Introduction

Phosphorus (P) is a key limiting nutrient for plant growth in many terrestrial ecosystems (Elser et al. 2007; Vitousek et al. 2010; Du et al. 2020). Because plants allocate P to vital functions that include photosynthesis, gene expression, protein synthesis, dark respiration and cell membrane structures (White and Hammond 2008; Hawkesford et al. 2023; Lambers and Oliveira 2019), P allocation among different functions relates to indicators of plant performance such as photosynthetic capacity (Hidaka and Kitayama 2013; Ellsworth et al. 2015; Mo et al. 2019) and cold tolerance (Yoshida and Sakai 1973; Willemot 1975; Yan et al. 2021). P in leaves can be chemically separated into five functional fractions: orthophosphate P (Pi), metabolite P (PM; low-molecular-weight metabolites such as RuBP, NADP, ATP, ADP, sugar phosphates, and pyrophosphate, as well as phytate, which is strictly not a metabolite but rather, a storage compound), nucleic acid P (PN; RNA and DNA), lipid P (PL; phospholipids, predominantly cell membrane lipids), and residual P (PR; phosphorylated proteins and unidentified compounds that contain P) (Chapin and Kedrowski 1983; Kedrowski 1983; Hidaka and Kitayama 2011; Yan et al. 2019). Hereafter, we describe the concentration of each leaf P fraction as [Pi], [PM], [PN], [PL] and [PR]; and their relative proportions per total leaf P as rPi, rPM, rPN, rPL and rPR, respectively. Allocation of P among different functional fractions within leaves varies substantially among species (Veneklaas et al. 2012; Crous and Ellsworth 2020; Lambers 2022; Suriyagoda et al. 2023).

Leaf ‘economic’ traits, such as leaf dry mass per unit area (LMA), light-saturated net photosynthetic rate per unit dry mass (Amass) and leaf lifespan (LL) provide a potential means to link leaf P fractions to ecological strategies. The leaf economic spectrum (LES) is an important trait-related dimension of plant ecological strategy among the world’s species (Reich et al. 1997; Westoby et al. 2002; Wright et al. 2004). The LES is underpinned by a strong positive relationship between LMA and LL, and negative relationships between Amass and each of these traits. Functionally, the LES involves trade-offs in resource allocation between photosynthesis and leaf structures (Chabot and Hicks 1982; Shipley et al. 2006; Lambers and Poorter 2004; Onoda et al. 2004). For example, thicker cell walls increase LMA, which is related to longer LL. Higher-LMA species tend to have a higher fraction of leaf nitrogen (N) in cell walls but a lower fraction in Rubisco, the enzyme responsible for CO2 fixation, often leading to a lower rate of photosynthesis per unit N (i.e. photosynthetic N-use efficiency; Onoda et al. 2004 and 2017; Takahashi et al. 2004; Feng et al. 2009; Guan and Wen 2011).

Whether leaf P fractions relate to the LES is not yet clear. Cell walls contain little P (Aoyagi and Kitayama 2016). However, a trade-off in P allocation may be posited between photosynthesis and P-rich cell membrane lipids (i.e. PL). Greater rPL may cause reduced allocation of leaf P to photosynthesis because PL is largely independent of photosynthetic reactions. Indeed, light-saturated net photosynthetic rate per unit area (Aarea) is unrelated to [PL], unlike other fractions, including [PM], [Pi], and [PR], which are positively related to Aarea (Hayes et al. 2022). A negative correlation between rPL and photosynthetic rate per unit P (photosynthetic P-use efficiency, PPUE) has been reported among broad-leaved woody species from Bornean rainforests (Hidaka and Kitayama 2013), among broad-leaved woody species from forests at early successional stages on the Tibetan Plateau (Lei et al. 2021), and within a domesticated rice variety (Hayes et al. 2022). Therefore, we expect that high rPL is correlated with low PPUE via reduced allocation of leaf P to photosynthesis (Hypothesis 1).

There is a knowledge gap for P-limited ecosystems that are dominated by species with low concentrations of total leaf P ([Ptotal]), e.g., from infertile sandy soils in eastern and western Australia (Wright et al. 2001; Hayes et al. 2014), or ultrabasic soils on the slopes of Mt. Kinabalu, Borneo (Kitayama and Aiba 2002). These species often exhibit high PPUE, high LMA (Denton et al. 2007; Hidaka and Kitayama 2013; Guilherme Pereira et al. 2019), and long LL simultaneously (Denton et al. 2007; Wright et al. 2001). It is reported that the high PPUE of some Proteaceae species on low-P soils is associated with the substitution of phospholipids by galactolipids and/or sulfolipids during leaf development (Lambers et al. 2012). Moreover, rPL was negatively correlated with PPUE but unrelated to LMA among 10 Bornean tree species (including five species on low-P ultrabasic soils and five species on relatively high-P sedimentary soils; Hidaka and Kitayama 2013). However, the relationships between LL and leaf P fractions remain unclear.

It is also possible that variation in leaf N concentration ([N]) is coordinated with leaf P fractions. In general, [N] is positively correlated with [Ptotal], with a power scaling of ~ 2/3 (Elser et al. 2010; Reich et al. 2010; Tian et al. 2018). This relationship can be considered part of the LES (Wright et al. 2005). In principle, [N] may be related to [PN] because PN consists mainly of ribosomal-RNA (rRNA) involved in the turnover of proteins (Veneklaas et al. 2012), which account for the dominant part of leaf N despite a large fraction of leaf N is allocated to chlorophyll (Evans and Clarke 2019). Protein turnover contributes strongly to leaf dark respiration rate (Thornley and Cannell 2000), also part of the LES (Wright et al. 2004). Previous reports have found [PN] to be correlated positively with [N] among 21 Bornean woody species (Hidaka and Kitayama 2011) and across two Banksia species in a pot experiment (Han et al. 2021). It is also demonstrated that [PN] was positively correlated with night-time leaf respiration rate per unit area across six mangrove populations grown in a glasshouse (Yan et al. 2021). Therefore, positive correlations are expected among [PN], [N], and Rdmass (Hypothesis 2). A positive correlation may also be expected between [N] and [PR] because PR contains phosphorylated proteins (Veneklaas et al. 2012).

In this study we aimed to understand how leaf P fractions are related to leaf economic traits among 12 evergreen woody species co-occurring on P-impoverished soils. The Australian continent is an ideal model system to address this question because (1) Australian soils are for the most part extremely impoverished in P due to the geologically ancient and stable landscapes (Kooyman et al. 2017); and (2) Australian plants on P-impoverished soils have a range of strategies in response to P limitation such as: ‘cluster roots’ that help plants ‘mine’ recalcitrant P forms from the soil (Shane and Lambers 2005); sclerophylly that extends leaf lifespans and nutrient residence times, reducing uptake requirements (Beadle 1954; Wright et al. 2002); preferential allocation of P to mesophyll cells, away from the epidermis (Shane et al. 2004; Hayes et al. 2018; Guilherme Pereira et al. 2018); and near-complete resorption of P before leaf abscission (Wright and Westoby 2003; Hayes et al. 2014). We analysed the correlation patterns between leaf P fractions and key leaf economic traits, including LMA, LL, Amass, carboxylation capacity per unit dry mass (Vcmax), dark respiration per unit dry mass (Rdmass), [N], leaf N:P ratio (N:P), and PPUE among 12 common, co-occurring woody species in south-eastern Australia. The 12 species included three gap-dependent pioneer species and nine non-pioneer species, widening sampling along the LES and across different ecological strategy groups. We also quantified P resorption before leaf abscission as this strongly determines P residence time in plants (Aerts 1990; Escudero et al. 1992; Kazakou et al. 2007; Tsujii et al. 2020). P resorption was expressed both as an ‘efficiency’ (i.e. % withdrawn) and as a ‘proficiency’ (i.e. P concentration in senesced leaves, [Psenesced]), describing the degree of completion of the resorption process (Killingbeck 1996). Based on knowledge from previous studies, we proposed and tested the following three hypotheses:

Hypothesis 1 (trade-off in P allocation to photosynthesis versus lipid pools): We expected that PPUE would be negatively correlated with rPL.

Hypothesis 2 (respiration): As protein turnover and its associated rRNA machinery is a major contributor to leaf dark respiration (Thornley and Cannell 2000), we expected that [PN] would drive variation in both [N] and Rdmass.

Hypothesis 3 (resorption): On the premise that different pools of leaf P have different per-unit costs for resorption (Wright et al. 2003) – specifically, the need for hydrolysis of organic P pools would confer higher per-unit resorption costs than that for labile P pools such as Pi (Tsujii et al. 2017) – we expected that rPi would be positively correlated with P-resorption efficiency (PRE), the percentage of P resorbed before leaf abscission (Killingbeck 1996).

Materials and methods

Study sites

The present study was carried out in the Macquarie University Ecology Reserve (33°77S, 151°12E, c. 50 m above sea level), a small patch (c. 5 ha) in Lane Cove National Park (Sydney, south-eastern Australia). The geology of Lane Cove National Park is characterised by Hawkesbury sandstone with occasional Wianamatta shale bands. The concentration of total soil P (total soil [P]) was 78 ± 10 µg g−1 dry mass (mean ± SD, n = 7) (the methodology of sampling and chemical analyses is provided further below), which is much lower than the global average (570 µg P g−1 dry mass) (He et al. 2021) but similar to another sandstone site in Sydney (94 ± 28 µg P g−1 dry mass, n = 5; Wright et al. 2001). However, the concentration of soluble soil P (soluble soil [P]; Bray-1 extraction) was 6.6 ± 0.6 µg P g−1 dry mass (n = 5, see further below for methodology), which is higher than other sandstone sites in Sydney (on average 1.2 µg P g−1 dry mass; Orpheus Butler unpubl. data). At the nearest weather station (33°78S, 151°11E), mean maximum and minimum annual temperature are 22.8 and 11.2 °C (during February 1971–October 1995), while mean annual precipitation is 1157.7 mm yr−1 (during November 1970–February 2023; data from Bureau of Meteorology, Australian Government). The study site is characterised by native bushland vegetation with some human disturbances (e.g., bush walking). Our target species (12 species in total) included: canopy trees (c. 5–10 m tall), including Eucalyptus haemastoma (Myrtaceae) and Allocasuarina littoralis (Casuarinaceae); understorey trees and shrubs, including three Banksia species, Hakea sericea and Lambertia formosa (all Proteaceae) and Angophora hispida and Leptospermum trinervium (Myrtaceae); and gap-dependent pioneers (locally known as “disturbance specialists”), including Dodonaea triquetra (Sapindaceae), Pittosporum undulatum (Pittosporaceae), and Acacia longifolia (Fabaceae). Allocasuarina littoralis and Acacia longifolia have N2-fixing root symbionts (Lawrie 1982; Rodríguez-Echeverría et al. 2009). All species are evergreen woody angiosperms.

Leaf sampling

Fully expanded outer-canopy leaves (or phyllodes for Acacia longifolia) were sampled from five individuals of each species during early summer (November–December 2019). Hereafter, these sampled leaves are described as green leaves. We randomly sampled healthy leaves from several branches. Green leaves included newly-expanded leaves (100% expanded) and leaves produced in the previous year or before. Green leaves were snap-frozen in liquid N2 in the field and brought back to the laboratory at Macquarie University. The samples were freeze-dried and powdered in a wire mill (mesh size 0.08 mm, Retsch ZM200, Haan, Germany) prior to chemical analyses. Recently senesced leaves of the same trees were sampled by shaking the trees (leaves with a fully formed abscission layer easily detached). Senesced leaf samples were oven-dried at 60 °C for 48 h and powdered in the same manner.

Photosynthesis, LMA, and leaf lifespan measurements

Gas exchange measurements were performed using branch samples in parallel with leaf sampling (November–December 2019). Approximately 1 m-long branches were cut with pruning shears from 2–6 mature individuals per species (one branch per individual; on average 3.5 individuals per species) in the early morning (c. 6–8 am during daylight-saving time) to avoid high daytime air temperatures. The proximal ends of sampled branches were re-cut under tap water in a bucket and transported to the laboratory at Macquarie University. A Li-Cor LI-6800 portable photosynthesis system (LICOR, Lincoln, NE, United States) was used to measure light-saturated net photosynthesis and dark respiration. Aarea was measured under 1800 µmol m−2 s−1 of photosynthetic photon flux density (PPFD), with reference to a recent continental-scale study across Australia (Westerband et al. 2022). Dark respiration rate per unit area (Rdarea) was recorded once gas exchange became stable after the light in the photosynthetic chamber was turned off (i.e. PPFD = 0 µmol m−2 s−1). The time for dark acclimation was c. 5–30 min. Stomatal conductance to water vapour (gsw) and internal CO2 concentration in leaves (Ci) were recorded to estimate carboxylation capacity (see the following section). Leaf temperature and CO2 concentration in the chamber were kept at 25 °C and 400 µmol mol−1, respectively. LMA was used to convert area-basis rates to mass-basis rates (i.e. Amass and Rdmass). Arithmetic means by species (hereafter, species means) were used for this calculation because the individuals used for the gas exchange measurements often differed from those for the following LMA measurements (also from those for LL monitoring or leaf P and N analyses). Therefore, we did not have within-species error estimates for Amass and Rdmass.

Four to 34 leaves per individual (on average 12 leaves per individual) were collected and scanned on a flat-bed scanner (CanoScan LiDE220; Canon, Tokyo, Japan). Leaves were collected from five individuals per species during the same period (not always the same day) as the gas exchange measurements were performed (i.e. November–December 2019). The areas of the scanned leaves were measured using image analysis software (Image J; US National Institutes of Health, Bethesda, MD, USA). The scanned leaves were oven-dried at 60 °C for 48 h and their dry mass determined to calculate LMA. The area of needle leaves (i.e. the leaves of Hakea sericea and Allocasuarina littoralis) was adjusted upwards by π/2 on the assumption of circular cross-sections (Wright and Westoby 2002).

Leaf censuses were conducted to estimate LL for five individuals per species following the procedure of Wright et al. (2002). The individuals used for the LL monitoring did not always correspond to those for the LMA measurement or chemical analyses. Five age-sequences of leaves were randomly selected per individual in late 2019 (in total 25 sequences per species). For Acacia longifolia, in total 25 branches were chosen across six individuals due to the limited replications within individuals. The number of leaves was counted along the sequence of leaf age from the youngest leaf (> 25% expanded) on the shoot top to the oldest leaf on the same or older ordered branch, or sometimes on the stem. On average, 178 leaves were counted per sequence (13–772 leaves per sequence). Wires with plastic tags were used to mark each branch (or stem). Six months later, the remaining leaves on the same branch or stem were counted and new sequences with newly-expanded leaves were marked. This census was repeated six months later (i.e. the total monitoring period was 1 year). The leaf census of D. triquetra was conducted every three months because of short LL (often less than six months). LL was defined as the inverse of leaf turnover rate. We estimated average LL across the two census periods. Leaf turnover rate during each census period was calculated by Eq. 1.

$$\mathrm{Leaf}\quad\mathrm{turnover}\quad\mathrm{rate}\quad\mathrm ({\mathrm{year}}^{-1})=\left\{\left({\mathrm{L}}_{\mathrm{d}1}-{\mathrm{L}}_{\mathrm{d}2}\right)/{\mathrm{L}}_{\mathrm{d}1}\right\}/\left\{\left({\mathrm{d}}_{2}-{\mathrm{d}}_{1}\right)/365.25\right\}$$
(1)

where Ld1 and Ld2 represent the number of leaves at the start date (d1) and the end date (d2) of census, respectively. One year was defined as 365.25 days.

Estimation of carboxylation capacity

Carboxylation capacity at 25 °C (Vcmax area; on an area basis) was estimated by the one-point method (De Kauwe et al. 2016). In this estimation, day respiration rate (i.e. non-photorespiratory CO2 evolution in the light) was assumed as 1.5% of Vcmax area. The Michaelis constants for CO2 and O2 and the CO2 compensation point were estimated using the equations provided by Bernacchi et al. (2001) and Crous et al. (2013), respectively. LMA was used to convert area-basis rates to mass-basis rates (i.e. Vcmax).

Soil sampling

In June 2020, seven soil cores (3 cm diameter and 0–15 cm depth beneath the litter layer) were randomly sampled in the ecology reserve, taken to the laboratory, and sieved with 1.18-mm mesh. The sieved soils were air-dried prior to chemical analyses.

Leaf P fraction, total leaf P and N, and soil total and soluble P analyses

Five chemical fractions in green leaves (i.e. Pi, PM, PN, PL, and PR) were determined by the sequential extraction methodology of Chapin and Kedrowski (1983) and Kedrowski (1983), with modifications by Hidaka and Kitayama (2011) and further by Yan et al. (2019). Briefly, 50 mg of powdered sample was extracted with 1 mL of chloroform–methanol-formic acid (12:6:1 v/v/v; CMF) in a 2-mL tube (tube 1) for 30 s. After being centrifuged, the supernatant was transferred into a new 15-mL tube (tube 2). This procedure was repeated three times (in total 3 mL of CMF). The pellet in the tube 1 was further extracted with 1.26 mL of chloroform–methanol-water (1:2:0.8 v/v/v; CMW) for 30 s. The extract in the tube 1 was transferred into tube 2. This extraction was repeated three times. An aliquot of 1.9 mL of chloroform-washed water was added to tube 2, mixed by inverting, and centrifuged to separate into a lipid-rich organic bottom layer and an aqueous upper layer. The lipid-rich bottom layer was transferred to a Kjeldahl tube labelled ‘PL’. The remaining aqueous layer with precipitates was transferred to another tube (tube 3). The pellet in tube 1 was vacuum-dried and then extracted with 1 mL of 85% (v/v) methanol for 30 s. The extract in tube 1 was transferred to tube 3. An aliquot of 1 mL of cold 5% (v/v) trichloroacetic acid (TCA) was added to the residual pellet in tube 1, and extracted for 1 h (1 min shaking, with 9 min intervals). The extract was transferred into tube 3. This extraction was repeated twice. The extract in tube 3 contains Pi and PM and was labelled ‘Pi + PM’. The pellets after the cold-TCA extraction (tube 1) were extracted three times with hot-TCA (2.5% w/v) at 95 °C (1 mL × 3) for one hour each. The hot-TCA extractant and the remaining residue in the tube 1 corresponded to PN and PR, respectively. Samples were kept at 4ºC during the experiments, except for the hot-TCA extraction. An Eppendorf shaker (Thermomixer, at 1400 rpm) was used for the cold extractions. All solid–liquid separations were accomplished by centrifugation with 1028 × g at 4ºC for 10 min. We accepted the data with > 90% of recovery rates (i.e. the percentage of the sum of [P] in the fractions per [Ptotal]).

[PM] was determined by subtracting [Pi] from the concentration of the composite fraction of Pi and PM. During the Pi assay, the activated-charcoal filtering procedure by Dayrell et al. (2022) was used to remove interferences such as turbidity, colouration, and precipitation. Briefly, 25 mg of powdered sample was extracted with 0.5, 0.5, and 0.75 ml of 1% (v/v) acetic acid by shaking at 1400 rpm for 15 s, with 5 min intervals. After the homogenate was centrifuged at 21,000 × g for 15 min, 1.6 mL of the supernatant was transferred to the 2 mL tube with 16 mg acid-washed activated charcoal. The tube was shaken for 30 s, and then the solution with activated charcoal in the tube was filtered through a syringe filter (0.45 μm, diameter = 25 mm). The same procedure was applied also for the standard solutions of Pi. Sample temperature was always kept at 4 °C. Pi in the filtered solution was quantified by the molybdenum blue method (Murphy and Riley 1962) using a spectrophotometer (UV-mini 1240, Shimadzu Co., Kyoto, Japan). The detection wavelength was set at 712 nm to reduce the colour interference by yellow organic matters (Tiessen and Moir 1993).

Powdered leaves and the dried lipid fraction were digested with 2.5 mL of concentrated H2SO4 and 1 mL of 30% H2O2 at 360 °C for 3 h. The liquid samples (i.e. cold- and hot-TCA extractants) and wet solid samples (i.e. the residual fraction) were digested with 2.5 mL H2SO4 for 3 h, after water in the samples was evaporated by a gradual increase in temperature up to 150 °C. P concentrations in the digests were determined by the molybdenum blue method (Murphy and Riley 1962), with 882 nm as the detection wavelength. Leaf [N] was determined by the combustion method using a Vario MICRO cube elemental analysers (Elementar Analysensysteme GmbH, Hanau, Germany).

Total soil [P] was determined by the molybdenum blue method (Murphy and Riley 1962) after digestion with H2SO4 and HClO4 (Olsen and Sommers 1982). Regarding soil soluble [P], 25 ml of Bray-I solution was added to 3.55 g of air-dried soils and shaken by hand for 30 s. Orthophosphate in the Bray-I extract was quantified by the molybdenum blue method (Murphy and Riley 1962), with 712 nm as the detection wavelength. The data for total and soluble soil [P] are provided in the site description section above.

Calculations of PPUE and PRE

PPUE was calculated as Amass per [Ptotal]. PRE was calculated by Eq. 2. Mass loss during senescence was corrected by a mass loss correction factor (MLCF = 0.78; the global average of evergreen angiosperms) (Vergutz et al. 2012).

$$\mathrm{PRE}\left(\%\right)=\left\{1-\left[{\mathrm{P}}_{\mathrm{senesced}}\right]\times \mathrm{MLCF}/\left[{\mathrm{P}}_{\mathrm{total}}\right]\right\}\times 100\left(\%\right)$$
(2)

Statistical analyses

One-way ANOVA with Tukey’s HSD post hoc test was used to test for trait differences among species, with individuals as the units of replication. Pearson’s correlation tests quantified relationships among leaf P fractions and leaf economic traits. In the correlation tests, species means were used as the units of replication. Exceptionally, medians by species were calculated for LL to avoid potential bias from outliers of some individuals with unnaturally long LLs (> 10 years; see Fig. 2). [Psenesced] was log-transformed before the correlation test to normalise the data. We set the significance level at P < 0.05.

The same analyses were also applied to area-basis data to check whether the basis of expression (mass vs area) made a difference to the main trends. The area-based contents of leaf P and N were calculated as the product of species-mean mass-based concentration data and species-mean LMA data. Therefore, there is no replication within species. Hereafter, we describe the area-based contents of total leaf P, N, Pi, PM, PN, PL, and PR, as Ptotal area, Narea, Pi area, PM area, PN area, PL area, and PR area, respectively.

All statistical analyses were done in R version 3.0.1 (R Core Team 2020). The ggplot2 package (Wickham 2016) was used to visualise data.

Results

Variation in the concentration of leaf total P and P fractions

Mean [Ptotal] significantly varied among the species, from 0.26 to 0.74 mg g−1 dry mass (Fig. 1, P < 0.01), with the lowest value in the sclerophyllous shrub Hakea sericea and the highest in a pioneer, Pittosporum undulatum. Indeed, all three pioneer species exhibited higher [Ptotal] than the non-pioneer species. The concentration of each P fraction also varied significantly among species (Fig. 2, P < 0.01). [PR] varied the least among the species (c. two-fold; 0.03–0.06 mg g−1, Fig. 2d), while [Pi] and [PM] varied the most (c. five-fold; 0.05–0.24 and 0.02–0.10 mg g−1, respectively, Fig. 2a and b), with variation in [PN] and [PL] intermediate (c. three-fold among species: 0.06–0.22 and 0.07–0.20 mg g−1, respectively, Fig. 2c and d). Among families, the species in the Proteaceae and Casuarinaceae exhibited the lowest [PL] (the fraction including cell membranes; 0.07–0.09 mg g−1). Proteaceae generally had low [PN] (0.08–0.13 mg g−1). The three pioneer species stood out in having higher [PN] (0.18–0.22 mg g−1) and [PL] (0.14–0.20 mg g−1) than non-pioneer species. [PM] (including substrates and products of photosynthetic enzymes) varied widely between all species combined. High [Pi] was found in a non-pioneer species, Eucalyptus haemastoma, as well as pioneer species. No clear differences in [PR] were found among families.

Fig. 1
figure 1

Box plots showing [Ptotal] for 12 Australian woody species. The colours of each box pertain to families. The central box shows the interquartile range and median by species; the whiskers extend 1.5 times the interquartile range or to the most extreme value. Pairwise significant differences among species are shown as different letters (Tukey HSD, P < 0.05)

Full size image
Fig. 2
figure 2

Box plots showing the concentrations of each P fraction for 12 Australian woody species. The colours of each box pertain to families. The central box shows the interquartile range and median by species; the whiskers extend 1.5 times the interquartile range or to the most extreme value. Pairwise significant differences among species are shown as different letters (Tukey HSD, P < 0.05)

Full size image

Next, we considered the ‘relative’ P allocation to different pools (Fig. 3). On average, the largest relative fraction was rPN (28%), followed by rPi and rPL (27 and 24%). The smallest fractions were rPR and rPM (9 and 12%, respectively). Most notably, rPN and rPL varied the least among species (c. 1.5-fold; 22–33% and 19–27%, respectively, Fig. 3c and d), while rPM and rPR varied the most (c. four-fold; 5–22% and 3–12%, respectively, Fig. 3b and e). rPi varied c. two-fold (17–38%, Fig. 3a). No consistent patterns were apparent between families or within growth forms.

Fig. 3
figure 3

Box plots showing the relative allocation to each P fraction for 12 Australian woody species. The colours of each box pertain to families. The central box shows the interquartile range and median by species; the whiskers extend 1.5 times the interquartile range or to the most extreme value. Pairwise significant differences among species are shown as different letters (Tukey HSD, P < 0.05)

Full size image

Variation in leaf economic traits

Leaf economic traits significantly varied among the species (Fig. 4; P < 0.05). LMA varied 2.6-fold among species, with the lowest mean LMA in the pioneer species Dodonaea triquetra (141 g m−2) and the highest in Eucalyptus haemastoma (362 g m−2) (Fig. 4a). LL varied > six-fold among species, from 0.48 year in Dodonaea triquetra to 3.2 year in Banksia oblongifolia (Fig. 4b). Amass varied > three-fold, from 24.4 nmol CO2 g−1 s−1 (Angophora hispida) to 88.5 nmol CO2 g−1 s−1 (Dodonaea triquetra; Fig. 4c). Vcmax varied > three-fold, from 133 nmol CO2 g−1 s−1 (Hakea sericea) to 449 nmol CO2 g−1 s−1 (Dodonaea triquetra; Fig. 4d). Rdmass varied > four-fold, from 2.7 nmol CO2 g−1 s−1 in Hakea sericea to 11.5 nmol CO2 g−1 s−1 in Leptospermum trinervium (Fig. 4e).

Fig. 4
figure 4

Box plots showing leaf economic traits for 12 Australian woody species. The colours of each box pertain to families. The central box shows the interquartile range and median by species; the whiskers extend 1.5 times the interquartile range or to the most extreme value. For Amass and Rdmass (panels C and D), one data point is provided per species (see Methods). Different letter indicates significant differences among species (Tukey HSD, P < 0.05) for each trait

Full size image

[N] varied > four-fold among species, from 6.2 mg g−1 dry mass (Hakea sericea) to 25.7 mg g−1 dry mass in the N2-fixing pioneer, Acacia longifolia (Fig. 4f). N:P varied < three-fold, with the lowest value in a pioneer, Pittosporum undulatum (16.5) and the highest in the N2-fixing pioneer, Acacia longifolia (40.3) (Fig. 4g). [Psenesced] varied > three-fold from 0.08 mg g−1 (Banksia serrata) to 0.25 mg g−1 in the pioneer, Pittosporum undulatum (Fig. 4h).

PPUE varied c. four-fold among the species (1.7–6.7 mmol CO2 mol P−1 s−1), with the lowest PPUE in Pittosporum undulatum and the highest PPUE in a non-pioneer, Leptospermum trinervium (Fig. 4i). PRE varied from 66% in Hakea sericea to 86% in Eucalyptus haemastoma (Fig. 4j; both non-pioneer species).

Correlation patterns among leaf P fractions and economic traits

The results from correlation tests are summarised in Table 1. First, we report correlation patterns between leaf P fractions and between economic traits, followed by those of leaf P fractions to LMA, LL, Amass (Hypothesis 1); those to Rdmass and [N] (Hypothesis 2); and those to resorption (Hypothesis 3).

Table 1 The results from Pearson’s correlation tests between the concentrations of leaf P fractions and leaf economic traits (above diagonal) and between the relative allocations to each P fraction and leaf economic traits (below diagonal) for 12 Australian woody species. * P < 0.05. † [Psenesced] was log-transformed to normalise the distribution
Full size table

Correlations between leaf P fractions

[Ptotal] was significantly positively correlated with each of [Pi], [PN] and [PL] (r = 0.91–0.95, P < 0.01); however, it was unrelated to [PM] or [PR] (r = 0.19–0.54, P = 0.07–0.56). Significant positive correlations were found among [Pi], [PN] and [PL] (r = 0.83–0.87, P < 0.01). [PM] or [PR] was unrelated to the other P fractions.

Considering relative P allocations, no significant correlations were found among P fractions, except for a negative correlation between rPi and rPM (r = -0.75, P < 0.01). However, rPi and rPM were not statistically independent as [PM] was estimated as the difference between Pi and the composite fraction of PM and Pi. [Ptotal] was correlated only with rPR (r = -0.82, P < 0.01).

Correlations between leaf economic traits

Correlations among leaf economic traits were much as expected. LMA was positively correlated with LL (r = 0.88, P < 0.01). LL was negatively correlated with both Amass and Rdmass (r = -0.72 and -0.76, respectively; both P < 0.01). Amass and Rdmass were themselves positively correlated (r = 0.87, P < 0.01). Vcmax showed the same pattern as Amass. Species with high leaf [Ptotal] also had higher [N] and [Psenesced] (r = 0.68 and 0.83, P < 0.05 and < 0.01, respectively). [N] was positively correlated with Rdmass, and [Psenesced] (r = 0.60, and 0.65, P < 0.05, and < 0.05, respectively). Higher-LMA species had lower [Psenesced] (r = -0.70, P = 0.01).

Correlations of leaf P fractions with LMA , LL , A mass , and PPUE (hypothesis 1)

Leaf P fractions were statistically independent of LMA or LL (r = -0.55–0.37, P = 0.06–0.91). Amass and Vcmax were significantly positively correlated with [PR] (r = 0.62 and 0.63, P < 0.05; see Fig. 5a for Amass) but unrelated to [PM] (r = 0.00 and 0.08, P = 1.00 and 0.80, respectively), despite this fraction containing substrates and products of photosynthetic enzymes. PPUE was significantly positively correlated with rPR (r = 0.70, P < 0.05; Fig. 5b) but, unexpectedly, unrelated to rPM (r = -0.03, P = 0.92) or rPL (including cell membranes; r = -0.07, P = 0.84).

Fig. 5
figure 5

Correlations between Amass and [PR] (panel a); between PPUE and rPR (panel b); and between N:P and rPN (panel c) among 12 Australian woody species. Solid lines indicate significant correlations (P < 0.05). Different colours pertain to different families. Triangles and circles correspond to N2-fixing species and non-fixing species, respectively

Full size image

Correlations of leaf P fractions to Rd mass and to [N] (hypothesis 2)

As predicted, [PN] was positively correlated with [N] (r = 0.83, P < 0.01). Higher Rdmass was positively associated with [N] (r = 0.60, P < 0.05). [PR] was correlated with Rdmass and [N] (r = 0.65 and 0.76, P < 0.05 and < 0.01, respectively).

Correlations between leaf P fractions and P resorption (hypothesis 3)

[Psenesced] (i.e. P-resorption proficiency) was positively correlated with [Pi], [PN], and [PL] (r = 0.76–0.84, P < 0.01). PRE was not significantly correlated with the relative allocation to each P fraction (P = 0.16–0.79), in contrast to our hypothesis 3.

Area basis data

Aarea varied > twofold, from 6.6 µmol CO2 m−2 s−1 (Pittosporum undulatum) to 15.6 µmol CO2 m−2 s−1 (Banksia serrata; Fig. S1a). Vcmax area varied > two-fold, from 34.9 µmol CO2 m−2 s−1 (Hakea sericea) to 81.2 µmol CO2 m−2 s−1 (Banksia oblongifolia; Fig. S1b). Rdarea varied > three-fold, from 0.7 µmol CO2 m−2 s−1 in Hakea sericea to 2.3 µmol CO2 m−2 s−1 in Lambertia formosa (Fig. S1c). There were no significant variations among species in either Aarea, Vcmax area, or Rdarea (Figs S1a–c; P = 0.05–0.06).

Ptotal area varied > three-fold among species, from 64.4 to 204.1 mg m−2 (Fig. S1d), with the lowest value in the low-LMA and low-[Ptotal], Leptospermum trinervium and the highest value in the high-LMA and high [Ptotal], Eucaluyptus haemastoma. Narea varied c. threefold among species, ranging from 1563 mg m−2 in the low-LMA and low [N], Leptospermum trinervium to 4633 mg m−2 in a N2-fixing, Allocasuarina littoralis (Fig. S1e). As for leaf P fraction, PN area and PL area varied the least among species (c. twofold; 19.0–55.6 and 16.5–48.3 mg m−2, Figs S1h and i), while PM area varied the most (> eightfold; 3.4–30.4 mg m−2, Fig. S1g), followed by Pi area and PR area (c. fivefold and > threefold; 15.5–80.5 and 4.4–15.6 mg m−2 respectively, Figs S1f and j). There were no systematic differences in the area-basis leaf P fractions among families or growth forms.

Correlation patterns among these area-basis traits were similar to those among mass-basis traits (Table S1). Exceptionally, Rdarea was not related to leaf P fractions although Rdmass was significantly related to [PR].

Discussion

Among all 12 study species, LL was correlated negatively with Amass and Rdmass. LL was also positively correlated with LMA, representing the LES (Reich et al. 1997; Westoby et al. 2002; Wright et al. 2004) that spanned low-LMA pioneer species to non-pioneers with high LMA. Differences in the concentrations of leaf P fractions were found between the ecological strategy groups, with higher [Pi], [PN] and [PL] in pioneer species; these species showed notably high [Psenesced]. Considering P allocations as a proportion of [Ptotal], the N2-fixing species exhibited higher rPN than the non-fixing species. This is in agreement with a previous report for a pioneer Fabaceae species in Western Australia (Yan et al. 2019), presumably reflecting high capacities for protein synthesis. Presence of N2-fixning species was involved in the positive correlation between rPN and N:P (Fig. 5c), with no significant pattern when analysed excluding the N2-fixing species (r = 0.54, P = 0.22, data not shown). Therefore, growth strategy and N2-fixation capacity described the association between the LES and leaf P fractions. In the following sections, we discuss how leaf P fractions are coordinated with the LES in relation to (1) trade-offs in P allocation; (2) respiration; and (3) P resorption.

Trade-offs in P allocation (Hypothesis 1)

Hypothesis 1 is based on the premise that there is a trade-off in P allocation between photosynthesis and phospholipid-rich cell membranes. rPL is negatively related to PPUE (Hidaka and Kitayama 2013; Hayes et al. 2022; Lei et al. 2021). In contrast to these previous studies, here PPUE was unrelated to rPL, this implies that there was no trade-off in P allocation to photosynthesis versus PL among our study species. One explanation comes from the Proteaceae, in which low expression of rRNA may be involved in high PPUE (Sulpice et al. 2014) as well as membrane phospholipids with low-P galactolipids and/or sulfolipids may substitute for high-P phospholipids (Lambers et al. 2012). Indeed, many Proteaceae species studied here exhibited low [PN] and [PL]. In addition, the non-significant link between [Ptotal] and LMA (r = -0.47, P = 0.12) in this dataset might cause the decoupling between relative P fractions and the LES.

Species with higher Amass had higher [PR]; species with higher PPUE had higher rPR. This suggests that PR reflects some aspect of photosynthetic P metabolism, perhaps via the phosphorylation of photosynthetic proteins (Veneklaas et al. 2012). It was therefore surprising that [PM] was not correlated with Amass, even though this fraction captures the P incorporated into soluble photosynthetic reactants. This may be because PM included phytates, that may be present as storage compounds, rather than metabolically active compounds (Lambers 2022), or involved in cellular signalling (Kumar et al. 2021). Polyphosphate, a storage form of inorganic P reported for Banksia ornata in an Australian heathland (Jeffrey 1964), might be also contained in PM (and perhaps also in PN and/or PR) because it can be partly extracted by 8% cold-TCA solution (Miyachi and Tamiya 1961). Supporting the suggestion that inorganic P storage might be elevated at the expense of PM, there was a negative correlation between rPM and rPi implying that partitioning among P storage pools such as Pi (stored largely in vacuoles; White and Hammond 2008), polyphosphate, and phytate, differs among species. Detailed chemistry of PM and Pi needs to address this possibility, because the P-fractionation method that we used does not distinguish metabolically active forms from storage forms, particularly in PM (Lambers 2022).

We estimated Vcmax (or Vcmax area) to allow for potential stomatal limitation in photosynthesis. There was a significant positive correlation between gsw and Aarea (r = 0.91, P < 0.01; Fig. S2), implying that stomatal limitation constrained Aarea. However, the correlations of leaf P fractions with Vcmax (or Vcmax area) were similar to those with Amass (or Aarea). This suggests that the correlations between leaf P fractions and Amass (or Aarea) were independent of stomatal limitation.

Respiration (Hypothesis 2)

As we expected, [N] was more tightly correlated with [PN] than with [Ptotal] or other P fractions. This may be because the synthesis of rRNA involves the turnover of proteins. PR may correlate with the degree of phosphorylation of photosynthetic proteins (Veneklaas et al. 2012) which is the reason why [PR] and Rdmass were correlated. However, it should be noted that Rdarea was not correlated with PR area. This suggests that the Rdmass–[PR] relationship partially resulted from covariation of each trait with LMA.

P-resorption patterns (Hypothesis 3)

PRE was independent of P-allocation patterns. This contradicts Hypothesis 3 but accords with a previous study for Bornean woody species (Tsujii et al. 2017), in which higher PREs involved greater degradation of organic P (i.e. PN, PL, and PR). Similarly, our target species must have hydrolysed substantial organic P because PRE (often > 80%) was much higher than rPi (max < 40%). In principle, the correlations between P resorption and leaf P fractions are determined by the balance between the unit-cost of resorption-derived P and the unit-cost of soil P (Wright and Westoby 2003); that is, preference would be given to obtaining P from the cheapest sources (i.e. Pi). Indeed, PRE was positively correlated with rPi under relatively P-rich conditions, such as freshwater marshes in northeast China (Mao et al. 2015). We surmise that plants on low-P soils may expend more energy hydrolysing organic P (Hidaka and Kitayama 2011; Tsujii et al. 2017).

Conclusions

The present study is, so far, the most comprehensive survey of leaf P fractions and the LES across wild woody species co-occurring on P-impoverished soils. While the concentration of each P fraction was associated with P-use strategies and N2-fixation capacity, the relative P fractions were largely independent of the LES, at least among the present set of species, most of which tended towards the ‘slow-return’ end of the LES. The Proteaceae species in our study, for example, all exhibited low allocations of P to structural constituents (such as cell membrane lipids) and nucleic acids. Presumably, these strategies allow increased P allocation to photosynthesis which, in turn, contributes to high PPUE, regardless of LMA or LL.

Author contribution statement

This study was originated from the research plan designed by YT and IW. YT and BF led the fieldwork with assistance of ZL. YT, and partly ZL, conducted the lab experiments with contributions from BA and HL. YT conducted the data analyses. YT and IW led writing of the manuscript. All authors contributed to the final writing.