In the frame of the priority program SPP 1685 “Ecosystem Nutrition, Forest Strategies for limited Phosphorus Resources” (Lang et al., 2016), two temperate beech forest sites were selected for the purpose of this study. The site “Bad Brückenau” (BBR) is located in Bavaria, southeastern Germany in a midrange mountain area of the Rhön (50°35′N, 9°92′E). The potential natural vegetation at the site BBR is Hordelymo-Fagetum (Lang et al., 2017). The forest stand comprises 99% Fagus sylvatica L. and 1% Acer pseudoplatanus L. Mean stand age of beech is 137 years with an average height of 26.8 m, a mean (breast height) diameter of 36.8 cm and a tree density of 335 individuals per hectare. At this site, mull-like Moder forest floors on dystric skeletic Cambisols developed on alkaline igneous rocks/metamorphites (Haußmann and Lux, 1997). The order of organic soil horizons (IUSS WG WRB, 2015) was as follows, for LUE: Oi (3 cm) – Oe (4 cm) – Oa (1 cm) and for BBR: Oi (3 cm) – Oe (3 cm). Mean annual air temperature and the mean sum of annual precipitation are 5.8°C and 1031 mm, respectively (Level II ICP-forest monitoring; Lorenz, 1995). At the site BBR yearly atmospheric N deposition was 13.8 kg N ha^–1 (measured between 2009 and 2012; personal communication Dietrich, H.-P., LWF-Bayern, Germany). For more detailed site characteristics see Lang et al. (2017). During the experimental course, soil temperature at a depth of 0.05 m was, on average, of 13.0 ± standard deviation (SD) 0.4°C. During this period of time, 26.6 mm throughfall were reported (personal communication Dr. Meesenburg, NW-FVA, Germany). The site “Lüss” (LUE) is located in Lower Saxony, northwestern Germany in the Lüneburg Heath (52°83′N, 10°36′E). The potential natural vegetation at the site LUE is Luzulo-Fagetum (Lang et al., 2017). The forest stand comprises 91% F. sylvatica and 9% Quercus petrea (MATTUSCHKA) LIEBL. Mean stand age of beech is 132 years with an average height of 27.3 m, a mean (breast height) diameter of 27.5 cm and a tree density of 480 individuals per hectare. Organic layers and soil types are classified as a mor-like Moder forest floor on hyperdystric folic Cambisols developed from Pleistocene sands. The mean annual air temperature is 8°C and the annual sum of precipitation amounts to 779 mm (Level II ICP-forest monitoring; Lorenz, 1995). During the experimental course, the mean soil temperature mean at a depth of 0.06 m was 9.3 ± SD 0.1°C and 23.8 mm throughfall were observed (personal communication Mr. Dietrich, LWF-Bayern, Germany). At the site LUE, yearly atmospheric N deposition was 10 kg N ha^–1 (measured between 2009 and 2012; personal communication Dietrich, H.-P., LWF-Bayern, Germany). Sites were selected according to their bioavailability of P, thus showing the ecosystem adaptions to different parent materials. We assume the influence of the climate conditions to be important but not fundamental for the development and adaptions of the forest ecosystems on these sites.

At both sites, a full-factorial fertilization experiment was established in April-May 2016. The experiment comprised a control and three fertilization treatments: N addition (150 kg N ha^–1 year^–1 as NH₄NO₃), P addition (50 kg N ha^–1 year^–1 as KH₂PO₄), N+P addition (combination of N and P addition). Control and treatment plots (20 m × 20 m) were established in triplicates which were randomly distributed in the stands while keeping a minimum distance of 20 m between plots (Figure 1). Each plot contained at least two individuals of adult beech trees (F. sylvatica L.) with a diameter at breast height of 46.8 ± SD 3.65 cm at the site BBR and 48.8 ± SD 5.81 cm at the site LUE and an approximate age of 120 years.

The control and the N addition treatments received KCl to account for the addition of K in P-containing treatments. The addition of N was split into five parts (30 kg N ha^–1 each) in spring, summer, and autumn to account for the seasonal demand of trees and to reduce leaching losses. The addition of P took place once a year along with the N addition in spring. The fertilizer addition was carried out in dissolved form by means of garden sprayers. The experiment was conducted in July/August 2017. We acknowledge that by restricting the fertilization to N and P, we did exclude the potential limitation of forest growth by elements other than N and P which was speculated by Jonard et al. (2012). We tried to simulate a conservative estimate of maximum N input for the next decade of atmospheric N deposition. Nevertheless, similar to many other forest fertilization experiments, our fertilization approach did not take into account direct interactions between the fertilizer and the canopy which would have occurred had we manipulated atmospheric deposition directly.

The isotope labeling took place June 19–21, 2017 at the site BBR and June 12–14, 2017 at the site LUE. We used an area of 10.8 m² around each tree individual for the application of ¹⁸O-enriched water. This area was formed by two circles with an identical center in the middle of each tree with radii of 0.4 and 1.9 m. We prepared ¹⁸O-enriched water by diluting tap water with ¹⁸O-enriched water (>98 atom% ¹⁸O, Hyox ¹⁸O, rotem, Arava, Israel) aiming at δ¹⁸O values of soil solution of around +40‰. Assuming thermodynamic equilibrium fractionation (Chang and Blake, 2015), δ¹⁸O_Pi values of +61.7‰ would result (see calculations for realized δ¹⁸O_W values of soil solution in Supplementary Figure S1). We applied the ¹⁸O-enriched water using a syringe (100 ml, Vektenxi, amazon.de) that was composed of a 12 cm long needle with a closed tip and five outlet holes (1 mm diameter). We injected the ¹⁸O-enriched water into the Oe horizon and the injection depth was adapted to the site-specific forest floor structure in order to be centered in the Oe horizon. To assure homogeneous application of the label, the area was divided into eight subareas. 60 injections of 2.5 ml water were allocated randomly to each subarea. In total, 1.2 l of ¹⁸O-enriched water was applied in the prescribed area around each tree individual and this equals 2.1% (BBR) and 1.7% (LUE) of the average soil water content in the organic layers. To check for resulting labeling of soil water, organic layer samples were taken and stored in gas-tight vials until isotope analysis.

In total, we sampled 48 trees (2 sites × 4 treatments × 2 time steps × 6 replicates). Two time steps were necessary to cover the potential duration until ¹⁸O-enriched phosphate originating from incorporation of ¹⁸O from the water label in soil were taken up by the trees. To follow the uptake of ¹⁸O-enriched phosphate, we extracted xylem sap. To this end, branches (diameter of 0.8–2 cm) were harvested by tree climbers from the middle canopy. Directly after harvesting, we removed 2 cm of bark and cambium to prevent contamination with phloem constituents and tightly pulled a PE hose over the twigs. The twigs were placed in the Scholander pressure bomb (Soilmoisture 3000 Series Plant Water Status Console). The pressure was applied as N₂ gas (Grade 3). This procedure was repeated until a minimum amount of 30 ml xylem sap was extracted. Directly after sampling the xylem sap was filtered through 1.2 μm PET syringe filters (CHROMAFIL Xtra PET-120/25, MACHEREY-NAGEL GmbH & Co. KG, Düren, Germany). The samples were stored in a portable fridge at 4°C. Aliquots were split for δ¹⁸O_W (1 ml) and δ¹⁸O_Pi analyses.

Because some organic P compounds might be rapidly transformed into Pi in xylem sap, we processed xylem sap samples each day directly after returning from the field. We purified Pi following the protocol of Weiner et al. (2011) and added anion exchange resin membranes (VWR International GmbH, Bruchsal, Germany) to the xylem sap samples. After 16 h of shaking, the membranes were removed, rinsed with H₂O and eluated by HNO₃. Eluates were again stored in a portable fridge at 4°C until preparation for isotope analysis in the laboratory. In the laboratory, the eluate was used for precipitation of silver phosphate as described by Weiner et al. (2011). In brief, the mineral precipitation and dissolution of ammonium phospho-molybdate was followed by mineral precipitation and dissolution of magnesium ammonium phosphate. After removal of cations and proof of the absence of chloride, silver phosphate was precipitated.

Concentration of inorganic P (Pi) in xylem was determined spectrophotometrically with a continuous flow analyzer (CFA, AA3, XY2, Seal-Analytic, Norderstedt, Germany) at λ = 660 nm, using the method of Murphy and Riley (1962). Concentration of total P in xylem as well as total dissolved P concentrations in soil extraction solutions were measured by means of Inductively Coupled Plasma/Optical Emission Spectrometry (ICP-OES, PerkinElmer Optima 5300 DVGermany) at λ P 213.617. Organic P concentrations (Po) were calculated as the difference between total P and Pi concentrations (Toor et al., 2006). Limits of detection were 0.02 mg P l^–1 for CFA and 0.05 mg P l^–1 for ICP-OES.

Soil water for O isotope analysis was gathered by cryoextraction (Orlowski et al., 2013). δ¹⁸O values of soil water extracts as well as of xylem sap were measured by cavitiy ring down spectroscopy (PICARRO Santa Clara, California, USA). The analysis of O isotope ratios of silver phosphate was carried out by means of a TC/EA (PYRO cube) coupled in continuous flow to an IRMS (IsoPrime100, both Elementar Analysensysteme; Hanau, Germany). Three triplicate subsamples (if enough silver phosphate was available) of each sample were weighed in silver capsules together with a small amount of glassy carbon powder to promote CO formation during combustion (Thermofisher scientific, Type 1, Kandel, Germany). The purity of the precipitated silver phosphate was ensured by the close match of standards and samples regarding the regression of O yield on analyte weight (Supplementary Figure S2). Calibration and drift-corrections were accomplished by repeated measurements of two international benzoic acid standards, IAEA 601 and IAEA 602 (δ¹⁸O = +23.3 and +71.4‰, respectively; distributed by the International Atomic Energy Agency, Vienna, Austria), and one internal Ag₃PO₄ standard (δ¹⁸O = +10.2‰). The standard deviation of triplicate measurements was ±0.6‰.

General site effects were derived from a comparison of control plots between sites. In these cases, differences between sites were tested using a student’s t-test. For single time steps and the complete design (i.e., including all treatments), we used a repeated measures ANOVA with site as between-subject factor and treatment (fertilization) as within-subject factor. Furthermore, we tested whether calculated differences between treatments (Δ) significantly deviated from zero based on a t-test against zero. For time series, a repeated measures ANOVA was conducted with site as between-subject factor and time step as within-subject factor. If the prerequisite for statistical analyses was violated (non-normal distribution), Wilcoxon tests were performed instead. Statistical analysis was carried out by IBM SPSS Statistics 22.

Across control plots of both sites, Pi and Po concentrations ranged from 0.1 to 4.8 and 1.2 to 6.7 mg P l^–1, respectively. On average, we found Pi concentrations of 1.1 ± standard error (SE) 0.4 mg P l^–1 and Po concentrations of 3.4 ± SE 0.6 mg P l^–1. In general, the variability of P concentrations in xylem was high at both sites (coefficient of variation [CV] 41% in BBR and 63% in LUE). The mean Pi:Po ratio was 0.33 ± SE 0.08. We found significantly higher Ptot, Pi, and Po concentrations and Pi:Po ratios in BBR than in LUE (Figure 2). In LUE, Po concentrations were 4.3 times higher than Pi concentrations (Figure 2).

The application of ¹⁸O-enriched water in the Oe horizon increased δ¹⁸O_W values of soil solution to +41 ± SE 5‰ across sites and sampling dates and this represented a substantial increase as compared to δ¹⁸O_W values before label application (BBR: −8.2‰ ± SD 0.9‰, LUE: −9.1‰ ± SD 0.5‰, Figure 3A). This increase was similarly observed in BBR and LUE (compare boxplots and dashed red line in Figure 3A). Despite the successful and consistent ¹⁸O labeling of soil water at both sites, δ¹⁸O_W values in xylem sap after labeling did not differ from those observed before label application in BBR (Figure 3B). By contrast, δ¹⁸O_W values of xylem sap in LUE were higher after than before labeling and significantly higher than in BBR at both sampling dates (Figure 3B). In LUE, δ¹⁸O_W values of xylem sap significantly decreased with time (Figure 3B).

At both sampling dates after the application of ¹⁸O-enriched water (2 and 14 days), δ¹⁸O_Pi values of xylem sap were significantly different to the background δ¹⁸O_Pi values in xylem sap (mean δ¹⁸O_Pi background = −9.15‰ in BBR and −2.26‰ in LUE). In BBR, the difference significantly decreased from 26.1 ± SE 1.7‰ to 7.8 ± SE 2.4‰ during the period from 2 to 14 days after label application (Figure 4). In LUE, the difference was 9.8 ± SE 3.1‰ and 13.9 ± SE 2.2‰ 2 and 14 days after label application, respectively (Figure 4). Notably, the variation in δ¹⁸O_Pi values of xylem sap was smallest 2 days after label application in BBR (coefficient of variation [CV] 16% as opposed to 75% after 14 days), whereas in LUE δ¹⁸O_Pi values of xylem sap varied the least 14 days after label application (CV 78 and 39% after 2 and 14 days, respectively).

We found only few effects of fertilization on P concentrations in xylem sap. In BBR, N+P addition significantly increased Ptot and Pi concentrations while P addition resulted only in increased Pi concentrations in xylem sap of fertilized as compared to non-fertilized trees (Figures 5A,B and Supplementary Figure S3). As a consequence, the Pi:Po ratio of xylem sap was significantly higher in the P-fertilized than the non-fertilized trees in BBR (Figure 5D). Total P concentrations in xylem sap of fertilized trees were increased as compared to non-fertilized trees by P addition in LUE (Figure 5, but see Supplementary Figure S3). δ¹⁸O_Pi values of xylem sap were enriched in ¹⁸O irrespective of fertilizer treatments in BBR two days and in LUE 14 days after labeling (Supplementary Figure S4). Strikingly, only P fertilization (P and NP) resulted in an ¹⁸O-enrichment of phosphate in xylem sap on the other sampling dates at each site (14 days in BBR and 2 days in LUE, Supplementary Figure S4). However, δ¹⁸O_Pi values of xylem sap were not affected by fertilization treatments (Supplementary Figure S5).

The ranges of 0.1 to 4.8 mg P l^–1 (Pi) and 1.2 to 6.7 mg P l^–1 (Po) in our study were in the lower range of Pi and Po concentrations of xylem sap reported in the literature (Pi: 0.5 to 70 mg P l^–1; Po: 0.5 to 80 mg P l^–1) (Bollard, 1960; Saur et al., 1995; Prima-Putra and Botton, 1998; Netzer et al., 2017). In line with our hypotheses, xylem sap total P concentrations were 2.3-fold higher in BBR than in LUE (Figure 2) and this was corroborated by 10-fold higher NaHCO₃-P-extractable P concentrations in soil in BBR as compared to LUE (Hauenstein et al., 2018). Therefore, site-specific P availability in soil was reflected in P concentrations of xylem sap. The significantly higher Pi:Po ratios in xylem sap at the P-rich site BBR compared to the P-poor site LUE indicates an intensified uptake of inorganic P from soil solution. By contrast, the low Pi:Po ratios at the P-poor site LUE could be indicative of the uptake of Po. In line, Scheerer et al. (2019) showed that small P-containing organic molecules can be taken up directly (i.e., intact) by trees. Alternatively, low Pi:Po ratios might be indicative of tree-internal P cycling. For example, Netzer et al. (2017) described higher Po concentrations in xylem at P-poor sites postulating intensified tree internal P remobilization and enhanced internal P cycling efficiency. This is further corroborated by a larger difference in P concentrations between leaf litter and fresh leaves and thus, higher P resorption, in LUE as compared to BBR (Lang et al., 2017). In summary, not only transformations of Po to Pi in organic layers (Hauenstein et al., 2018) but also Po uptake and tree-internal cycling are decisive for P nutrition of trees at the P-poor site.

Because neither site nor date influenced δ¹⁸O values of soil solution (Figure 3A) and because of constant temperatures as well as negligible rainfall events during the experimental course, we can exclude both the ¹⁸O-enrichment of soil water via evaporation (Kendall and McDonnell, 1998) and the dilution of the ¹⁸O tracer in soil solution via rainfall. Despite the continuous ¹⁸O-enrichment of soil solution water at both sites, we observed an ¹⁸O-enrichment of xylem sap in LUE only (Figure 3B). Therefore, trees in LUE took up water from the organic layer to which we added the isotope label. In line, >40% of the total fine roots (down to 1 m) are located in the organic layers in LUE (Hauenstein et al., 2018) and trees rely on the organic layer for water uptake. By contrast, 9% of the total fine roots can be found in the organic layer in BBR (Hauenstein et al., 2018) and thus, water uptake to a main proportion likely originates from deeper soil layer that we did not label. We observed differences in δ¹⁸O values of xylem water between sampling dates in LUE (Figure 3). These might indicate water uptake from deeper non-labeled soil layers in LUE as well. However, the absolute difference of 3.3‰ is too small to pinpoint shifting patterns during the experimental course.

We found significantly increased δ¹⁸O_Pi values of xylem sap at both sites (Figure 4). Notably, at both sites differences in δ¹⁸O_Pi values after and before labeling were sufficiently large (+15‰) to exceed potential “small” isotope fractionation of ca. 3‰ associated with active transporter-driven P uptake (Blake et al., 2005; Gross and Angert, 2015). Furthermore, the incorporation of ¹⁸O into phosphate in xylem sap might originate from tree-internal biological processes. According to textbook knowledge, neither do microorganisms colonize the protoplast-free tracheids enabling xylem sap flow nor do extracellular enzymes exist in xylem sap. Therefore, xylem-internal biological processes resulting in an incorporation of ¹⁸O-enriched xylem water into phosphate in xylem sap are highly unlikely. The only plausible explanation of a tree-internal ¹⁸O incorporation into phosphate would be an exchange between phloem – containing P-containing metabolites produced in metabolically active plant organs such as leaves – and xylem. During metabolic reactions in leaves, O atoms from cell water are incorporated into phosphate (Pfahler et al., 2013). Because in our study, xylem sap water was enriched in ¹⁸O, this biological incorporation would be visible in xylem if there was a connection between phloem and xylem. We did not measure δ¹⁸O values of cell water or δ¹⁸O_Pi values in leaves. However, we can roughly estimate δ¹⁸O_Pi values in metabolically active P fractions in leaves. Leaf cell water is enriched in ¹⁸O caused by preferential evaporative losses of H₂¹⁶O (Dongmann et al., 1974). Gan et al. (2002) found a difference between xylem and leaf mesophyll water ranging from +22.1 to +24.4‰ in cotton leaves. Based on δ¹⁸O values of xylem sap water in our study (Figure 3), δ¹⁸O values of leaf water of +14‰ (BBR) and +21‰ (LUE) would result. During metabolic reactions in leaves, all four O atoms in phosphate molecules are exchanged with those of cell-water-O and this reaction is accompanied by a temperature-dependent equilibrium fractionation (Pfahler et al., 2013) similar to observations of microbial metabolism (Blake et al., 2005). Applying the equation provided by Chang and Blake (2015) for a temperature of 13 and 8.5°C (mean soil temperature in 5 cm depth during the experiment in BBR and LUE, respectively), δ¹⁸O_Pi values in leaves of 39‰ in BBR and 47‰ in LUE would result. Because estimated δ¹⁸O_Pi values in leaves are far from those we observed and because we are not aware of studies on an exchange between xylem and phloem, we postulate that ¹⁸O-enriched phosphate in xylem sap originates from the uptake of phosphate previously biologically cycled in soil rather than from tree-internal biological cycling of P. This assumption is in line with our finding that the biological incorporation of ¹⁸O from ¹⁸O-enriched ambient water into phosphate was visible at both sites and was in line with the ¹⁸O enrichment of soil water but not of xylem sap water.

Interestingly, for the P-rich site BBR, Pi and water uptake seemed to be decoupled since phosphate contained the ¹⁸O label while xylem sap did not (Figures 3, 4). A decoupling of P and water uptake from different soil layers is plausible due to the energy-dependent active Pi uptake (Schachtman et al., 1998). Trees can adapt the active Pi uptake to changing conditions, for example to the seasonal variation in P supply (Netzer et al., 2017). Therefore, an active uptake of Pi is not necessarily linked to water uptake. Conversely, the location of water and P uptake seem to be identical, namely the organic layer, at the P-poor site LUE since the ¹⁸O enrichment of soil water, xylem sap and phosphate in xylem sap match (Figures 3, 4). This is corroborated by the site-specific fine root distribution (Hauenstein et al., 2018) and in line with our hypotheses identifying distinct ecosystem adaptions to P-poor sites. Beside water acquisition, this might also be an adaption of trees to P limitation, by expanding their fine roots to locations of P acquisition. Nevertheless, we cannot pinpoint the exact location of water uptake due to a lack of data on δ¹⁸O_W values of deeper soil layers.

Since the δ¹⁸O_Pi values of xylem sap in BBR exhibit little variability and are highly enriched in ¹⁸O already 2 days after label application (Figure 4), ¹⁸O from ambient water seem to be rapidly incorporated into phosphate which was subsequently taken up by trees in BBR. This short time scale is in line with the time scale of minutes to a few weeks until all four O atoms of a phosphate molecule are exchanged by O from isotopically labeled ambient water (Gross et al., 2015; von Sperber et al., 2017; Helfenstein et al., 2018). The incorporation of ¹⁸O from ambient water into phosphate is mediated by enzymes such as pyrophosphatases in microorganisms (Blake et al., 2005), or mono-/diesterases (Liang and Blake, 2006; Sperber et al., 2014). The fast occurrence of the ¹⁸O label in xylem sap of BBR is corroborated by higher microbial activity (Heuck and Spohn, 2016; Pistocchi et al., 2018) as well as higher phosphatase activity (Lang et al., 2017) reported for BBR as compared to LUE. At the P-poor site LUE, a distinct occurrence of the ¹⁸O label occurred after 14 days (Figure 4). The slower appearance of the ¹⁸O label corresponds to a slower forest floor turnover rate of 39 years in LUE compared to 5 years in BBR (Lang et al., 2017) illustrating a hampered biological activity in LUE. The slow turnover rates are linked to a greater contribution of fungi as compared to bacteria in organic matter decomposition. This is corroborated by Zavišić and Polle (2018) who highlights the importance of ectomycorrhrizal fungi (EMF) for P uptake by beech trees on P-poor sites. Similarly, Zavišić et al. (2016) pinpoint the intense colonization of root tips by EMF in the organic layers. In summary, high availability of P in soil is reflected in tree P nutrition in acquiring ecosystems such as BBR. In these ecosystems, because of high biological activity, P supply of trees originates from biological processes and is not necessarily linked to water supply. The source of P supply is similar in acquiring and recycling ecosystems, however, the retarded biological processes result in less favorable P nutrition of trees in recycling as compared to acquiring ecosystems.

We found that P addition alone and in combination with N increased Pi concentrations in xylem sap of BBR (Figure 5). Because data on tree growth or biomass are not available, it remains unclear whether the increased P uptake we observed for xylem sap was driven by fertilizer-induced increased biomass production or by growth-independent “luxury” P uptake. Other studies indicate that increased P uptake is mostly associated to growth-related increased P demand of trees (Goswami et al., 2018). At first sight, one might guess that the increased P uptake derives from the fertilizer source itself. If this was the case, the O isotope signature of the fertilizer would predominate without shifts caused by biological cycling. However, ¹⁸O enrichment of phosphate in xylem sap took place (Figure 4) and cannot be explained by fertilizer uptake because the fertilizer was not enriched in ¹⁸O (δ¹⁸O_Pi values = +19‰). Furthermore, δ¹⁸O_Pi values were ¹⁸O-enriched irrespective of treatment at the dates where the label was most intensely recovered in xylem sap at each site (BBR: 2 days, LUE: 14 days; Figure 4 and Supplementary Figure S5). At the dates where label recovery was not as pronounced (BBR: 14 days; LUE 2 days), P fertilization (P,NP) still significantly increased δ¹⁸O_Pi values in xylem sap which was not the case for the control and N treatment (Supplementary Figure S5). Because of our interpretation of the ¹⁸O-enrichment as a result of biological ¹⁸O incorporation into phosphate prior to P uptake (see previous chapter on site effects), we infer that (i) fertilizer-P was cycled biologically i.e., taken up and released by soil microorganisms, and/or (ii) P fertilization stimulated biological activity in soil (Mo et al., 2015) and thereby increased biological incorporation of ¹⁸O into phosphate prior to uptake by trees.