DISCUSSION

Foliar total phosphorus and nitrogen concentrations

Our first hypothesis was that with decreasing soil P availability, the mass-based foliar total P concentrations of both B. attenuata and B. sessilis would decrease in response to decreased P availability, and this was supported. An important finding was that the area-based foliar P concentrations was higher in B. attenuata than in B. sessilis in all three substrates, but there was no significant difference in mass-based foliar P concentration in any of the three substrates. We found the same patterns of mass-based foliar total P concentration for both B. attenuata and B. sessilis grown in all three substrates; the foliar total P concentrations of both species were greatest in sand, and about 35% lower in SLIM. The greatest carboxylate-extractable P concentration was in sand, and the lowest was in limestone gravel (Fig. S1).
In contrast to the similar mass-based foliar total P concentrations in the two species in each substrate, the mass-based foliar total N concentration of B. sessilis was almost twice that of B. attenuata grown in the same substrate. Leaf N concentrations in B. attenuata and B. sessilis grown on SLIM were approx. 20% lower than those in plants grown in sand and SLAT. Our result differ results on Hakea prostrata (Proteaceae) showing that P availability did not influence leaf N concentration (Prodhan et al. , 2016). The foliar total N concentrations of both species were very low compared with those of plants from other environments (Reich et al. , 1991), which reflects the low foliar rRNA concentration in B. attenuata (Sulpice et al. , 2014), and, presumably, in B. sessilis , based on the similar size of their nucleic acid P pools. In our study, the relatively low foliar N concentrations in B. attenuata and B. sessilis indicate that protein concentrations were very low, which implies a low demand for rRNA and, thus, P.
Whilst the leaf N concentrations in both species were low compared with the global average (Reich et al. , 1991), they were distinctly higher inB. sessilis than in B. attenuata . The higher N concentration correlated with greater allocation of P to the nucleic acid fraction inB. sessilis . However, rates of photosynthesis and leaf N concentrations expressed on an area basis were similar for the two species, and hence so was the photosynthetic N-use efficiency (PNUE). Therefore, the ‘extra’ N in B. sessilis on a mass basis was a reflection of a lower investment in sclerenchymatic tissue, as evidenced by its lower LMA). A low ribosome abundance can be expected to decrease the rate of protein synthesis, and hence the protein and N concentrations. Therefore, lower leaf N concentrations in B. attenuata compared with B. sessilis on the three substrates tested is consistent with lower rRNA concentrations and lower rates of protein synthesis. However, since we studied mature non-growing leaves, which do not rapidly change in protein concentration (Kuppusamy et al. , 2014), the faster rate of protein synthesis must have been balanced by faster rate of protein breakdown, and hence protein turnover.
We found that RGR was strongly correlated with leaf N and nucleic acid P concentrations in both species, and that RGR and N concentration inB. sessilis were significantly greater than those in B. attenuata . This supports our second hypothesis that B. sessilis , which exhibits a more opportunistic growth strategy than B. attenuata (Shi et al. , 2020), will have a higher foliar NTotal : PTotal ratio than B. attenuata and invest more P in nucleic acid P to support the higher N concentration. The higher leaf N concentration found here and higher capacity to acquire P (Shi et al. , 2020) in B. sessilisthan in B. attenuata when grown in the more P-limiting SLIM may explain the different distribution patterns of the two species in the environment. This higher capacity to acquire P presumably allows it to colonise and become established on different P-impoverished soils (sand over laterite or over limestone), compared with B. attenuata,which is restricted to deep sand (FloraBase, http://florabase. dpaw.wa.gov.au/).

Foliar traits and P fractions

The different foliar P-allocation patterns combined with differences in LMA between the two species reflects differences in their life history strategies and resource requirements. Plants like B. sessiliswith an r selection life history typically grow fast (Clarkeet al. , 2013) and produce seeds before the next catastrophe,i.e. fire or drought (Bowen & Pate, 2017, Knox & Clarke, 2005, Pate et al. , 1990). This strategy may require relatively greater investment in P-rich rRNA and, thus, ribosomes, to support rapid protein synthesis and turnover, including replacement of damaged proteins (Raven, 2012). A high protein synthesis capacity may provide flexibility to acclimate to variable and changing environments (i.e. shallow sand over laterite or limestone, where water availability may fluctuate) and complete the life cycle quickly. Unlike B. sessilis , B. attenuata with larger seeds (Shi et al. , 2020) and higher LMA, has the ability to resprout from epicormic buds or lignotubers (Groom & Lamont, 2011, Pate et al. , 1991), a strategy associated with a slower RGR (Bowen & Pate, 2017, Knox & Clarke, 2005, Pate et al. , 1990). Thus, selection inB. attenuata was based on a lower investment in nucleic acid P, as well as the ability to allocate more biomass to deep roots compared with B. sessilis (Shiet al. , 2020). Thus, it does not need to grow fast and complete its life cycle quickly (Bowen & Pate, 2017, Knox & Clarke, 2005, Pateet al. , 1990).
The Pi concentration in slow-growing B. attenuata was higher than that in the faster-growing B. sessilis when grown in sand and SLAT, slightly higher than when grown in SLIM. Cell vacuoles serve as a reservoir for excess Pi in most plants, which can then be drawn upon as P availability decreases (Mimura, 1995). Changes in total foliar P concentration with Pi supply generally reflect the accumulation of Pi in vacuoles, which is typically greater in slow-growing species than in fast-growing ones (Güsewell, 2004). Thus, fast-growing species convert Pi into growth-sustaining organic P, rather than accumulate Pi, as in slow-growing species.
The metabolite P concentrations and the proportions of total P in metabolites for B. sessiliswere significantly higher than those for B. attenuata for plants grown in SLIM. Moreover, the PPUE of B. sessilis was greater than that of B. attenuata grown on all substrates. Hidaka & Kitayama (2009) suggested that high PPUE is sustained by the allocation of a greater proportion of P to metabolic P (metabolite P + Pi) than to structural P, as we have showed here. In addition, B. sessilis had a lower LMA than B. attenuata on all substrates tested; however, B. attenuata had higher lipid P concentrations when grown in SLAT and SLIM in response to the lower P availability compared with sand alone. This finding was partially in line with a study that showed that the concentration of structural P is greater in slow-growing plants with high LMA than in fast-growing plants with low LMA (Villar et al. , 2006). In other words, a greater proportion of nucleic acid P, a lower proportion of lipid P and a lower LMA in B. sessilis than in B. attenuata are all traits associated with a higher RGR and shorter leaf life-span (Veneklaaset al. , 2012). Overall, the unique foliar traits of the two species revealed different patterns of P allocation in response to soil P availability and associated with growth strategy that may define the ecological niches in which they are found (Figure 6).