This study was conducted in Zhuji City, Zhejiang Province, China (120°14′E, 29°43′N). The 3,869 hectare research region features a subtropical monsoon environment with four distinct seasons, plenty of sunshine, and copious amounts of rainfall. With an average sunshine length of 1888 h per year and an average annual precipitation of 1,374 mm, the average annual temperature is roughly 16.3°C. Ultisol is the type of soil (Wang et al., 2022).

This area is famous for the biggest ancient T. grandis gathering place in China. All ancient T. grandis trees were independent. Because the T. grandis were planted in different years, the soil conditions and nutrient status will be different when planting (It is also difficult to keep the same). However, the soil management measures and intensity (turning the soil, weeding, and fertilizing) were basically the same after planting. Chemical fertilizer (N: P₂O₅: K₂O = 15%:15%:15%, ~30 kg per plant) was applied to the surface in March, and organic fertilizer (livestock manure, ~250 kg per plant) was applied to the surface with chemical fertilizer in September. The total amount of chemical fertilizer per unit area was 0.7 kg·m⁻², the total amount of organic fertilizer per unit area was 7.5 kg·m⁻², and the application depth was 30 cm (Yuan, 2020; Wang et al., 2022). Before the application of fertilizers, farmers were in the habit of planting green manure crops under the trees, but now nothing is planted under the trees in order to save labor.

In early September 2019, two independent sampling sites with basically the same site conditions and management measures were selected (Figure 1), namely, Zhaojiazhen (120° 30′ E, 29° 41′ N) and Dongbaihu (120° 29′ E, 29° 33′ N), which are both situated in the Torreya National Forest Park in Zhejiang, China (Wang et al., 2022). Zhaojia town is the main producing area of Torreya in Zhuji City (Meng et al., 2004). Dongbaihu is the main distribution area of Torreya in Zhuji City (Min et al., 2009).

According to the results of the local ancient tree census and the planting time of the new T. grandis forest provided by farmers, the sample trees with the same management measures were selected according to five tree age gradients (0–50, 50–100, 100–300, 300–500, and more than 500 year). The sample area of each age gradient depends on the number of T. grandis trees surveyed in corresponding age gradient. There are at least 13 and a maximum of 30 plants, with an average of about 19 plants per age gradient.

The growth cone method was used in determining the age of T. grandis. Tree cores were collected from the trunk or lateral branches of T. grandis, and the number of tree cores extracted was determined with a Leica LINTAMTM6 tree ring analyzer. Given that T. grandis is a grafting tree, the trunk and lateral branches were significantly eccentric. The eccentricity rate was calculated from the cross-section of T. grandis, and tree age was estimated according to the average tree ring width calibrated by eccentricity (Neri et al., 2005). Supplementary Table S1 provides detailed information about the trees and sampling sites.

An appropriate amount of branches (diameter 0.7 ~ 1 cm) and needles (at least 1,000 g fresh samples each) were collected from the east, south, west, north and upper, middle, and lower parts of the crown of each selected plant. The plant samples of each tree were mixed and placed in bags based on different organs. Plant roots (diameter less than 2 cm) were collected in the 0–60 cm deep soil profile in four directions. The leaves, twigs, and roots were dried in an oven at 65°C for 48 h. After being finely powdered to fit through a 40-mesh screen, the dried plant tissues were kept at a temperature of −80°C until the total carbon, nitrogen, and phosphorus contents were determined.

To gather soil from the 0–10, 10–20, 20–40, and 40–60 cm soil layers, 60 cm-deep soil profiles were dug in the east, south, west, and north directions, each one 1 m from the sample trees’ trunks (Yuan, 2020). The soil was then mixed and passed through a 2 mm standard sieve. A portion of a soil sample was utilized to measure the water content, and the remaining portion was divided into two parts. One part was kept in a refrigerator at 4°C for the determination of microbial biomass (C, N, and P), enzyme activity, nitrate N (NO₃⁻-N), ammonium N (NH₄⁺-N), and available P. The other part was air-dried, passed through a 20-mesh sieve for the measurement of pH and electrical conductivity and then passed through a 100-mesh sieve for the determination of soil organic C, total N, and total P content.

The potassium dichromate volumetric method and the outside heating method were used to determine the C concentration in each plant sample (Mebius, 1960). The Kjeldahl method was used to determine total N (Moraghan et al., 1983). Molybdenum antimony colorimetry was used to determine the total P (Yuen and Pollard, 1954).

Soil organic C content was determined in the same manner as that of a plant sample (Mebius, 1960), the total N content was evaluated using the Kjeldahl method (Moraghan et al., 1983), and the total P content was measured using molybdenum antimony colorimetry (Yuen and Pollard, 1954). According to Tel and Heseltine (2008), the concentrations of NO₃^--N and NH₄⁺-N were measured using an Auto Analyzer3 continuous flow analyzer, and the concentration of available P was estimated using the 0.5 mol·L⁻¹ sodium bicarbonate method (Olsen, 1954). A pH and conductivity meter was used in measuring the pH and conductivity of the soil and water (1,5, W/V) suspensions. Fresh soil was dried for 24 h at 105°C, and the weight moisture content of bulk soil was calculated.

The content of microbial biomass C and N was determined through chloroform fumigation-potassium sulfate extraction with a total organic carbon and total nitrogen analyzer (multi N/C 3100 TOC/TN), and the content of P was determined using the chloroform fumigation-potassium sulfate extraction-phosphoric acid determination plus correction method (Brookes et al., 1984, 1985; Vance et al., 1987). Chloroform fumigant is used to lyse the microbial cells in moist soil, followed by fumigant removal. Method for estimating soil microbial biomass by determining the total amount of organic matter that can be extracted from fresh soil microorganisms (Rainer et al., 2011).

The activities of dehydrogenase and urease were evaluated using triphenyltetrazolium chloride colorimetry and indophenol colorimetry (Małachowska-Jutsz and Matyja, 2019). P-nitrophenol-phosphate disodium colorimetry was used to evaluate the activity of the acid phosphatase (Dick, 2011). The 3,5-dinitro-salicylic acid colorimetry method was used to evaluate the activities of sucrase and cellulase (Frankeberger and Johanson, 1983; Deng and Tabatabai, 1994).

All data were analyzed using SPSS 26 (SPSS Inc., Chicago, United States), and normality and homogeneity variances were examined. Multiple comparisons and analysis of variance (ANOVA) were used in determining differences in C, N, and P content in T. grandis, microbes, and soils, and the least significant difference test was used for multiple comparisons. In all tests, statistical significance was accepted at p < 0.05. Pearson correlation analysis was used in measuring the correlation of C:N:P ecological stoichiometry among T. grandis leaves, twigs, roots, and microbes (0–20 cm). The relationships of C:N:P ecological stoichiometry in each component (leaves, twigs, roots, and microbes) with soil environmental factors were investigated through redundancy analysis (RDA), which was performed on Canoco 5.0. Soil and T. grandis leaves, twigs, roots, or soil and 0–10 and 10–20 cm soil microbial biomass at five tree age gradients were further examined using a linear regression model (ordinary least squares). All count data were transformed to their logarithmic base 10 (log) values.

The stoichiometric homeostasis of T. grandis leaves; twigs; roots; and C, N, P ecological stoichiometry or 0–10 and 10–20 cm soil microbial biomass C, N, P ecological stoichiometry at different tree ages were determined by Sterner and Elser (2002). The homeostasis model was proposed as follows:

where y represents the C, N, P content and stoichiometry of T. grandis leaves, twigs, roots or 0–10 and 10–20 cm soil microbial biomass; x represents soil C, N, and P content and stoichiometry; c is a constant, and H is the homeostatic regulation coefficient. The values of H and c were obtained through linear regression analysis. All datasets with significant regressions and 0 < 1/H < 1 were classified as 0 < 1/H < 0.25 “H-homeostatic,” 0.25 < 1/H < 0.5 “WH-weakly homeostatic,” 0.5 < 1/H < 0.75 “WP-weakly plastic,” or 1/H > 0.75 “P-plastic.” When p > 0.05, 1/H was set at zero, and the organism was regarded as “SH-strictly homeostatic.” Origin 2021 (Origin Lab Corp, Northampton, United States) was used to draw Figures 2–7. Data were shown as mean + SE. Table 1 provides definitions for each abbreviation.

At various age gradients (Figure 2), N and P concentrations had similar increasing tendencies in the leaves, twigs and roots, and N concentrations in the leaves of 0-50-year-old trees were significantly lower than other age gradients. The P concentration in the roots was significantly higher than those of the leaves and twigs at all age gradients. When the trees were more than 500 years old, the concentrations of N and P decreased in the leaves and twigs but remained at a high level in the roots.

The C:N ratios in the twigs were significantly higher than that in the leaves and roots at all age gradients. The C:P ratios in the roots were significantly lower than those in the leaves and twigs, except for the age gradient of 50–100 years. The N:P ratios were ranked as follows: leaves > roots > twigs, and N:P ratios in the leaves were significantly higher than that in the twigs, and higher than that in the roots when the forest age was over 50 years. The C:N ratios in the leaves, twigs, and roots all showed a declining tendency, and C:N ratios in the leaves of trees between 0 and 50 years old were significantly higher than those of other age gradients. The variation tendency of the C:P ratios was similar to the C:N ratios in the twigs and roots. The C:P ratios in the leaves increased and subsequently decreased with tree age, and the N:P ratios in the leaves and roots showed similar trends. A slight increase in N:P ratio was observed in the twigs with increasing tree age.

The C and N content in the four soil layers all exhibited similar increasing trends (Figure 3). At depths of 0–10 and 40–60 cm, the 50–100 year old T. grandis forests had significantly lower soil organic C (SOC) content than forests that were 300–500 and above 500 years old. The variations in soil N at different tree ages and soil depths were nearly identical to variations in soil C. At all four depths, the SOC and soil total N (STN) content increased with the age of T. grandis, but the soil total P (STP) content decreased (Figure 3). The SOC, STN, and STP content were significantly higher in the 0–10 cm depth than those in the 20–40 and 40–60 cm soil layers. In the 50-100-year-old forests, the soil had lower C:N, C:P, and N:P ratios than the other forests did. The C:N ratio did not drastically change with soil depth, whereas it did marginally rise with tree age. Soil P fluctuated with tree age but declined with soil depth, whereas C:P and N:P ratios generally increased, and the growing tendency is similar to that of C and N content.

At different age gradients (Figure 4), the N and P content of microbial biomass in the 0–10 and 10–20 cm soil layers showed similar trends. Microbial biomass N (MBN) content increased with tree age in the 0–10 cm soil layer but declined in the 10–20 cm soil layer. In the 10–20 cm soil depths, the microbial biomass P (MBP) content of the 0-50-year-old forests was significantly higher than the MBP content in the age gradients of 100–300, 300–500, and above 500 years old. Except the age gradient of 0–50 years, MBN of different age gradients in the 0–10 cm soil layer were significantly higher than that in the 10–20 cm soil layer. The MBC content in the 0–10 cm soil layers were generally higher than that in 10–20 cm soil layers at different age gradients, and the MBC contents in the 0–10 cm soil layers at the age gradient of 50–100 year were significantly higher than that at the age gradient of 100–300 year. In the 0–10 cm soil layers, MBC:N ratios showed a downward trend as tree age increased, and the MBC:N ratios at the age gradient of 0–50 year were significantly greater than that at the age gradient of 100–300, 300–500, and above 500 years. At the 10–20 cm soil depths, the MBC:P and MBN:P ratios followed the same change patterns as MBC:N.

The correlations of C:N:P ecological stoichiometry of T. grandis leaves, twigs, roots, and microbe in five age gradients are discussed in Figures 5–7. As T. grandis grew older, the correlation between the concentrations of C, N and P and their ratios in leaves indicated a declining trend, while that of twigs and roots showed little change. The correlations of soil microbe C, N and P contents and their ratios gradually got stronger as trees get older, and the correlations were highest at the age gradient of 300–500 year. There was little correlation between the C, N and P concentrations and their ratios of leaves, twigs and roots along the 0–50 year gradient. With the increase of tree age, the correlations among leaves, twigs and roots of T. grandis forests steadily increased, and the correlations reached the highest at 100–300 year gradient. Furthermore, the C, N and P concentrations and their ratios in different organs at 0–50 year gradient have a high correlation with that in soil microbe, while there had a significantly decreased at above 500 year gradient. Overall (0–900), there was a strong positive correlation existed between C, N, P, C:N, C:P and N:P of different organs, although the correlation of those indexes between microbe and T. grandis was weak.

The RDA results highlight how the influence of C:N:P stoichiometry of T. grandis is mediated by soil properties. For instance, soil C, N content and leaves C, N concentrations were negatively correlated, while soil C, N content and roots C, N concentrations were positively correlated. P concentrations in leaves, twigs and roots were positively correlated with soil P content (Figures 8A–C). In the soil, NO₃⁻ or NH₄⁺ contents were negatively correlated with the N concentrations and N:P ratios in the leaves (Figure 8A) but positively correlated with the N concentrations in the roots (Figure 8C). The first two axes can explain a total of 23.66, 21.65, and 23.58% variations of C, N, and P concentrations and its ratios in T. grandis leaves (F = 1.5, p ≤ 0.05, Table 2), twigs (F = 1.4, p > 0.05, Table 2) and roots (F = 1.6, p ≤ 0.05, Table 2) with 96 cases. The factors of NO₃⁻, EC, APA, NH₄⁺, N_s, WC, and pH were significant in Figure 8A, and the factors of URE, APA, pH, and DHA were significant in Figure 8C. No significant factors were found in Figure 8B (Table 3). The strength of five enzymatic activity were ranked as follows: APA > DHA > CEL > SUC > URE (Figures 8A), URE>APA > DHA > SUC > CEL (Figures 8C). A more detailed analysis by age is provided in Supplementary Tables S2–S4.

The ecological stoichiometry of microbes can be determined according to the RDA results (Figures 8D,E). For instance, in the 0–10 and 10–20 cm soil layers the soil N and NH₄⁺ content were positively correlated with microbial biomass N, while the soil enzyme activity URE and APA were positively correlated with microbial biomass C and N. The 0–10 and 10–20 cm soil layers’ microbial biomass benefited greatly with APA. Available P was positively correlated with microbial biomass P in 0–10 cm soil layer. A total of 31.10 and 29.03% variance of C, N, and P content and its ratios of soil microbial biomass in the 0–10 cm (F = 2.4, p ≤ 0.01, Table 2) and 10–20 cm (F = 2.2, p ≤ 0.01, Table 2) soil layers can be explained by the same environmental variables with 96 cases. The factors of pH, WC, APA, DHA, URE, N_s, EC, SUC, N:P_s, and CEL were significant, as shown in Figure 8D, and the factors of WC, APA, pH, EC, DHA, N_s, NO₃⁻, URE, NH₄⁺, and C_s were significant in Figure 8E (Table 3). The strength of five enzymatic activity were ranked as follows: APA > DHA > URE>SUC > CEL (Figures 8D), APA > SUC > DHA > URE>CEL (Figures 8E). A more detailed analysis is provided in Supplementary Tables S5, S6. Many similarities between Figures 8D,E were found, indicating that microbes in the 0–10 and 10–20 cm soil layers were closely related.

The homeostases of T. grandis leaf, twig, root, and soil microbe were analyzed based on the homeostasis model (Table 4). The results showed that C_l, N_l, and C:P_l displayed homeostasis or weakly homeostasis at low age gradients (0–50 and 50–100 years), N:P_l, N_r, P_r, N:P_m10, C_m20 and N_m20 displayed plasticity or weakly plasticity at low and high age gradients, and other nutrients and stoichiometry displayed strict homeostasis. The 1/H values of C content in the leaf and soil microbe in the 10–20 cm soil layer ranged from 0.022 to 0.089 and 0 to 0.574, respectively. The 1/H values of N content in the leaf, root, and soil microbe ranged from 0.070 to 0.289, 0.060 to 0.703, and 0.041 to 1.463, respectively. The 1/H values of P concentrations in the twig and root ranged from 0.066 to 0.490 and 0.132 to 0.981, respectively. The plasticity of P concentrations in the root was relatively high.

Plants of different ages have different nutrient utilization efficiency (Weih et al., 2018). In our study, N concentration in T. grandis was low during the early growth stages (Figure 2). With the increase of tree age, T. grandis rapidly grows, various physiological metabolic pathways became active and absorbs large amounts of N and P, and the concentrations of N and P in various organs increase (Yuan et al., 2019). As the tree reaches old age and the physiology deteriorates, organic matter accumulation decreases. Plants selectively absorb various nutrient elements in the soil according to their growth needs, and accumulate photosynthates due to enhanced metabolism, thereby promoting N and P dilution and decrease in N and P concentration in the leaves and twigs, while the concentration of N and P in roots remained relatively high because of soil fertilization (Zhang et al., 2019b). It is likely that nutrient concentration in T. grandis decreases at more than 500 year gradient because it acquires a adaptive mechanism for nutrient redistribution and storage during long-term survival competition in nature.

Studies have shown that the C:N ratios of leaves does not show significant change in the forest development stage, whereas twigs and roots have significant differences (Liang et al., 2018). In our study, little difference in N content was found in the roots, but significant variations in N content and C:N ratios in the leaves were found among different tree ages. The reason may be that trees disturbed by long time fertilization had sufficient nutrients in the soil, which destroyed the stability of N and P concentrations in the roots, so plant root absorption is not limited by nutrients, and the N content in the roots was significantly higher than that in the leaves and twigs.

Soil C content increases with tree age (Figure 3). Although subtropical regions generally have higher rates of organic matter decomposition and lower rates of organic matter accumulation (Vitousek et al., 2010), our study demonstrated that organic matter accumulated over a long time frame. The C:N ratio of soil organic matter increased slightly with tree age, indicating that the decomposition rate of soil organic matter slowed down in a long time range. During a 900-year timescale, little change in soil N contents in the 0–40 cm soil layers and P contents in the 0–10 cm soil layers was found. A possible explanation is that the soil N and P nutrients have been controlled by artificial fertilization in this area.

Soil permeability and porosity of plantations decreased dramatically with age and fertilizer use (Huang and Wen, 2007), and these decreases can be indirect reasons for the decline in plantation soil microbial biomass. In forests, N fertilization enhances microbial biomass immediately after application (Zhang and Zak, 1998; Hart and Stark, 2016), but microbial biomass declines in the long run (Wallenstein et al., 2006). These changes may be the reason that the N or P content in soil microbial biomass increased in the 0–10 cm soil layer and decreased in the deep soil layer (Figure 4). Soil microbial C:N ratio is higher than 10:1 (Figure 4), microbes cannot get enough nitrogen to form their bodies, thus affecting their reproduction rate. Figures 8C–E also showed that the strength of cellulase, a significant predictor of soil carbon cycling rate, was at its lowest. The accumulation of soil C in the T. grandis forest may be explained by the rate of C accumulation in litters and rhizosphere deposits exceeding the rate of C decomposition by soil microbes (Vesterdal et al., 2002).

Plants, soil, and soil microbes have complex interactions (Wardle et al., 2004). Plants provide C sources and energy for soil microbes in the form of root exudates, and soil microbes provide mineralized nutrients for plant communities (Liu and Wang, 2021; Shi et al., 2021; Ali et al., 2022). Soil C, N, and P content provide energy and nutrients for soil microbes (Elser and Hamilton, 2007; Gusewell and Gessner, 2009), and soil C:N:P ratios regulate microbial community composition to maintain stability between element absorption and release in plant-microbe-soil systems (Makino et al., 2003). Several studies have revealed strong correlations between C, N, and P in soils and plants (McGroddy et al., 2004; Yang and Luo, 2011; Liao et al., 2014). Our results showed the correlations among leaves, twigs and roots of T. grandis forests gradually increased with tree age, and the correlation was the highest at the 100–300 year gradient. Given that the land has been fertilized-managed and the soil nutrient supply is not deficient, we hypothesized that T. grandis over 100 years old has a high degree of N and P coupling in the leaves and twigs (Figures 5–7). Plant physiology usually assumes that N and P are active in young trees. The supply of aboveground plant nutrients are constrained by the availability of belowground nutrients (Townsend et al., 2007; Bui and Henderson, 2013; Wang and Zheng, 2020). For example, in tropical and subtropical forests where available P is scarce, the dynamics of plant C, N, and P are mainly influenced by the supply of soil P (Hedin, 2004; Morris and Blackwood, 2007; Chen et al., 2021). As a economic forest with long-term human activities, T. grandis may change the availability and utilization efficiency of N and P nutrient through physiological and biochemical mechanisms, such as resorption, to improve their adaptation.

In general, N and P concentration in the leaves have a strong coupling relationship (Reich et al., 2010). Photosynthesis, respiration rate, and productivity are positively correlated with N concentration in the leaves (Evans, 1989; Wright et al., 2004; Reich et al., 2009). The P concentration in the leaves regulates the relationship between plant physiological processes (especially growth rate) and leaf N-productivity (Reich et al., 2009; Wang and Zheng, 2020). In our study, the correlations between the concentrations of C, N and P in leaves and their ratios showed a general trend of decline with the increase of T. grandis forest age, while the changes of twigs and roots were not significant. These results are consistent with the growth-rate hypothesis, indicated that ancient T. grandis trees with long-term human interference have passed the life stage characterized by high growth rate and do not need large amounts of P transport from roots to leaves to support protein synthesis (Sardans et al., 2021).

The leaf N:P ratio is used to indicate the N limitation or P limitation in an ecosystem, that is, an N:P ratio < 14 indicates N limitation, and N:P ratio > 16 is the P limitation (Reich, 2005). In our study, the N:P ratio was always lower than 10, indicating that the plants had a limited N supply. This result is inconsistent with the fact that subtropical areas are usually limited by P (Chen et al., 2021), and also lower than the leaf N:P ratio (11.4) of seeded male T. grandis in this area (Yuan et al., 2019). Male T. grandis trees mainly provide pollen for female T. grandis trees, and they are fruitless and rarely managed by local farmers. The reason for this may be related to the long-term fertilization that has changed the inherent N and P cycles and processes of T. grandis forest ecosystem. Given that the fruits of T. grandis is primarily harvested by farmers who managed the ancient T. grandis trees, it is likely that the use of P fertilizer is more important than N fertilizer in order to promote higher yields, which can induce a shift from P to N limitation in plant production (Yuan and Chen, 2015). In our study, the P concentrations in the leaves and roots were positively correlated with the P content in the soil, also showing no P deficiency in the soil (Figures 8A–C).

Soil N deficiency showed in RDA is that N contents in soil microbes are more strongly correlated with its availability in soil than that of P contents (Figures 8D,E). This is because that in contrast to soil P, N is particularly low in soil and microbes may absorb N from soil when it is available. In Figures 8A,D,E, APA enzyme activity was much higher than URE, indicating that the decomposition, transformation and availability of organophosphorus in soil were much higher than the nitrogen supply capacity of soil. Soil extracellular enzymes are the main media of soil organic matter decomposition, and their activities are often related to the biochemical cycle of nutrients. Through the nutrient elements and energy material transfer established between extracellular enzymes, microbes complete material and energy cycles between soil and T. grandis during metabolic process (Carrasco et al., 2012; Ali et al., 2022).

Stoichiometric homeostasis reflects the physiological, biochemical, and growth adaptation of organisms to environmental changes (Hessen et al., 2004; Elser et al., 2010; Xing et al., 2016). Studies on the correlation and homeostasis between different plant organs, soil and microbial stoichiometry can facilitate study of plant nutrient allocation and utilization strategies (Bardgett and Wardle, 2010) and plant responses and adaptation mechanisms to the environment. In our research, ancient T. grandis has formed a stable homeostasis mechanism during its long-term growth (Table 4; Supplementary Table S7). Our results showed that roots and soil microorganisms are more sensitive to changes in soil environments than leaves and twigs. Studies on eucalyptus (Olsen and Bell, 1990), figs (Garrish et al., 2010), and lowland tropical forests in the Republic of Panama (Schreeg et al., 2014) have indicated that the N:P ratios of stems and roots are stronger indicators of soil nutrient availability than leaves. However, these studies were limited to seedlings, and the extent to which the stems, old leaves, and roots of mature trees can follow similar patterns remains to be determined. Our study showed that even 900-year-old mature trees also follow this hypothesis. There is evidence that metabolically active elements in tissues with higher metabolic activity (such as leaves) are less sensitive to environmental gradients than that with lower activity (Tang et al., 2018). Although microbial biomass comprises only a small proportion of soil organic matter, it is sensitive to environmental change and can thus be used as an indicator to some degree (Liu and Wang, 2021; Shi et al., 2021). The N:P ratio of soil microorganisms responds more strongly to nutrient availability than the N:P ratio of leaves, and is a better indicator of soil nutrient availability than that of leaves in our study.