Lower P contents and more widespread terpene presence in old Bornean than in young Hawaiian tropical plant species guilds

Leaf elemental and secondary metabolite contents and morphological traits are important measures of time-dependent ecosystem changes. We aimed to test whether plants from older tropical forests have lower nutrient contents and different elemental stoichiometry than plants from younger ecosystems (‘‘soil age’’ hypothesis) and whether they had different contents of carbon based secondary compounds (CBSC) and morphological traits as a result of a longer evolution under tropical conditions. We conducted a phylogeny-independent study of the foliar chemical and structural traits in two sets of 86 species each measured in two different-aged tropical forests, a young soil forest in Hawaii and an old soil forest in Borneo. The leaf contents of nutrients and micronutrients tended to be higher in Hawaii than in Borneo but leaf N:P content ratio was not different. The ‘‘soil age’’ hypothesis was thus only partially supported by the results indicating that several other factors influence plant elemental content. Total phenolic content was twice larger in Hawaiian than in Bornean plant species. Terpene contents were not different in terpene-containing species but the percentage of species containing terpenes was much higher in Borneo (97%) than in Hawaii (34%) suggesting that the longer time of evolution in Borneo has allowed a more widespread development of very diverse defensive, allelopatic and information relationships of plants with specialist herbivores and other plants. Principal component analyses separated Hawaii and Borneo species on the basis of leaf elemental composition, total phenolics and terpene contents and leaf dry mass per area (LMA). The results collectively support the ‘‘leaf economic spectrum’’ and ‘‘carbon excess’’ paradigms because in both sets of species and also in the combined set of Borneo and Hawaiian species, there are negative relationships of N content with LMA and total phenolics. The results suggest thus that changes throughout time in N and P availability can be important but do not explain all the variability underlying the evolutionary changes in leaf chemistry and structure in these tropical forests. Other factors determining species biogeochemical niche such as K, Mg or S elemental stoichiometry, leaf economic traits and changes in plant defence and communication strategy are also likely to be involved.


INTRODUCTION
Leaf chemical and morphological traits are key determinants in plant adaptation to biotic and abiotic environmental factors and therefore in ecosystem function (Wright et al. 2004and 2007, Peñuelas et al. 2010a).Quantifying these leaf traits in tropical plants and understanding their change through evolutionary time and between different sites is fundamental for advancing knowledge of tropical forest function.The ''soil substrate age'' hypothesis claims that older tropical soils might be more P-limited than younger soils in cooler regions of the globe (Walker andSyers 1976, Chadwick et al. 1999).Because P is derived primarily from rock weathering, ecosystems begin their existence with a fixed complement of P from which even very small losses can not readily be replenished.Consequently, ecosystems on old soils can become depleted in P. Thus, the long time soil evolution of tropical soils should diminish P content, whereas there may be N input by N 2fixation, thus explaining the generally observed trend of soils evolving from N-limitation to Plimitation over time in wet tropical ecosystems (Walker andSyers 1976, Chadwick et al. 1999).These assumptions were corroborated by Walker and Syers (1976) in New Zealand soil chronosequence studies.This hypothesis was thereafter applied to explain the results of several reports that observed increases of N:P ratios in plants towards the tropics.This is because over the last several million years, high latitude ecosystems have been systematically and cyclically rejuvenated by cyclical glaciations and the resulting major surface movements with consequent replenishment of P. In contrast, no glaciers have existed in the lowland tropics for hundred of millions of years-and so a larger fraction of tropical sites could approach Walker and Syers (1976) terminal state of P depletion (Vitousek et al. 2010).However, metadata studies have reached contradictory conclusions which sometimes support, and sometimes do not support these predictions (Vitousek et al. 2010), and other causes of P limitation have been suggested to explain P limitation in ecosystems.For example, P can be scarce in the first stages of soil and ecosystem development because of ''transactional limitation'' due to lower P leaching, implying that most P remains in primary materials leaving low P availability in soils (Crews et al. 1995, Vitousek et al. 1995).Other circumstances, for example low P in parent materials, or recent increase of other nutrients in soil such as N due to human activities means that changes in relative N and P availabilities in ecosystems throughout time is complex and can depend on several different circumstances (Vitousek et al. 1995 and2010).Most studies have focused mainly on P and N, whereas other important nutrients and elements have not received much attention.On average, leaf N:P ratios of tropical forest are elevated (Davidson and Howarth 2007) and do not seem to vary much with either latitude or annual mean precipitation within the tropical area (Townsend et al. 2007, Lovelock et al. 2007, He et al. 2008).
Changes in C:N:P leaf stoichiometry with ecosystem age can change several other ecosystem traits.C:N:P content stoichiometry is related to several ecological and ecophysiological strategies such as growth rate (Elser et al. 1996, Elser and Hamilton 2007, Mulder and Elser 2009), and to community biotic relationships such as plantherbivore interactions (Schade et al. 2003, Urabe et al. 2003, Perkins et al. 2004, Carline et al. 2005).Long-term changes in ecosystem C:N:P stoichiometry could result from soil changes and species adaptation processes and it is worthwhile to investigate this important trait in tropical forests of different ages.Because different elements have different chemical properties (e.g., solubility) and different nutrient value, they can be leached from parent materials and taken up by plants at different rates thus changing their stoichiometry with time.These changes can be more critical in wet tropical ecosystems due to the great climate impact on materials.
Nutrient availability changes on a timescale of years can also affect leaf content of molecular compounds.The growth-differentiation balance (Herms and Mattson 1992) or the ''excess carbon'' (Coley et al. 1985, Peñuelas andEstiarte 1998) hypotheses, although severely criticised in recent years (Hamilton et al. 2001, Koricheva 2002, Nitao et al. 2002) may still help to explain these changes.They predict a higher production of carbon-based secondary compounds (CBSC) under lower nutrient availability.Lower N and P leaf contents would result in less active primary metabolism and growth and a higher accumulation of CBSC, such as phenolics and tannins.These species with lower N and P foliar contents also present higher LMA and leaf mechanical properties linked to C-rich structural compounds that can also actuate as herbivore deterrents (Kursar and Coley 2003).Interestingly, a great diversity of other hypotheses and theories have been proposed to explain plant defensive strategy in the face of herbivory.Most of these have a more evolutionary basis (e.g., Hamilton et al. 2001).According to these theories, the contents of chemical defence compounds can differ in tropical ecosystems of different ages mainly if nutrient availability and plant productive capacity differs, and/or as a result of the evolutionary time-scale of selective pressure exerted by herbivory.Thus, for old tropical forests, the longer time-scale of coevolution of plants with herbivores might have enhanced the suitability of accumulating defence compounds against specialist herbivores.
Although some secondary metabolites are phylogenetically conserved in specific plant Orders, Families or Genera, e.g., glucosinolates in the Order Capparales (Rask et al. 2000), more frequently the chemical defences are secondary metabolites widely present throughout the plant phylogenetic spectrum, such as phenolics and terpenes.Phenolics include low molecular mass phenolics and condensed polyphenolics such as tannins.They are considered quantitative defences (Feeny 1976, Meyer and Karasov 1989, Adams et al. 2009).They are present in high contents (frequently between 5-40% dry weight) (Meyer andKarasov 1989, Adams et al. 2009) and thus represent a high initial investment cost (Coley et al. 1985).Their effects are the result of the tissue content (Feeny 1992, Nomura andItioka 2002), acting mainly by reducing digestibility rather than through direct toxicity (Bernays 1991, Eichhorn et al. 2007).They are effective against a broad herbivore spectrum including specialist herbivores (Beck and Schoonhoven 1980, Coley et al. 1985, Bernays and Chapman 1994).Converserly, terpenes are present in lower concentrations, and therefore are considered to be less costly qualitative defence compounds (Coley et al. 1985, Langenheim 1994, Stamp 2003).Terpenes act as deterrents because of their toxicity, or as modifiers of insect development (Bennet and Wallsgrove 1994).They are effective against non-adapted specialist herbivores (Sorensen et al. 2005), and generalist herbivores (Mihaliak et al. 1987, Landau et al. 1994, Mote et al. 2007).The capacity to produce carbon-based secondary compounds (CBSC) such as phenolics, tannins and terpenes, can be an involved factor in the competition between plant species by conferring chemical defences against herbivores, allelochemical defences against neighbouring competing plants (Peñuelas et al. 1996, Peñuelas and Estiarte 1998, Kursar and Coley 2003, Khanh et al. 2008), greater anti-stress capacity (Filella and Peñuelas 1999, Peñuelas and Llusia 2003, 2004), or/and greater signalling capacity (Peñuelas et al. 1995, Peñuelas andStaudt 2010).
Borneo is the largest Indonesian Island and has a considerable extent of tropical rainforest.The geological origin of Borneo goes back 56 million years, during the Alpine orogeny when the Pacific plate crashed against Asian plate.In contrast, the Hawaiian Islands that constitute the most isolated terrestrial ecosystem of the Earth (Vitousek and Walker 1989) were formed more recently.Their origin is dated at 2.6 to 5 million of years ago (depending on author; Clague andDalrymple 1987, Guillou et al. 2000).The Oahu island formation began approximately 3.7 millions of years ago (Clague and Dalrymple 1987).In contrast to the Borneo flora, which has thus been established a long time and is currently one of the most diverse ecosystems of the world, the Hawaiian flora was formed from 279 (at most) immigrant plant species that by an intense speciation processes have resulted in a total of 956 native plant species to date (Wagner et al. 1999).In addition to these natives, there are 861 naturalized plant species that have been added relatively recently by humans and that are considered to be either alien or invasive species (Wagner et al. 1999).Some reports have observed that rainforest soil P in Borneo strongly determines woody plant productivity (Paoli et al. 2005, Paoli 2006) and that some areas are N-limited while others are Plimited (Kitayama et al. 2000, Nomura and Kikuzawa 2003, Paoli et al. 2005).On the other hand, Hawaiian soils have also been intensively leached and in spite of being younger than Borneo soils, soil N and P limitation have also been reported (Herbert andFownes 1995, Marti-nelli et al. 1999).But the data suggest that on average Hawaiian soils are less poor in nutrients than Borneo soils, particularly regarding total soil P content (Table 1).
Using statistical phylogenetically independent analyses, we conducted a study to compare leaf chemical traits in two populations of 86 woody dominant species, one population in Hawaii and another in Borneo.Our aims were to investigate (i ) the elemental and stoichiometry differences between the two island forest species sets, (ii ) the differences in allocation to foliar C-rich secondary metabolites (phenolics and terpenes) and LMA and (iii ) the differences between these two island floras regarding the relationships among these leaf chemical variables.The overall main objective was to test the effect of soil aging and of longer evolution time on all these foliar chemical traits.

Field sites and studied species
In Hawaii, the study was conducted in May 2007 on the island of Oahu (Peñuelas et al. 2010a).As typical of larger Hawaiian Islands, the climate is characterized by very steep rainfall gradients over short distances (Mu ¨ller-Dombois and Fosberg 1998).While precipitation is distributed almost uniformly in lowland and mountain rain forests, the lowlands at the leeward side have a pronounced dry summer season.Due to the oceanic tropical climate, temperature oscillations are small, with winters having on average 2-38C cooler temperatures than summers.As large differences in composition of vegetation occur in response to rainfall gradients, four sites with distinct precipitation regimes were selected for plant sampling in the leeward lowlands of Oahu and at the leeward side of Koolau mountains (see details in Peñuelas et al. 2010a).The annual mean temperature averages range from 23 to 278C and precipitation from 1600 to 4200 mm in the studied sites.The four key soil types found across the sites rank according to the state of weathering as oxisols .ultisols .mollisols .inceptisols (Deenik and McClellan 2007) and are fully described in Peñuelas et al. (2010a).A total of 86 dominant Oahu species were sampled in the four sites (species abundant enough to get three replicates per species).
The Borneo study was conducted in The Danum Valley Field Centre that is located at 1178 48.75 0 E and 58 01 0 N on the east coast of the Malaysian state of Sabah, Borneo Island.The station lies on the edge of the 438 km 2 Danum Valley Conservation Area which itself lies within the Ulu Segama Forest Reserve, as part of the ca.10,000 km 2 Yayasan Sabah Forestry Concession.The Danum Valley Conservation Area (DVCA) is a Class I (Protection) Forest Reserve located on the western side of the upper reaches of the Segama River in Southeast Sabah, and is the largest remaining area of undisturbed lowland dipterocarp forest in Sabah.
The climate at Danum is equatorial with a mean annual temperature of 26.88C.Temperatures in excess of 348C are rare, occurring only during prolonged dry periods.Minimum temperatures rarely fall below 198C.Mean annual rainfall  is 2,825 mm; the lowest annual rainfall of 1,918 mm occurred in 1997, which was an ENSO year, and the highest annual total of 3,539 mm occurred in 2003.Mean monthly rainfall ranges from 153 mm in April to 309 mm in January and tends to be highest in the transition months following the equinoxes (May-June and October-November) and also during the northerly monsoon months of December-January.Rainfall is generally lowest during March and April, which are the most drought-prone months during ENSO events, and also in August and September when the southwesterly monsoon is at its height.The climate of Danum Valley is aseasonal but subject, as in 1997-98, to occasional severe droughts and is intermediate between the less drought-prone north-western Borneo and the more droughtprone east coast.For more details of the study area see Peñuelas et al. (2011).A total of 86 dominant species were sampled (Fig. 1).Species nomenclature follows the local floras (Peñuelas et al. 2011).

LMA and elemental analyses
For plant sampling details see Peñuelas et al. (2010a and2011).In the laboratory, the areas of 3-10 leaves (depending on leaf size) per each sampled plant were measured using a LICOR LI-3100 area meter (LI-COR, Lincoln, Nebraska, USA).For later analysis of trace element content only in foliar tissues, leaves were thoroughly washed with distilled water in the laboratory to remove the elements deposited on leaf surfaces.After washing, the samples were dried in an oven at 708C to a constant mass.Thereafter, dry mass of leaves was determined and leaf dry mass per area (LMA, g m À2 ) was calculated.
Dried plant material was further ground by a CYCLOTEC 1093 sample homogenizer (Foss Tecator, Ho ¨gana ¨s, Sweden).In all cases, we washed the grinding system with bi-distilled water between each other sample grinding to avoid sample contamination.C and N contents were determined by the combustion of 1-2 mg of pulverized dried sample mixed with 2 mg of V 2 O 5 as oxidant.We coupled the combustion to gas chromatography using a Thermo Electron Gas Chromatograph model NA 2100 (C.E.instruments-Thermo Electron, Milan, Italy).For analyses of other elements, dried and ground samples were digested with concentrated HNO 3 and H 2 O 2 (30%, p/v) (MERCK, Darmstadt, Germany) in a microwave oven.Blank solutions were regularly analyzed.To assess the accuracy of digestion and the analytical biomass procedures, standard certified biomass (NIST 1573a, leaf tomato, NIST, Gaitherburg, MD) was used.After digestion, the contents of As, Cd, Cr, Cu, Mo, Ni, Pb, V and Zn were determined using ICP-MS (Mass Spectrometry with Inductively Coupled Plasma) and Ca, Fe, K, Mg, Mn, Na and P were determined using ICP-OES (Optic Emis-sion Spectrometry with Inductively Coupled Plasma).For As analyses, As (V) was reduced to As (III) by a mixture of HCl (30% v/w), KI (1% w/v) and ascorbic acid (0.2% w/v) was added to a digestion solution aliquot of each sample.This solution was then pumped into a gas-liquid separator where it reacted with NaBH 4 (1.3% w/v solution in 0.1 M NaOH) to form arsenic hydrides and analyzed with ICP-MS.

Leaf phenolic, tannin and terpene analyses
Total phenolics content of leaves was determined by the improved Folin-Ciocalteu assay (Singleton andRossi 1965, Marigo 1973).The improvement relative to standard assay was the use of a blank of polyvinylpolypyrrolidone (PVPP).PVPP retains the phenolic compounds avoiding their reaction with Folin-Ciocalteu solution, thereby providing a true blank sample (see Peñuelas et al. 2010b).The absorbances of the samples were determined at 760 nm using a spectrophotometer Helios Alpha (Thermo Spectronic, Cambridge, UK).
Total soluble tannins (Ttan) were extracted from 20 mg of leaf powder with 70% acetone.Tubes containing the sample and the acetone were sonicated three times for 1 min allowing the tubes to cool for three minutes between successive sonications.After centrifugation, the extract was assayed with the butanol/HCl method (Porter et al. 1986) modified according to Makkar Leaf terpene contents were analyzed according to Sardans et al. (2010).Briefly, leaf samples were  v www.esajournals.orga split of 0.5:80, thus allowing only 0.625% of the injected sample to enter the column (HP-5 cross linked 5% PH Me Silicone (Supelco Inc.).Solvent delay was 3 min.The initial temperature of 408C was immediately increased with a ramp of 308C min À1 to 608C.The second ramp was 108C min À1 to 1508C which was maintained for 3 min.The third ramp was 708C min À1 to 2508C which was maintained for 5 min.The carrier gas was helium at 0.7 ml min À1 .A mass detector was used with an electron impact of 70 eV.Identification of monoterpenes was conducted by GC-MS and comparison with standards from Fluka (Buchs, Switzerland), literature spectra, and GCD Chemstation G1074A HP.Calibration with common terpenes a-pinene, D-3-carene, b-pinene, b-myrcene, p-cymene, limonene, sabinene (monoterpenes) and a-humulene (sesquiterpenes) standards (Sigma Aldrich, Gilingham, Dorset, UK).Terpene calibration curves (n ¼ 4 different terpene contents) were always highly significant (r 2 .0.99 for the relationships between signal and terpene contents).

Phylogenetic and statistical analyses
The program Phylomatic (Webb and Donoghue 2005) was used to build a phylogenetic tree of the species studied (Fig. 1).Briefly, this program assembles a phylogeny for the species of interest employing a backbone plant megatree based on a variety of sources involving primarily DNA studies.Our phylogenetic hypothesis was based on the conservative megatree, where unresolved nodes were included as soft politomies.We employed programs in the PDAP package (Garland et al. 1993) to transform the phylogenetic tree into a matrix of phylogenetic distances, and assessed if the studied traits showed significant phylogenetic signal-i.e., the tendency of closely related species to resemble each other due to shared ancestry-employing the randomization procedure in the PHYSIG module developed by Blomberg et al. (2003).For more information see Peñuelas et al. (2010a).These analyses were performed to determine if phylogenetic correction was necessary in subsequent regression analyses.We employed generalized linear models (GLM) to analyze how leaf elemental chemical composition, leaf C:N:P:K stoichiometry, LMA and chemical defenses were related to plant site (Hawaii versus Borneo).We employed ordinary least square regressions (OLS) when the dependent variable did not show significant phylogenetic signal, and phylogenetic generalized least square regressions (PGLS) otherwise.PGLS controls for phylogenetic relatedness by adjusting the expected variance/ covariance of regression residuals employing the matrix of phylogenetic distances (this approach is mathematically equivalent to analyzing the data employing phylogenetically independent contrasts).These analyses were performed in Matlab 7.6.0employing the RegressionV2 module (Lavin et al. 2008).When necessary for statistical analysis, data was log-transformed for it normalization.
Finally, we used principal component analyses (PCA) to explore how the contents of chemical elements discriminated between Hawaii and Borneo plants.We employed a one-way ANOVA to determine how PCA scores obtained for the main components differed between Hawaii and Borneo species.These analyses were performed with Statview 5.0.1 (SAS Institute Inc., Cary, NC, USA) and Statistica 6.0 (StatSoft, Inc.Tule, Oklahoma, USA).

RESULTS
The leaf contents of nutrients and micronutrients tended to be higher in Hawaii than in Borneo plants, although the differences were statistically significant only for Mg, Na and S, and marginally significant for P (Table 2).Regarding trace elements, leaf Zn contents were higher in Hawaii than in Borneo plants whereas Mn, As, Cr and V were higher in Borneo than in Hawaii plants (Table 2).Hawaii plants had higher leaf N and P content per leaf area since LMA was on average 30% lower in the studied species of Borneo compared with those of Hawaii (Table 2), but the leaf C:N, C:P and N:P ratios both in function of leaf weight and leaf area were not different between Hawaii and Borneo plants (Table 2).In the Hawaii species leaf N content scaled as a 0.47 power function of P content (N ¼ 5.9 3 P 0.47 , P , 0.0001) whereas in the Borneo species the scaling factor was 0.30 (N ¼ 4.03 3 P 0.30 , P , 0.0001) (Fig. 2).Similar relationships were found when including Fabaceae species (many legumes harbor colonies of nitrogen-fixing bacteria in their roots) than when they were not included and also when Fabaceae species were considered alone (data not shown).The leaf C:K, N:K and P:K ratio were also higher in Hawaii than in Borneo (Table 2).Several leaf element contents had a significant (C, Ca, Na, Mo, Cr, Cd, Zn) or marginally significant (N, V, Cu, As) phylogenetic fingerprinting (Table 2).N and P content per leaf area and LMA had also phylogenetic fingerprinting (Table 2).LMA was negatively correlated to foliar N content with a steeper slope for Hawaii than for Borneo plants (Fig. 3).
Hawaii species had leaves with higher total leaf phenolics contents (P , 0.0001) (Table 2), total phenolics/N content ratio (P , 0.0001) and total phenolics/ P content ratio (P , 0.0001).No significant differences were observed between Hawaii and Borneo plant species regarding total leaf tannin content.No significant differences were either found for foliar and terpene contents neither when considering only terpene-containing species nor when considering all species studied including species for which we did not detect terpene content (Table 2).However, the percentage of terpene containing species was much higher in Borneo (97%) than in Hawaii (34%) (Chi-square ¼ 64.0, P , 0.0001) (Fig. 4).Leaf phenolics content was negatively correlated with leaf N, P and K content both in Borneo and Hawaii studied species (Fig. 5).
In the Principal component analysis (PCA) conducted with leaf N, P, S, Mg, Na, Ni, Mn, V, Note: Significant effects are indicated in bold type.Abbreviations are: LMA ¼ leaf dry mass per unit area; TP ¼ total phenolics; Tta ¼ total tannins; TT ¼ total terpenes considering all studied species.
Considering only terpene-containing species.
v www.esajournals.orgCr, Zn and As contents, total leaf phenolics, total leaf terpenes and LMA as variables, the scores of the first 2 principal component axes were different between Hawaii and Borneo species (Fig. 6).Most Hawaii species were located towards higher P in the PC1 and towards higher Na, S, Mg and As contents and higher LMA and phenolics contents in the PC2 (Fig. 6).Borneo and Hawaii species had significantly different PC1 and PC2 scores (P , 0.001) (Fig. 6).

DISCUSSION
The results showed that leaf chemical composition differed between Hawaii and Borneo plants.There were differences in important nutrients, micronutrients and trace elements such as P, Na, Mg, S, Cr, Ni among others.Although Hawaii ecosystems can be considered young at a global scale, studies of fertilization in the field have proved that young Hawaiian mountain soils with native forest of Metrosideros polymorpha are both N and P limited (Raich et al. 1996).The older Borneo flora had generally lower nutrient contents including a marginally significant 12% lower leaf P content.This partially supports the geochemical ''soil age'' hypothesis since Borneo flora is more than 56 million years old compared with the 3 million year old Hawaiian flora.
There are several possible explanations for not presenting greater differences in leaf N and P contents in these two contrasting age ecosystems.The different bedrock elemental content, the changes through time of other elements such as K, Mg or S that could also affect differently N and P cycle, and the N and P allocation to different organism functions such as growth, respiration, defence or storage are some of these additional factors that may counteract the ''soil age'' effect.Another counteracting factor might be the increase of some nutrients in the soils as a result of human activities (Vitousek et al. 1997).The proximity and frequency of volcanic eruptions and the loading of volcanic ash could be an important determinant of soil nutrient availability in both forest sites that needs also consideration in future studies.
The ''soil age hypothesis'' also predicts an increase of N:P contents ratio with time that was not observed in this comparison of Hawaii and Borneo floras.In fact, not all the studies have detected general patterns of leaf N:P content ratios between and within the climate areas (Lovelock et al. 2007, Townsend et al. 2007, Vitousek et al. 2010) and there are soil chronosequence studies that have observed that soil available-P was greater in the intermediate ages than in the young and old ages within the time sequence (Crews et al. 1995, Vitousek et al. 1995).The average of leaf N:P content ratios observed in Hawaii plants (21.6 6 1.0) and in Borneo plants, (21.8 6 0.7) are high compared to other studies for woody plants throughout the world that report an average leaf N:P content ratio between 14 and 16 (Koerselman andMeuleman 1996, Gu ¨sewell andBollens 2003).In general leaf N:P content ratios of tropical forest are elevated (Davidson and Howarth 2007) and no relation- v www.esajournals.orgships are observed between leaf N:P content ratios and either latitude or annual mean precipitation within the tropical area (Townsend et al. 2007, Lovelock et al. 2007).He et al. (2008) investigated the changes of N:P ratios of 213 plant species across 199 research sites located in a climatic gradient.They observed that climatic variables had very little relationships with leaf P and N:P ratios.Growing season precipitation and temperature together explained less than 2% of the variation in leaf N:P ratio.In contrast, intersite differences and within site phylogenetic variation explained 55 and 26% of the total variation of leaf P and N:P ratios respectively.In any case, N:P ratios do not always correlate with current N and/or P limitation, and moreover, N and P can be co-limiting in highly diverse ecosystems (Niinemets and Kull 2003).Niklas et al. (2005) proposed that leaf N contents scale as some power function of leaf P contents to maintain optimum growth capacity, showing a relationship of P content with relative plant growth rates (Niklas 2006).They reported that leaf N contents scaled, on average, as 0.75 power of leaf P contents (Niklas et al. 2005, Niklas andCobb 2005).Reich et al. (2010) in a review study of leaf N to P in the major plant groups have reported that N content scales 0.66 to P content.Thus, the results of this study show a different scaling (only ca 0.47 in Hawaii species and even less, only 0.30 in Borneo species), which supports these previous observations only in the sense that when the foliage content of both elements increase, P content increases in higher proportion than N content, but not in the scaling factor.
The results also support the ''leaf economic paradigm'' (Wright et al. 2004) with inverse relationships of LMA with leaf N content both in Hawaii and Borneo floras.The results for both floras also support the ''carbon excess'' hypotheses (Coley et al. 1985, Peñuelas andEstiarte 1998) in the sense that plants with higher nutrient content (N, P and K) and production capacity invest less in C-rich secondary compounds.Thus, the results link the two paradigms because higher leaf economic capacity is linked to less C investment in compounds not directly involved in growth.When comparing plants of the same island, different strategies can be favoured to diminish interspecific competition or to be useful in different successional stages.These strategies range from those plants with low nutrient content and high investment in carbon based defences to those plants with high nutrient  with higher LMA and leaf phenolics content.There are two reasons why these results showing higher phenolics in plants from the younger Hawaiian island with slightly higher soil nutrient content were not expected.Firstly, there has been a shorter evolutionary period in the Hawaii ecosystem for selective pressure to act; and secondly, from an ''economic'' perspective, Hawaii plants have a higher nutrient content compared with Borneo plants.The results thus indicate that higher or lower concentrations of different chemical compounds may not only depend on a defence strategy or soil nutrient content.Other environmental factors such as UV radiation, temperature or drought could explain changes of secondary metabolites concentrations such as phenolics since they are also involved in protection mechanisms in the face of abiotic stressors (Filella and Peñuelas 1999, Lewis et al. 2006, Georgieva et al. 2010).In Hawaii, we had both the dry sites that receive more UV, and the wet sites where it is very cloudy.Since the climate varies tremendously over short distances in Hawaii, we could not clearly discern the UV effect.The results also indicate that environments with higher soil nutrient availability are associ-ated with higher leaf nutrient contents and therefore higher leaf nutritive value and also greater C fixation.
In summary, the results provide considerable evidence of differences in leaf elemental composition (mainly in P, Mg, S, Na, Mn, Ni, Zn, Cr and V), LMA, phenolic content and terpene content distribution between Hawaii and Borneo plant species.There was no difference in N content and N:P content ratio, and only marginally overall higher P in Hawaii plants, giving only partial support to the ''soil age'' hypothesis.These results suggest that the evolution of leaf composition and structure in tropical forest involved not only N and P changes due to evolution of their different soil nutrient availability or to plant production capacity.Other factors may underlie the changes of the leaf composition and morphology throughout time, e.g., change in ecosystem status of other elements with time, or the evolution of plant capacity to respond to biotic and abiotic relationships.Terpenes were found in most Borneo plant species, demonstrating the importance of the evolution time of biotic and abiotic relationships in the acquisition of these compounds.The results of the LMA and pheno-Fig.6. PCA using leaf N, P, Mg, S, Na, Mn, V, Ni, Cr, Zn and As leaf contents, LMA, total leaf phenolics and terpene contents as variables.(Species acronyms are depicted in Fig. 1.)The arrows indicate the mean value for Hawaii and Borneo whole sets of species.They are labelled with a different letter, a or b, when they are significantly different (P , 0.05) or with the same letter when they are not.v www.esajournals.orglics contents of the combined set of 172 different woody plant species of Hawaii and Borneo are consistent with the ''leaf economic spectrum'' paradigm and with the ''carbon excess'' hypotheses when considering all the species of the two sets together, or when considering the species of the two sets separately, but not when comparing the two sets between them, thus showing, once more, the complex multifactorial multiscale nature of chemical ecology.

Fig. 1 .
Fig. 1.Phylogenetic tree of the woody plant species studied in Hawaii and Borneo.The phylogenetic tree was constructed with the program PHYLOMATIC (Webb and Donoghue 2005).Analyses were performed with Mesquite (Maddison and Maddison 2009).Acronyms for each species used in other figures are also depicted.

Fig. 2 .
Fig. 2. Relationships between leaf N content and leaf P content (%, w/w).(Species acronyms are depicted in Fig. 1.)The arrows indicate the mean value for Hawaii and Borneo whole sets of species.They are labelled with a different letter a and b when they are significantly different (P , 0.05) or with the same letter when they are not.The different letters between brackets indicate statically marginal differences (P , 0.10).

Fig. 3 .
Fig. 3. Relationships between leaf mass area (g m À2 ) (LMA) and leaf N content (%, w/w).(Species acronyms are depicted in Fig. 1.)The arrows indicate the mean value for Hawaii and Borneo whole sets of species.They are labelled with a different letter, a or b, when they are significantly different (P , 0.05) or with the same letter when they are not.

Fig. 4 .
Fig. 4. Total terpene contents (mean 6 S.E.mg g À1 ) and the frequency of terpene-containing species in the studied plant species sets of Hawaii and Borneo.

Fig. 5 .
Fig. 5. Relationships between leaf phenolics content and leaf content of N, P and K. (Species acronyms are depicted in Fig. 1.)The arrows indicate the mean value for Hawaii and Borneo whole sets of species.They are labelled with a different letter, a or b, when they are significantly different (P , 0.05) or with the same letter when they are not.

Table 1 .
Soil N and P contents reported by previous studies conducted in Borneo and Hawaii forests.

Table 2 .
Association between species leaf traits and phylogeny (phylogenetic effects were estimated with the PHYSIG randomization procedure) and values of the studied chemical and structural leaf variables (mean 6 S.E.) in Bornean and Hawaiian woody plant species studied.P-values indicate the results of general linear models.OLS ¼ Ordinary least squares regression, PGLS ¼ phylogenetic generalized least squares regression.