 1997; Westoby 1998). Indeed, some authors (Westoby, Leishman & Lord. 1997) have even suggested that putting species into a ‘comparative context’ is the key to ecological research over the coming years. The search for plant traits that simplify our understanding of the ecological and evolutionary processes influencing plant distributions has concentrated on such attributes as seed size (Leishman et al. 2000), relative growth rate (Marañón & Grubb 1993), and leaf morphology (Wilson, Thompson & Hodgson 1999). Studies of this kind have identified physiological trade-offs constraining the evolutionary responses of plants as they interact with their biotic and abiotic environments. Significant among these trade-offs are those between traits conferring herbivore resistance and those that influence plant growth (Herms & Mattson 1992).
The aim of our study was to determine whether there were consistent patterns in the physical and chemical defensive traits displayed by seedlings of 14 Western Australian Hakea (Proteaceae) species. In addition to measuring leaf phenolic and cyanide concentrations, we also quantified physical defences, including spinescence, leaf thickness and density, and specific leaf area (SLA). Besides its role as an (inverse) index of leaf toughness (Groom & Lamont 1999; Moles & Westoby 2000; Witkowski & Lamont 1991), considerable interest has centred on SLA as an index of plant growth potential and as a key trait in relating plant distribution to resource availability (Westoby 1998; Wilson et al. 1999; Wright & Westoby 2001). It has also been suggested that as a consequence of their high total nitrogen concentrations and smaller allocations to chemical defence, high-SLA leaves may be more vulnerable to herbivory (Grime et al. 1996). Nevertheless, the relationship between SLA and chemical defence remains conjectural.
Materials and methods
SPECIES SELECTION AND COLLECTION OF PLANT MATERIAL
The genus Hakea accounts for over 100 species in a total south-western Australian flora of 6000 species (Paczkowska & Chapman 2000). The seeds are enriched with nitrogen and phosphorus such that seedlings make little use of soil nutrients in the first few months of growth (Stock, Pate & Delfs 1990). Herbivory is a significant factor affecting recruitment of Hakea seedlings (Hanley & Lamont 2001; Lamont & Groom 1998), and there is considerable variation in their leaf morphology, and degree of physical defence (spinescence). Many species (e.g. Hakea erinacea) possess needle-shaped (terete) leaves with sharp, terminal spines. Others have broad, flat leaves that develop numerous marginal spines (Hakea oleifolia). Some broad-leaved species, in contrast, possess only small terminal calluses on their leaves (Hakea elliptica) and appear to lack physical defences (Groom, Lamont & Markey 1997).
The study species were selected because they are well represented in the region, seeds were readily procurable, and they covered the full range of leaf variations in Hakea; seven were broad-leaved and seven needle-leaved, with different amounts of spinescence in each (Groom et al. 1997). Although the geographical ranges of the individual species vary, each broad-leaved species occurs with at least one of the needle-leaved species. Fruits (follicles) of Hakea erinacea, H. lissocarpha, H. petiolaris, H. stenocarpa, H. trifurcata and H. undulata were collected in March 1999 from sites around Lesmurdie, on the Darling Scarp, 20 km east of Perth (31°57′ S, 115°51′ E). To extract the seeds, follicles were placed in direct sunlight and dried until the seeds fell out. Seeds of the remaining eight species were supplied from commercial sources. All had been collected from sites in south-western Australia during the previous 10 months. Nomenclature follows McCarthy (1999).
SEED MASS AND RELATIVE GROWTH RATE
Mean seed mass for each species was determined by weighing 50 seeds of each species. Seeds were germinated in Petri dishes in a dark incubator set at 15 °C. Following germination, seedlings were transferred to 50 mm diameter pots containing a 4 : 1 (by volume) quartzitic sand : peat compost mixture. This mix accurately reflects the texture and low nutrient status of soils in this region (Lamont 1995). Nutrient enrichments are likely to reduce growth rates, or even cause seedling death (Milberg, Pérez-Fernández & Lamont 1998), so this potting medium was considered to represent near-optimum soil nutrient conditions for hakeas.
Two seedlings of each species were planted into each of 16 pots, while four seedlings were planted into each of eight pots for the first harvest. These pots were randomly arranged on large trays placed inside an illuminated incubator (900 µmol PAR m−2 s−1) set at a 22 °C, 12 h day/15 °C, 12 h night regime. These conditions are similar to those experienced by Hakea seedlings during the winter establishment and growth period. Deionized water (500 ml) was added to each tray daily such that each pot received about 10 ml water. After 14 days, all the pots containing four seedlings were harvested. Seedlings were removed from their pots and any adhering compost washed from the roots before oven-drying overnight at 80 °C. The remaining seedlings were grown on for a further 14 days and harvested at 28 days. Watering of the remaining pots was adjusted to maintain 10 ml water per pot for the remainder of the experiment. Mean relative growth rate (RGR) was calculated using the formula given by Hunt et al. (1993).
To quantify total phenolic compounds, we germinated 300 seeds of each of the 14 experimental species in Petri dishes in a dark incubator set at 15 °C. Immediately following germination (10–14 days), seedlings were transferred to 7 × 7 cm square pots containing a 4 : 1 sand : peat compost mixture as above. Four seedlings of the same species were planted in each pot in a 2 × 2 array, such that each seedling was 6 cm from its nearest neighbours. There were 10 pots per species.
Plants were maintained in a greenhouse (25 °C max/12 °C min), and watered daily until 3 months old. At this time, all shoots within each pot were harvested and dried in an oven for 48 h at 50 °C. The relatively low drying temperature reduces the risk of denaturing secondary metabolites (Hagerman 1988). Following drying, the plant material was ground using a pestle and mortar. Each pot of four seedlings represented a single replicate (individual seedlings were too small to test accurately for phenolics). Total phenols were extracted from 50 mg ground leaf material added to 7·5 ml distilled water boiled for 2 h. The extract was centrifuged at 3000 r.p.m. for 15 min and a 1 ml sample removed and added to 7·5 ml distilled water. To this, 1 ml Folin–Denis reagent (Waterman & Mole 1994) and 0·5 ml sodium carbonate (35 g per 100 ml) were added, and samples were left to stand for 15 min before centrifuging at 3000 r.p.m. for 15 min. The absorbance of each sample was read at 760 nm and plotted against a standard curve produced from a similar analysis of tannic acid.
While other secondary metabolites such as alkaloids have not been recorded in the Proteaceae (Everist 1981), cyanogenic glycosides may sometimes be present (Lamont 1993). To determine whether cyanide was present in our species, we conducted a simple qualitative test (Lamont 1993). One 3-month-old seedling (about 0·1 g of freshly harvested plant material) was placed in a 10 ml screw-top vial and crushed with a metal rod. Three drops of deionized water and three drops of 0·1%β-glucosidase (Sigma, emulsin; Sigma Chemical Co., St Louis, MO, USA) were then added. Feigl–Anger test paper was cut into strips 60 mm long and 5 mm wide, and suspended just above the mixture in each vial via the cap. Samples were left for 2 h at room temperature, the presence of cyanide indicated by the test paper turning blue. Cyanide tests were replicated 10 times for each species.
Total leaf nitrogen was determined on a three-channel auto-analyser system, using a modified Kjeldahl digestion method (Anonymous 1977). Leaf samples were taken from the same plants used in the phenolic test and were prepared in a similar way. About 0·1 g dried and milled shoot material was digested in a 6 ml mixture of sulphuric acid and 6% salicylic acid overnight (Ekpete & Cornfield 1964). This mixture was boiled at 350 °C for 20 min, allowed to cool, and 2 ml of hydrogen peroxide added. The process was repeated three times until the samples became clear, after which a final heat treatment at 350 °C for 30 min was applied (Yuen & Pollard 1954). Nitrogen was colorimetrically determined (as NH4+-N) by indophenol blue (Searle 1984).
Leaf thickness was quantified using vernier callipers placed midway across the long axis of a single leaf. Total leaf area was determined by measuring the area of the same leaf using an image analyser (DIAS II, Delta-T Devices, Cambridge, UK). Spinescence was determined by dividing the number of spines on each leaf by its total area. The same leaves were oven-dried at 80 °C overnight and their dry mass determined. Specific leaf area (SLA) was calculated by dividing leaf area by dry mass, correcting for shape of needle leaves (Witkowski & Lamont 1991). Leaf density was obtained by dividing SLA by leaf thickness. Although closely related, we used leaf density, rather than percentage dry matter (Wilson et al. 1999), as density takes into account air as a component of leaf structure and is better related to leaf structure than dry matter content. Each of these leaf characteristics was determined for 20 individuals of each species.
The distributions of mean species data were checked for normality using the Anderson–Darling test (Dytham 1999). Those seedling attributes not conforming to a normal distribution were transformed where necessary. Seed mass, RGR, leaf area, thickness, and spinescence were all log10-transformed. Leaf density was square-root transformed; percentage leaf phenolics and nitrogen were arcsine transformed. We were unable to normalize either the seed mass or percentage leaf nitrogen data because those for H. platysperma were significant outliers. However, we normalized these data when this species was omitted. Following transformations, we compared the mean values of each trait within broad- and needle-leaved species using one-way anova. The entire data set was then subjected to principal components analysis (PCA) with vectors (Podani 1995). We also examined the relationships between traits for the 14 species using a Pearson’s correlation matrix. Given the large number of pairwise tests and the possibility of type I errors, the usual procedure would be to apply the Bonferroni correction (Rice 1989). However, as our overall aim was to look for general trends in the number of significant (P < 0·05) interactions for individual traits, rather than the significance of specific interactions, we did not apply the Bonferroni correction.
Needle and broad-leaved species differed markedly in terms of several of their trait characteristics (Table 1). The PCA ordination (Fig. 1) shows the separation of broad-leaved and needle-leaved species along the vectors describing spinescence and phenolics. The position of individual species within the ordination is only poorly associated with RGR, seed mass and leaf nitrogen (with the exception of H. platysperma). Cyanoglycosides were not detected in any of the seedlings.
Table 1. Summary of mean values (± SE) of nine seed and seedling attributes determined for 14 broad- and needle-leaved Western Australian Hakea species. Group mean values for each attribute within broad- and needle-leaved species are also shown, together with results of one-way anova tests comparing these attributes between the two leaf morphology groups
The importance of the defensive traits, particularly phenolics, is shown by the Pearson product-moment partial correlation coefficients (Table 2). Of a possible 36 pairwise comparisons, 12 were significant at P < 0·05 (two could be expected for random, independent data), and seven of these involved either chemical (percentage phenolics) or physical (spinescence) defence traits. Percentage phenolic concentrations were significantly associated with six other traits (of eight possible pairwise contrasts), the most exhibited by any one attribute. In contrast to the importance of phenolics, there were no significant associations involving seed mass or leaf nitrogen, only three involving RGR, and two SLA. When we did a Pearson’s correlation matrix using the non-normalized data sets for percentage nitrogen and seed mass (H. platysperma was included), only two significant relationships involving nitrogen (negative with RGR and positive with seed mass) were apparent. However, the same analysis identified five significant interactions for seed mass (negative with RGR, phenolics and SLA, positive with nitrogen and leaf thickness), highlighting the fact that seed mass for H. platysperma represented a significant outlier.
Table 2. Pearson’s partial correlation matrix (with significant P < 0·05 coefficients in bold) between all pairings of nine seed and seedling attributes for 3-month-old seedlings of 14 Western Australian Hakea species. Correlations were performed without seed mass or percentage leaf N data for H. platysperma as they prevented normalization of data
The two main components of seedling defence quantified in this study (leaf phenolics and spinescence) were significantly related to each other, and in the case of phenolics, with five of the seven other traits. This was by far the most significant interactions involving any single attribute. The negative correlation between phenolics and spinescence was highlighted by the separation of broad- and needle-leaved species along the respective vectors on the PCA, and the significant difference between phenolic concentrations for needle- and broad-leaved species. This finding points to a trade-off between physical and chemical defences in Hakea seedlings, similar to those noted for South American Prosopis species (Pisani & Distel 1998) and Western Australian Gastrolobium species (Twigg & Socha 1996). Given this apparent trade-off between chemical and physical defence, it is unsurprising that we also found phenolics to be significantly associated with leaf thickness (P = 0·006) and leaf density (P = 0·03).
The significant (P = 0·03) positive relationship between phenolics and SLA is, by contrast, inconsistent with the literature. Most studies predict that low-SLA leaves (slow-growing and long-lived) should be better defended against herbivore attack than their high-SLA counterparts (Grime et al. 1996; Westoby 1998). Furthermore, the positive relationship between phenolics and RGR reported here also conflicts with conventional wisdom (Herms & Mattson 1992; Jones & Hartley 1999). The explanation may rest with trade-offs between physical and chemical defences, coupled with large variation in leaf morphology of our species.
Large/thin-leaved species contained 50% more phenolics than the needle-leaved species. In the absence of well developed physical defences (low spinescence, reduced thickness and high SLA), it may be adaptive for broad-leaved Hakea species to exhibit increased chemical defence. Furthermore, as RGR was not nitrogen-limited in this study (there was no correlation between RGR and leaf nitrogen), it is unlikely that there was any trade-off between carbon allocated to protein (growth) or phenolics (defence) at the common phenylalanine step (Jones & Hartley 1999). Broad leaves are also more photosynthetically efficient than needle leaves (Groom, Lamont & Kupsky 1994), so for a given nutrient status and mass they can fix more carbon for growth and secondary metabolism. The net effect was increased RGR and SLA in the broad-leaved species in tandem with increased allocation of carbon to chemical defence.
The generality of our results, especially with regard to ecophysiological relationships involving SLA, may be limited by the relatively complex leaf morphologies within hakeas. Nevertheless, the fact that leaf phenolic concentrations were frequently correlated with other attributes suggests that, for this group of plants at least, herbivory is a significant influence on seedling responses to their environment. This likelihood is supported by the strong negative relationship between leaf phenolic concentrations and rates of herbivore attack reported previously for the Proteaceae (Hanley & Lamont 2001) and the prediction that stress tolerators, like Western Australian hakeas, should exhibit well developed defences (Grime et al. 1996; Grime et al. 1997; Grubb 1992).
However, seedling herbivory may be a significant selective force for many species, not simply those growing in resource-limited conditions. For example, several studies have shown the considerable selective effect exerted by molluscs on seedlings in temperate grasslands (Hanley et al. 1995); by crabs in Indian Ocean island forests (Green et al. 1997); and by crayfish in Scandinavian aquatic macrophyte communities (Nystrom & Strand 1996). Not only is there significant variation in seedling acceptability between co-existing species, this variation also exists across functional groups (Fenner, Hanley & Lawrence 1999).
By selectively removing plants at the seedling stage, herbivores can effectively exclude certain species from plant communities (Hanley et al. 1995; Wilby & Brown 2001). This mechanism affects individual species directly, and it has the potential to exert community-level effects through changes in species diversity, competitive interactions, and community succession (Brown et al. 1987; Hanley et al. 1995; Vasconcelos & Cherrett 1997; Wilby & Brown 2001). Nevertheless, despite the importance of seedling herbivory as a selective force within plant communities, attempts to examine trade-offs between seedling defence and other life-history traits are few. For instance, there has been no comprehensive analysis of whether defensive traits are correlated consistently across a broad spectrum of co-existing species, and whether these traits are shared within common functional groups. This study has shown that seedling defence is a key characteristic within seedlings of at least one genus. If we are to understand more fully what governs the evolution of plant life-history traits within seedlings, a deeper insight into the development and expression of seedling defence across generic and functional group boundaries is required.
We thank Ken Dods, Dave Allen and Warren Ayliffe of the Chemistry Centre WA for use of their facilities and expertise during the chemical analyses, and Christine Rafferty who helped with chemical and morphological analyses. This work also benefited greatly from the constructive comments of two anonymous referees. The project was funded by an Australian Research Council, IREX Fellowship, to M.E.H.