Understanding the adaptive mechanisms of trees to phosphorus (P) limitation is a major topic in tropical ecology, because the productivity of tropical forests is often limited by P (Cleveland et al. Reference CLEVELAND, TOWNSEND, TAYLOR, ALVAREZ-CLARE, BUSTAMANTE, CHUYONG, DOBROWSKI, GRIERSON, HARMS, HOULTON, MARKLEIN, PARTON, PORDER, REED, SIERRA, SILVER, TANNER and WIEDER2011, Vitousek Reference VITOUSEK1984). Trees under P limitation typically have low concentrations of total foliar P (total foliar [P]) (Hidaka & Kitayama Reference HIDAKA and KITAYAMA2009, Lambers et al. Reference LAMBERS, CAWTHRAY, GIAVALISCO, KUO, LALIBERTÉ, PEARSE, SCHEIBLE, STITT, TESTE and TURNER2012). They show fast photosynthetic rates per unit mass despite the low total foliar [P], consequently exhibiting high photosynthetic P-use efficiency (PPUE; photosynthetic rate per unit foliar P) (Hidaka & Kitayama Reference HIDAKA and KITAYAMA2009). The high PPUE contributes to a high carbon acquisition per unit P absorbed (P-use efficiency) at a plant individual level (Berendse & Aerts Reference BERENDSE and AERTS1987), which may be adaptive to P limitation. Thus, elucidating the mechanisms underlying the high PPUE is essential to understanding the adaptation of trees to P limitation.

One explanation of the high PPUE is a greater relative allocation of P to photosynthesis via the localization of foliar P in photosynthetic cells (Shane et al. Reference SHANE, MCCULLY and LAMBERS2004). Shane et al. (Reference SHANE, MCCULLY and LAMBERS2004) showed that Hakea prostrata (Proteaceae) adapted to P-limitation accumulates P in palisade mesophyll at high P supplies, which contrasts to the general pattern that eudicots accumulate P in the epidermis (Conn & Gilliham Reference CONN and GILLIHAM2010). The accumulation of P in mesophyll was found also in other Proteaceae species (Hawkins et al. Reference HAWKINS, HETTASCH, MESJASZ-PRZYBYLOWICZ, PRZYBYLOWICZ and CRAMER2008, Lambers et al. Reference LAMBERS, CLODE, HAWKINS, LALIBERTÉ, OLIVEIRA, REDDELL, SHANE, STITT, WESTON, Plaxton and Lambers2015), but has not been studied in tropical trees.

We studied the localization of P within a leaf for tropical rain-forest trees. Our study was performed in three forests on the southern slope of Mount Kinabalu, Borneo (6°05′N, 116°33′E, 4096 m asl). These sites are different in soil P availability due to different geological substrates consisting of ultramafic rocks (P-infertile), Tertiary sedimentary rocks (intermediate P) and Quaternary colluvial sedimentary rock deposits (P-fertile) (Kitayama & Aiba Reference KITAYAMA and AIBA2002, Takyu et al. Reference TAKYU, AIBA and KITAYAMA2002). Mean concentration of soluble soil P, which is extracted with hydrochloric-ammonium fluoride solution, is 0.44, 2.32 and 3.31 mg kg⁻¹ dry weight in topsoils (15 cm depth) at the P-infertile, P-intermediate and P-fertile sites (calculated from the data of Takyu et al. Reference TAKYU, AIBA and KITAYAMA2002), respectively. These sites were located at c. 1700 m asl and close to each other (within 10 km). The mean annual temperature was c. 18°C and mean annual precipitation was c. 2800 mm at the P-intermediate site (Kitayama Reference KITAYAMA1992).

Four to five dominant canopy tree species, belonging to widespread genera or families across all the sites (see Table 1), were selected in each forest. Fully expanded sun leaves were sampled from three canopy individuals per species using a catapult. Sampled leaves were freeze-dried (under 20 Pa for 24 h) after having been frozen at −30°C. A portion of these samples was powdered and used for determining total foliar [P]. Freeze-dried leaves of one or two individuals per species were used for examining the spatial distribution of P within a leaf.

Table 1. Species mean ± SE in total foliar [P] for 13 species from three forests on Mount Kinabalu, Borneo. Site mean, and P-value from an ANOVA test among sites in the site means were also reported. The site mean was calculated as the average of species means per site. †The numbers in parentheses represent the sample sizes taken for micro-PIXE experiments per species.

Total foliar [P] was determined by the following procedure. A powdered sample (c. 200 mg) was digested in 5 mL of concentrated H₂SO₄ and 2 mL of 30% (v/v) H₂O₂ at 380°C for 5 h. P concentration in the digest was determined using an inductively coupled plasma atomic emission spectrometer (ICPS-7510, Shimadzu, Japan).

The spatial distribution of P within a leaf was examined using a Micro-PIXE technique (particle-induced X-ray emission analysis with a focused ion beam). Micro-PIXE provides a semi-quantitative elemental map by detecting the gradient of characteristic X-ray emissions among pixels. A freeze-dried sample was dissected into a cross-section (thickness c. 250 μm) using a sliding microtome (TU-213, Yamato Koki, Japan). An elemental map on the cross-section was captured using the nuclear microprobe at National Institute of Radiological Sciences (NIRS), Japan. A proton beam with the energy of 3.0 MeV from the tandem type electrostatic accelerator was used for the micro-PIXE analysis system installed at NIRS-electrostatic accelerator facility (Oikawa et al. Reference OIKAWA, SUYA, KONISHI, ISHIKAWA and HAMANO2015). The beam spot was ~1 μm in diameter. A scanning area for the micro-PIXE analysis system was adjusted according to the size of a cross-section (250–1000 × 250–1000 μm). After the micro-PIXE analysis, the cross-section was examined under a fluorescence microscope, and its image was captured using a microscope camera (DS-Vi1, NIKON, Japan). The following five tissue categories were demarcated on a fluorescence image; i.e. the upper epidermal tissues (cuticle + epidermis + hypodermis), palisade mesophyll, spongy mesophyll and lower epidermal tissues (cuticle + epidermis) and vascular bundles. Boundaries among the tissue categories were visually demarcated as vector data, which were subsequently superimposed on the corresponding elemental map. The sample analysed by a micro-PIXE was partially discoloured by the ion beam, and the discoloured part was used as a marker to superimpose these two images.

Mean relative P concentration per tissue category on a cross-section was estimated as mean X-ray emission per pixel within the regions of each tissue category on an image. Mean X-ray emission per pixel represented the amount of P per pixel, because the strength of X-ray emission correlated with the amount of P. However, X-ray emission per P varied depending on the thickness of cross-section because we could not prepare cross-sections with a uniform thickness. Thus, X-ray emission per pixel was not comparable among cross-sections and we here compared the relative P concentration among tissue categories on a given cross-section only. Mean X-ray emission per pixel in each tissue category was calculated using an image analysis software (Image J, http://rsb.info.nih.gov./ij/). Paired t-tests were performed to test differences in mean relative P concentration between tissue categories. Nineteen cross-sections were used as units of replication in this analysis.

An among-site difference in total foliar [P] was evaluated based on the means of the four or five dominant species per site by ANOVA. The 13 species were used as units of replication in this analysis. Statistical analyses were performed with R version 3.0.1.

Mean total foliar [P] per species ranged from 0.25–0.68 mg g⁻¹ dry weight (Table 1), which was much less than the average of terrestrial plants (over 1 mg g⁻¹) (Han et al. Reference HAN, FANG, GUO and ZHANG2005). This suggests that these species are under P limitation. The extracted data from the elemental maps suggested that most species had higher P concentration in palisade mesophyll than in the other tissues when compared on a single cross-section. Scattergrams in Figure 1 show the relationships of the relative P concentration of the palisade mesophyll (y-axis) to that of the other tissues on the same cross-section (x-axis). Most points were located above the 1:1 line across the scattergrams (Figure 1a–d). When all samples were pooled, mean differences between the Y values (the relative P concentration of the palisade mesophyll) and the paired X values (that of the other tissue on the same cross-section) were significantly greater than 0 across the scattergrams (paired t-test, t = 3.13–4.16, n = 18–19, P < 0.01; Figure 1a–c), except for the vascular bundles (Figure 1d). This indicates greater P concentration in the palisade mesophyll than in the spongy mesophyll and epidermal tissues on a given cross-section. This greater P concentration in the palisade mesophyll is possibly associated with one or more of the following three possibilities: (1) the accumulation of P in mesophyll cells (Shane et al. Reference SHANE, MCCULLY and LAMBERS2004), (2) the dilution of P in the epidermal tissues by thick cell walls, and (3) very little P stored in the vacuoles of epidermis cells.

Figure 1. Relationships of relative P concentration of palisade mesophyll (y-axis) with that of the other tissues (x-axis) for 19 leaf cross-sections of 13 species from three forests on Mount Kinabalu, Borneo. The x-axis represents relative P concentration of upper epidermal tissues (a), spongy mesophyll (b), lower epidermal tissues (c) and vascular bundles (d). The relative P concentration of palisade mesophyll was estimated as mean X-ray emission per pixel within the regions of palisade mesophyll on a single cross-section and compared with that of the other tissues on the same cross-section. Black dotted lines indicate a 1:1 line. Crosses, circles and triangles represent the P-infertile, P-intermediate and P-fertile sites, respectively. Paired t-tests were performed between y values and the paired x values in each case.

In contrast to the localization of P across most species, mean total foliar [P] per site based on dominant species was significantly lower at the P-infertile site than the other sites (ANOVA, P < 0.01; Table 1). This low total foliar [P] may reflect the following three mechanisms: (1) the substitution of phospholipids with non-phospholipids; (2) a slower protein synthesis; and (3) a tougher leaf structure. First, trees may have low P concentrations in all cells through the substitution of phospholipids with non-phospholipids in membranes. Lambers et al. (Reference LAMBERS, CAWTHRAY, GIAVALISCO, KUO, LALIBERTÉ, PEARSE, SCHEIBLE, STITT, TESTE and TURNER2012) suggested that the Proteaceae species adapted to P limitation substitute phospholipids with galactolipids and sulpholipids. Second, the low total foliar [P] may be linked with a slower protein synthesis (Sulpice et al. Reference SULPICE, ISHIHARA, SCHLERETH, CAWTHRAY, ENCKE, GIAVALISCO, IVAKOV, ARRIVAULT, JOST, KROHN, KUO, LALIBERTÉ, PEARSE, RAVEN, SCHEIBLE, TESTE, VENEKLAAS, STITT and LAMBERS2014). Hidaka & Kitayama (Reference HIDAKA and KITAYAMA2011) showed that low total foliar [P] was correlated with low concentrations of nucleic acid P. The low nucleic acid P may be linked with low concentrations of ribosomal-RNA, leading to a slower protein synthesis (Sulpice et al. Reference SULPICE, ISHIHARA, SCHLERETH, CAWTHRAY, ENCKE, GIAVALISCO, IVAKOV, ARRIVAULT, JOST, KROHN, KUO, LALIBERTÉ, PEARSE, RAVEN, SCHEIBLE, TESTE, VENEKLAAS, STITT and LAMBERS2014). Third, trees may have a tougher leaf structure including a thick cuticle and abundant sclerenchyma (Choong et al. Reference CHOONG, LUCAS, ONG, PEREIRA, TAN and TURNER1992, Onoda et al. Reference ONODA, RICHARDS and WESTOBY2012). This leaf structure contributes to a high P-use efficiency by prolonged leaf longevity (Escudero et al. Reference ESCUDERO, DEL ARCO, SANZ and AYALA1992), which may be adaptive to P-deficiency. Mean ± SE upper cuticle thickness per site based on dominant species was significantly greater at the P-infertile site (9.0 ± 1.3 μm) than at the P-intermediate and P-fertile sites (4.9 ± 0.9, and 4.7 ± 0.2 μm, respectively) (ANOVA, F = 6.67, n = 13, P < 0.05; Tsujii & Kitayama, unpublished data). Sclerenchymatic cells, which have a greater specific density (Poorter et al. Reference POORTER, NIINEMETS, POORTER, WRIGHT and VILLAR2009), might be more abundant in the leaves of the P-infertile site, because leaf mass per area (LMA) of the species at the P-infertile site was greater (Hidaka & Kitayama Reference HIDAKA and KITAYAMA2011). Sclerenchymatic cells contain a greater portion of cell walls, which are devoid of P (Lambers et al. Reference LAMBERS, CLODE, HAWKINS, LALIBERTÉ, OLIVEIRA, REDDELL, SHANE, STITT, WESTON, Plaxton and Lambers2015).