Sowing method
Seeds of Banksia attenuata R.Br. and B. sessilis (Knight) A. R. Mast & K. R. Thiele (purchased from Nindethana Seed Company, King River, Western Australia), were sown on filter paper. One seedling was transferred into each experimental pot on 22nd May, 2016. According to the supplier, the seeds of B. sessilis were collected from a coastal population, growing over limestone near Jurien Bay, Western Australia (30°18 S, 115°3’ E). The provenance of the B. attenuata seed was unknown.
Experimental design
A pot experiment was carried out in a glasshouse at the University of Western Australia, Perth, Australia (31°59ʹ S, 115°53ʹ E) using a randomised complete block design. Glasshouse temperature fluctuated between 13 and 33°C over a whole year, and transmission of radiation into the glasshouse was 60% of natural light. The experiment was designed to explore why B. sessilis is able to grow across a wider range of P-impoverished soil types and maintain a greater relative growth rate (RGR) than B. attenuata by comparing the use and allocation of P among foliar P fractions in the two species. Three soil treatments were imposed, based on washed river sand: sand only, sand + laterite (SLAT), and sand + limestone (SLIM). The substrate total P availability was sand > SLAT > SLIM (Fig. S1, Shi et al. , 2020). The pots (100 mm inner diameter x 400 mm tall PVC cylinder) were lined with plastic bags. For each soil treatment, 3.0 kg substrate was added to the pots. For the SLAT and SLIM treatment, a 100 mm layer of laterite or limestone gravel, respectively, was added 50 mm below the soil surface, and other layers were filled with sand. There were ten replicates for each species in each treatment. Field capacity of soils in each treatment was calculated as [(wet mass – dry mass) / dry mass] × 100%. The pots were watered to a constant weight of 80% of field capacity three times a week. A 20 ml aliquot of basal liquid nutrient solution lacking P and containing (per kilogram of soil): 217.5 mg KNO3; 74 mg CaCl2; 140 mg K2SO4; 80 mg MgSO4.7H2O; 28.9 mg MnSO4.H2O; 10 mg ZnSO4.7H2O; 5 mg CuSO4.5H2O; 0.7 mg H3BO3; 0.5 mg CoSO4.7H2O; 0.4 mg Na2MoO4.2H2O; 20 mg FeNaEDTA, was applied to each pot once every second week.

Photosynthesis measurement

Prior to the final harvest, net photosynthetic rate (Pn) of attached leaves was measured between 10:00 and 11:00 on March 7th, 9th 2017 using a red/blue LED light source (LI-6400, LI-COR Inc., Lincoln, NE, USA). The plants were watered on the day before the photosynthesis measurement. One mature leaf of each plant was measured under a photosynthetic photon flux density of 1500 μmol m-2 s-1 and a CO2 concentration of 400 μmol mol-1. The leaves used for photosynthesis measurement were sampled, and the projected leaf area measured at 200 dpi (Epson 1680, Long Beach, CA, USA) and calculated (ImageJ 1.4, NIH, Bethesda, MD, USA). Leaves were then dried at 70 °C for 72 h to measure dry mass (DM).

Harvest

After 50 weeks of growing in pots, a total of 20 fully-expanded leaves with no visible damage or discolouration were harvested from each plant. The leaves were immediately scanned at 200 dpi to calculate leaf area (LA1), submerged in liquid nitrogen and stored at -80°C. Frozen leaves were freeze dried for seven days (VirTis Benchtop “K”, New York, USA) and dry mass (DM) was determined (DM1). The remaining leaves on each plant were harvested and scanned at 200 dpi to calculate the remaining leaf area (LA2). Total LA = LA1 + LA2. The remaining leaves, stem and roots were separated and dried at 70°C for 72 h. The DM was determined for the remaining leaves (DM2) and for stems plus roots (DM3). Total leaf DM = DM1 + DM2. Total plant dry mass M2 = DM1 + DM2 + DM3. Leaf mass per area (LMA) was calculated as total leaf DM / total LA. Seed weight (W 1) was measured using four lots of 10 (B. attenuata ) or 30 (B. sessilis ) seeds that were dried (70˚C, 48hr) and weighed before calculating the average seed weight. Relative growth rate (RGR) was calculated as (lnM 2-lnW 1) / (T 2-T 1), whereT 1 and T 2 were the dates of sowing and harvesting, respectively, expressed in weeks.

Leaf nutrient analyses

Freeze-dried leaves were ground to a fine powder (Geno/Grinder 2010, Spex SamplePrep, Metuchen, New Jersey, USA). A 50 mg sample was used to determine inorganic P (Pi) described by Yanet al. (2019).
The P allocated to nucleic acids, lipids, small metabolites (Pi + other metabolites) and a residual fraction was determined in a 50 mg portion of powdered leaves using the differential solubility method described by Hidaka & Kitayama (2013), as modified in Yan et al.(2019). Metabolite P is defined here as small metabolite P – Pi.
Phosphorus concentrations in extracts and residues from the above procedures were measured as in described by Matusiewicz & Golik (2004) using a molybdenum blue method (Ames, 1966). Total leaf P is the sum of Pi, nucleic acids, lipids, small metabolites and residual fraction. Total leaf P was confirmed by acid digestion of ground leaf material, followed by Pi assay. Total foliar N concentration was determined by combustion of approx. 30 mg of dried leaf sample (Vario Macro Combustion Analyser, Elementar Analysensysteme GmbH, Langenselbold, Germany).
Leaf area-based P concentration was calculated as Total leaf P mass concentration×LMA; PPUE was calculated as the ratio of photosynthesis rate to area-based P concentration; Leaf area-based N concentration was calculated as Leaf N mass concentration×LMA, and photosynthetic nitrogen-use efficiency (PNUE) was calculated as the ratio of photosynthesis rates to area-based N concentration.

Statistics

The differences in means between B. attenuata and B. sessilis on the same substrate were analysed by t test, while the differences in means within each species across substrate types were analysed by one-way ANOVA with 95% confidence intervals. The relationships of foliar P fractions to total foliar P concentration, leaf mass per area, relative growth rate (RGR) to nucleic acid phosphorus (P), foliar nitrogen (N) and foliar N to nucleic acid P were determined by linear regression analysis, the correlation coefficients were analysed by Student’s T-test. All statistical analyses were performed using the SPSS Statistics 19.0 (SPSS Inc., Chicago, US), and graphed with OriginPro 9.5 (OriginLab Corporation, Northampton, MA, USA).