This is the accepted version of the article: Wu, Huili; Xiang, Wenhua; Ouyang, Shuai; [et al.]. «Linkage between tree species richness and soil microbial diversity improves phosphorus bioavailabil- ity». Functional Ecology, Vol. 33, issue 8 (Aug. 2019), p. 1549-1560. DOI 10.1111/1365-2435.13355 This version is avaible at https://ddd.uab.cat/record/218195 under the terms of the license Linkage between tree species richness and soil microbial diversity improves phosphorus bioavailability Huili Wua,b, Wenhua Xianga,b,*, Shuai Ouyanga,b, David I. Forresterc, Bo Zhoua, Lingxiu Chen a,b, Tida Ged, Pifeng Leia,b, Liang Chena,b, Yelin Zenga, Josep Peñuelasf,g, Changhui Penge a Faculty of Life Science and Technology, Central South University of Forestry and Technology, Changsha, Hunan, 410004, China b Huitong National Station for Scientific Observation and Research of Chinese Fir Plantation Ecosystems in Hunan Province, Huitong, Hunan, 438107, China c Swiss Federal Institute of Forest, Snow and Landscape Research WSL, Zürcherstrasse 111, 8903 Birmensdorf, Switzerland d Institute of Subtropical Agriculture, Chinese Academy of Sciences, Changsha, Hunan, 410125, China e Institute of Environment Sciences, Department of Biological Sciences, University of Quebec at Montreal, Montreal, QC H3C 3P8, Canada f CSIC, Global Ecology Unit CREAF-CSIC-UAB, Bellaterra (Catalonia) E-08193, Spain; g CREAF, Cerdanyola del Vallès (Catalonia) E-08193, Spain *Corresponding author: Dr. Wenhua Xiang Faculty of Life Science and Technology, Central South University of Forestry and Technology, No. 498 Southern Shaoshan Road, Changsha 410004, Hunan, China. E-mail: xiangwh2005@163.com; Tel: +86-731-85623350; Fax: +86-731-85623350 To be submitted to “Nature Communication or Nature Plants” Type of article: Article 1 1 An increase in the availability of soil phosphorus (P) has recently been recognized as an 2 underling mechanism of the positive relationship between plant diversity and ecosystem 3 functioning. The effect of plant diversity on the bioavailable forms of P involved in 4 biologically mediated rhizospheric processes and how the link between plant and soil 5 microbial diversity facilitates soil P bioavailability, however, remain poorly understood. 6 We quantified four forms of soil bioavailable P in subtropical mature forests using a 7 novel biologically based approach and soil microbial diversity based on high-throughput 8 Illumina sequencing. Tree species richness was positively correlated with the four forms, 9 which was more pronounced in organic than mineral soil. A model of the link between 10 plants and soil microbes for each form indicated that soil bacterial and fungal diversities 11 played dominant roles in mediating the effects of tree species richness on the 12 bioavailability of soil P. The increasing biodiversity of trees and soil bacteria and fungi 13 could maintain the bioavailability of soil P in forest ecosystems and alleviate the 14 limitation of soil P. 15 16 Many studies have reported that plant biodiversity enhances ecosystem functions, 17 particularly above- and belowground biomass or productivity1, 2. Increases in biomass and 18 productivity (e.g. overyielding) in ecosystems with many species of plants can be attributed to 19 sampling (or selection) effects of the dominant species and to complementarity effects among 20 species3-5. The sampling effects are species-specific impacts on biomass due to the higher 21 probability of having highly productive species included and dominant in more highly diverse 22 ecosystems3, 4, 6. The complementarity effects refer to the various forms of niche partitioning 23 among species for acquiring resources in ways that are spatially or temporally complementary, 24 or plant-plant facilitation for increasing resource availability or other growing conditions, and 25 therefore increasing productivity3, 4, 6. Phosphorus (P) is an important nutrient for various 2 26 physiological processes and components7 (e.g. energy metabolism, signal transduction, energy 27 carriers, nucleic acids and membranes) needed for plant growth but is often deficient to meet 28 the demands of plants8, 9. An increase in soil P availability has therefore recently been 29 recognized as an underlying mechanism for the positive effects of plant diversity on ecosystem 30 biomass and productivity10. P, however, occurs in many inorganic and organic forms in the 31 soil, and the use of multiple forms of P by plants is complex and poorly understood11. 32 Understanding how plant diversity affects the availability of multiple forms of bioavailable P, 33 as opposed to single forms of available P or total P12, 13, may facilitate the development of 34 sustainable strategies to alleviate limitations of soil P. 35 Plants develop a range of mechanisms accompanied by microbial processes in response 36 to P deficiency to increase the mobility and bioavailability of soil P8, 11. Four potential 37 mechanisms can be generalized: (1) modification of root morphology and formation of 38 mycorrhizae14-16, (2) exudation of organic acids9, 17-19 , (3) exudation of enzymes (e.g. 39 phosphatase and phytase) 19-21 and (4) exudation of H+/OH−/HCO3−18, 22, 23 in the rhizosphere 40 by plant roots and soil microbes. The forms of bioavailable P involved in mechanisms 1 to 4 41 are defined as CaCl2-P, citric-P, enzyme-P and HCl-P, respectively9. 42 Increases in soil P bioavailability in ecosystems with diverse plant species are 43 hypothesized to involve plant-plant facilitation24, where P-mobilizing species improve P 44 nutrition for themselves and neighboring non-P-mobilizing species by secreting organic acids, 45 protons and enzymes into the rhizosphere to desorb and solubilize phosphates10, 12, 24 . 46 Facilitation has recently been identified in two-species intercropping ecosystems10, 24, 25 . 47 Forests are P self-nourishing ecosystems that depend on P retained in their own biomass and 48 supplied from litter decomposition26. The facilitation of soil P bioavailability, however, has 49 not yet been reported for forest ecosystems, which often consist of more than two plant species 50 or even dozens of species. 3 51 Soil microbes play important roles in returning nutrients to the soil by the decomposition 52 of litter (leaves and roots) and root exudations, which are key processes that bridge the link 53 between plant and soil P nutrition12, 16, 24, namely plant-microbe-soil interaction24 (Extended 54 Data Fig. 1). Diverse plant communities produce litter composed of more diverse traits of 55 leaves and roots (in amount and quality) and release more diverse root exudates27. The litter 56 and exudates can also influence soil organic carbon (SOC)24 and directly affect soil microbial 57 composition and activity12, 24, 28, 29 . Bioavailable soil P clearly has simultaneous multiple 58 forms9, and these forms can be mediated in natural ecosystems by the biodiversity of soil 59 microbes. For example, ectomycorrhizal (ECM) fungi are widely considered the main factor 60 for improving P uptake by plants24, 30, 31, and saprotrophic fungi are responsible for litter 61 decomposition and play a crucial role in the mobilization of organic P32. Bacteria can 62 solubilize mineral P or immobilize it in their biomass33. Plant and soil microbial communities 63 and their interactions can shape multiple forms of bioavailable P, but identifying and 64 quantifying their relative effects is difficult, perhaps because soil microbes obtain C 65 compounds from plants in exchange for mineral nutrients, including P30, 33. Plant-microbe-soil 66 interactions may thus be key mechanisms for understanding the biogeochemical processes 67 involved in P bioavailability in diverse plant ecosystems. 68 Bioavailable-P plant-plant facilitation and plant-microbe-soil interactions may strengthen 69 as forest stands develop34. We selected a total of 94 subplots (with areas of 10 × 10 m) along 70 diversity gradients from 1 to 12 tree species in three mature subtropical forests35 (Extended 71 Data Fig. 2) to quantify the four forms of soil bioavailable P (CaCl2-P, citric-P, enzyme-P and 72 HCl-P), tree species richness, soil bacterial and fungal diversity (Shannon index) and many of 73 the drivers hypothesized to be important for regulating their variation. We plotted bivariate 74 relationships to determine the influence of biodiversity on bioavailable P. We identified the 75 underlying mechanism of the effect of tree species richness on bioavailable P by formulating 4 76 a theoretical framework for the interconnections among all drivers and using structural 77 equation models (SEMs) to empirically evaluate the theoretical framework (Extended Data 78 Fig. 1). More details of the methodology are provided in the Methods section. 79 Tree species richness was positively associated with soil P bioavailability (Fig. 1), 80 consistent with other studies4, 13. Tree species richness may have been positively correlated 81 with bioavailable P because diverse tree species may produce more and diverse litter (leaves 82 and roots) to form SOC (Fig. 2), have various root morphological characteristics for secreting 83 more exudates (i.e. organic acids, phosphatases and H+/OH-/HCO3-) and increase tree growth 84 (i.e. basal area (BA), see Fig. 2), thereby increasing the requirements of the nutrients, including 85 P, that drive root exudation and intensify soil microbial activities. The positive effects of tree 86 species richness on bioavailable P were more pronounced in organic than mineral soil (Fig. 1), 87 reinforcing the premise that forests with many tree species generate diverse quantities and 88 qualities of litter24 and increase the density of fine roots distributed in the organic horizon, 89 which greatly increases P exudation. 90 The effects of tree species richness on bioavailable P varied with the form of bioavailable 91 P8 (Figs. 1 and 2). CaCl2-P is a labile P that is easily available to plants and is then depleted in 92 the rhizospheric soil14, 16. A CaCl2-P concentration gradient formed between the rhizosphere 93 and bulk soil, which could drive the mobilization of CaCl2-P from bulk soil to the rhizosphere. 94 Citric-P is an active form of inorganic P, adsorbing to clay particles and weakly binding to Ca, 95 Fe or Al precipitates, which can be easily released by organic acids9, 18, 19. Organic acids are 96 commonly secreted by living plants or dead roots, and their secretions are plant species- 97 specific. HCl-P is a recalcitrant inorganic P that can be solubilized by H+/OH-/HCO3- root 98 exudates. H+/OH-/HCO3- are secreted when roots take up ions in unbalanced proportions, 99 which is also plant species-specific8. More and diverse root morphological characteristics, 100 organic acids and H+/OH-/HCO3- may increase the bioavailability of CaCl2-P, citric-P and HCl- 5 101 P in diverse species communities (Fig. 2a, b and d). Enzyme-P, however, is an organic form of 102 P that will only be taken up by plants if mineralized by phosphatases9, 36 . Phosphatase 103 exudation by plants consumes energy and depends on the demand for P21, 37. If CaCl2-P, citric- 104 P and HCl-P increased by high diverse trees is sufficient for supporting P requirements of 105 plants, they contribute to reduce energy and substrate consumption36, and there is a weak 106 relationship between enzyme-P and tree species richness. 107 Our results indicated a strong, positive and linear correlation between the amount of 108 bioavailable P and bacterial and fungal diversity (Fig. 3), but the effect of microbial diversity 109 on bioavailable P differed among microbial taxa. The solubilization and immobilization of 110 inorganic P are the main mechanisms responsible for bacterial P bioavailability33. Bacterial 111 diversity also directly increased the amounts of the three forms of inorganic P (CaCl2-P, citric- 112 P and HCl-P; Fig. 2). The ability to solubilize inorganic P depends on the development of 113 extraradical mycelia by ECM fungi and the release of organic acids and H+/OH-/HCO3-30, 33. 114 Fungal diversity contributed more than bacterial diversity to the bioavailability of enzyme-P 115 (Fig. 3), suggesting that fungal communities had a dominant role in enzyme-P bioavailability 116 by the exudation of phosphatases. The effects of fungal diversity on citric-P and HCl-P were 117 similar to those of bacterial diversity and tree species richness, indicating that organic acids 118 and H+/OH-/HCO3- are commonly released by plants and microbes. A specific functional 119 group of ECM fungi has been documented as an important P predator and helped plants take 120 up P24, 30, 31. In addition, CaCl2-P is a readily absorbed and used form of inorganic P9, so the 121 lack of significant impacts of fungal diversity on CaCl2-P was not surprising, because highly 122 efficient CaCl2-P uptake by ECM fungi can offset the positive effects of other functional 123 groups of fungi. 124 Soil microbial diversity mediated the effects of tree species richness on soil bioavailable 125 P by three biological mechanisms (Figs. 2 and 4). Firstly, the roots of diverse tree species 6 126 release diverse exudates in rhizospheric soil as a “booster” for soil microbial activity and 127 diversity8. Our analysis found that tree species richness directly increased bacterial diversity. 128 Secondly, the plants in tree-rich communities have long fine roots, which provide more and 129 multiple hosts for soil microbes and thus multi-host-multi-microbe interactions38. Our results 130 indicated that tree species richness increased the length of fine roots and bacterial and fungal 131 diversity. Thirdly, tree species richness increased tree basal area (aboveground biomass) and 132 fine-root biomass, which would produce larger amounts and varieties of litter and thus more 133 SOC, which would then decrease bacterial and fungal diversity. The higher amounts of litter 134 produced by highly diverse species communities could affect resource availability or litter 135 leachates and alter microclimatic conditions, including soil-water content and temperature, 136 which might suppress the growth of some common microbial species or decrease their 137 competitive ability, thus lowering microbial diversity39. In contrast to diversity, microbial 138 activity and biomass could increase as the amounts of litter39 and SOC36 increased, which 139 could also increase mycorrhizal formation and exudation of organic acids, phosphatases and 140 H+/OH-/HCO3- to increase the amount of bioavailable P. 141 SOC had positive and direct effects on citric-P, enzyme-P and HCl-P24 (Fig. 2 and 142 Extended Data Figs. 3-6). Both biological and physical processes can account for this result. 143 Among the biological processes, communities with diverse tree species producing more SOC36 144 lead to higher microbial activity and thereby the production of more organic acids, 145 phosphatases and H+/OH-/HCO3-. The physical processes vary depending on the form of 146 bioavailable P. Citric-P and HCl-P bind weakly or create stable Fe and Al precipitates9 at 147 elevated concentrations of SOC in acidic forest soils, which can easily form soluble C 148 compounds-Fe(Al)-P complexes in which P is readily liberated40. The positive correlation 149 between SOC and enzyme-P may be due to the ability of SOC to adsorb phosphatases in an 150 active form41 and then maintain a high rate of enzyme-P mineralization. 7 151 The bivariate plots of tree species richness could only explain less than 7, 22, 7 and 12% 152 of the variation in CaCl2-P, citric-P, enzyme-P and HCl-P (Fig. 1), but the SEMs could explain 153 18, 41, 25 and 45% of the variation in CaCl2-P, citric-P, enzyme-P and HCl-P, respectively. 154 The SEM results indicated that the effects of tree species richness on bioavailable P were 155 mediated by other biotic and abiotic factors, such as soil microbes and SOC concentrations. 156 Not all of the variability of bioavailable P could be explained by the variables in these SEMs. 157 Other variables (e.g. soil pH; Extended Data Figs. 7 and 8) not included in these SEMs may 158 thus have also contributed to the effects of tree species richness on bioavailable P. 159 To the best of our knowledge, this study is the first to explore the mechanism of soil P 160 bioavailability in subtropical forests with diverse tree species by identifying the links between 161 trees, microbes and soil. Our findings have three important implications for understanding the 162 interactions between biodiversity and bioavailable P. Firstly, the increase in tree species 163 richness increased soil bioavailable P, including CaCl2-P, citric-P, enzyme-P and HCl-P, which 164 were more pronounced in organic than mineral soil. Secondly, soil bacterial and fungal 165 diversity can mediate the effects of tree species richness on bioavailable P. Tree species 166 richness can directly affect bacterial diversity and indirectly affect bacterial and fungal 167 diversity by increasing tree basal area and fine-root biomass and length, thereby affecting 168 bioavailable P. Thirdly, the SEMs indicated that SOC served as a link between tree species 169 richness and soil microbial diversity to affect bioavailable P, suggesting that soil abiotic factors 170 may be key drivers controlling the relationships between biodiversity and bioavailable P. More 171 observations and experiments that link plant and soil biodiversity to bioavailable P will 172 certainly be needed in the near future to evaluate and predict P bioavailability and mobilization 173 in forest ecosystems, because the loss of biodiversity is continuing and soil properties are 174 changing in forest ecosystems. 175 8 176 Online content Methods, additional Extended Data items and source data are available in the 177 online version; references unique to these sections appear only in the online version. 9 178 References 179 1. Grace, J. B. et al. Integrative modelling reveals mechanisms linking productivity and plant 180 species richness. Nature 529, 390-393 (2016). 181 2. Xiang, W. H. et al. Fine root interactions in subtropical mixed forests in china depend on 182 tree species composition. Plant Soil 395, 335-349 (2015). 183 3. Assaf, T. A., Beyschlag, W. & Isselstein, J. The relationship between plant diversity and 184 productivity in natural and managed grassland. Appl. Ecol. Env. Res. 9, 157-166 (2011). 185 4. Firn, J., Erskine, P. D. & Lamb, D. Woody species diversity influences productivity and 186 soil nutrient availability in tropical plantations. Oecologia 154, 521-33 (2007). 187 5. Tilman, D., Lehman, C. & Thomson, K. Plant diversity and ecosystem productivity: 188 theoretical considerations. P. Natl. Acad. Sci. USA 94, 1857-1861 (1997). 189 6. Cardinale, B. J. et al. Impacts of plant diversity on biomass production increase through 190 time because of species complementarity. P. Natl. Acad. Sci. USA 104, 18123-18128 191 (2007). 192 7. George, T. S., Hinsinger, P. & Turner, B. L. Phosphorus in soils and plants - facing 193 phosphorus scarcity. Plant Soil 401, 1-6 (2016). 194 8. Hinsinger, P. Bioavailability of soil inorganic P in the rhizosphere as affected by root- 195 induced chemical changes: a review. Plant Soil 237, 173-195 (2001). 196 9. Deluca, T. H. et al. A novel biologically-based approach to evaluating soil phosphorus 197 availability across complex landscapes. Soil Biol. Biochem. 88, 110-119 (2015) 198 10. Li, L., Tilman, D., Lambers, H. & Zhang, F. S. Plant diversity and overyielding: insights 199 from belowground facilitation of intercropping in agriculture. New Phytol. 203, 63-69 200 (2014). 201 11. Richardson, A. E., Barea, J. M., McNeill, A. M. & Pringet-Combaret, C. Acquisition of 202 phosphorus and nitrogen in the rhizosphere and plant growth promotion by 10 203 microorganisms. Plant Soil 339, 305-309 (2009). 204 12. Tang, X.Y. et al. Increase in microbial biomass and phosphorus availability in the 205 rhizosphere of intercropped cereal and legumes under field conditions. Soil Biol. Biochem. 206 75, 86-93 (2014). 207 13. Zeugin, F., Potvin, C., Jansa, J. & Schererlorenzen, M. Is tree diversity an important driver 208 for phosphorus and nitrogen acquisition of a young tropical plantation? For. Ecol. Manag. 209 260, 1424-1433 (2010). 210 14. Shimizu, A., Kato, K., Komatsu, A., Motomura, K. & Ikehashi, H. Genetic analysis of 211 root elongation induced by phosphorus deficiency in rice (oryza sativa l.): fine qtl 212 mapping and multivariate analysis of related traits. Theor. Appl. Genet. 117, 987-996 213 (2008). 214 15. Johnson, J. F., Vance, C. P. & Allan, D. L. Phosphorus deficiency in Lupinus albus: 215 altered lateral root development and enhanced expression of phosphoenolpyruvate 216 carboxylase. Plant Physiol. 112, 31-41 (1996). 217 16. Bolan, N. S. A critical review on the role of mycorrhizal fungi in the uptake of phosphorus 218 by plants. Plant Soil 134, 189-207 (1991). 219 17. Kpomblekou-A, K. & Tabatabai, M. A. Effect of low-molecular weight organic acids on 220 phosphorus release and phytoavailabilty of phosphorus in phosphate rocks added to soils. 221 Agr. Ecosyst. Environ. 100, 275-84 (2003). 222 18. Grinsted, M. J., Hedley, M. J., White, R. E. & Nye, P. H. Plant-induced changes in the 223 rhizosphere of rape (brassica napus var. emerald) seedlings: I. ph change and the increase 224 in p concentration in the soil solution. New Phytol. 91, 19-29 (1982). 225 19. Wei, L. L., Chen, C. R. & Xu, Z. H. Citric acid enhances the mobilization of organic 226 phosphorus in subtropical and tropical forest soils. Biol. Fert. Soils 46, 765-769 (2010). 227 20. Pant, H. K. and Warman, P. R. Enzymatic hydrolysis of soil organic phosphorus by 11 228 immobilized phosphatases. Biol. Fert. Soils 30, 306-311 (2000). 229 21. Fatemi, F. R., Fernandez, I. J., Simon, K. S. & Dail, D. B. Nitrogen and phosphorus 230 regulation of soil enzyme activities in acid forest soils. Soil Biol. Biochem. 98, 171-179 231 (2016). 232 22. Hedley, M. J., White, R. E. & Nye, P. H. Plant-induced changes in the rhizosphere of rape 233 (brassica napus var. emerald) seedlings. III. changes in l value, soil phosphate fractions 234 and phosphatase activity. New Phytol. 91, 45-56 (1982). 235 23. Devau, N., Cadre, E. L., Hinsinger, P., Jaillard, B. & Gérard, F. Soil pH controls the 236 environmental availability of phosphorus: experimental and mechanistic modelling 237 approaches. Appl. Geochem. 24, 2163-2174 (2009). 238 24. Faucon, M. P. et al. Advances and perspectives to improve the phosphorus availability in 239 cropping systems for agroecological phosphorus management. Adv. Agron. 134, 51-79 240 (2015). 241 25. Li, L. et al. Diversity enhances agricultural productivity via rhizosphere phosphorus 242 facilitation on phosphorus-deficient soils. P. Natl. Acad. Sci. USA 104, 11192-11196 243 (2007). 244 26. Osman, K. T. Forest soils-properties and management. Switzerland: Spring international 245 publisher (2013). 246 27. Smith, A. P., Marín-Spiotta, E. & Balser, T. Successional and seasonal variations in soil 247 and litter microbial community structure and function during tropical postagricultural 248 forest regeneration: a multiyear study. Global Change Biol. 21, 3532-3547 (2015). 249 28. Spohn, M. & Kuzyakov, Y. Phosphorus mineralization can be driven by microbial need 250 for carbon. Soil Biol. Biochem. 3, 69-75 (2011). 251 29. Garnier, E., Navas, M. & Grigulis, K. Plant functional diversity - organism traits, 252 community structure, and ecosystem properties. London: Oxford University Press (2016). 12 253 30. Battini F, Grønlund M, Agnolucci M, Giovannetti M & Jakobsen I. Facilitation of 254 phosphorus uptake in maize plants by mycorrhizosphere bacteria. Sci. Rep. 7, 4686 255 (2017). 256 31. Baldrian, P. Forest microbiome: diversity, complexity and dynamics. FEMS Microbiol. 257 Rev. 41, 109-121 (2017). 258 32. Koukol, O., Novak, F., Hrabal, R. & Vosatka, M. Saprotrophic fungi transform organic 259 phosphorus from spruce needle litter. Soil Biol. Biochem. 38, 3372-3379 (2006). 260 33. Lladó, S., López-Mondéjar, R. & Baldrian, P. Forest soil bacteria: diversity, involvement 261 in ecosystem processes, and response to global change. Microbiol. Mol. Biol. Rev. 81, 262 e00063-16 (2017). 263 34. Turner, B. L. Resource partitioning for soil phosphorus: a hypothesis. J. Ecol. 96, 698- 264 702 (2008). 265 35. Ouyang, S. et al. Significant effects of biodiversity on forest biomass during the 266 succession of subtropical forest in south china. For. Ecol. Manag. 372, 291-302 (2016). 267 36. Hacker, N. et al. Plant diversity shapes microbe-rhizosphere effects on P mobilisation 268 from organic matter in soil. Ecol. Lett. 18, 1356-1365 (2015). 269 37. Allison, S. D. and Vitousek, P. M. Responses of extracellular enzymes to simple and 270 complex nutrient inputs. Soil Biol. Biochem.37, 937-944 (2005). 271 38. Rottstock, T., Joshi, J., Kummer, V. & Fischer, M. Higher plant diversity promotes higher 272 diversity of fungal pathogens, while it decreases pathogen infection per plant. Ecology 95, 273 1907-1917 (2014). 274 39. Sayer, E. J. Using experimental manipulation to assess the roles of leaf litter in the 275 functioning of forest ecosystems. Biol. Rev. 81, 1-31 (2006). 276 40. Porcal, P., Frejlachová, K., Kopáček, J., Nedoma, J. and Šavrdová, T. Photochemical 277 cleaving of allochthonous organic-metal complexes contributes to phosphorus 13 278 immobilization in surface waters. Chemosphere 167, 374-381 (2017). 279 41. Nohrstedt, H. Ö. Biological Activity in Soil from Forest Stands in Central Sweden, as 280 Related to Site Properties. Microb. Ecol. 11, 259-266 (1985). 281 14 282 283 Extended Data are available in the online version. 284 285 Acknowledgements This work was supported by the National Natural Science Foundation of 286 China (31570447 and 31170426) and the Huitong Forest Ecological Station funded by State 287 Forestry Administration of China. JP was funded by the European Research Council Synergy 288 grant SyG-2013-610028 IMBALANCE-P. We thank Prof. Xinhua He from Southwestern 289 University for his valuable comments on our manuscript. Thanks also go to the laboratory staffs 290 of the Institute of Subtropical Agriculture, Chinese Academy of Sciences, Changsha, Hunan 291 Province for their experimental support. 292 293 Author contributions Concept and study design: WX; data collection and analysis: HW, SO, 294 TG, BZ, LC, LP, SZ, LC and YZ; manuscript writing: HW, WX, DIF, JP and CP. 295 296 Author information The authors declare no competing financial interests. Correspondence 297 and requests for materials should be addressed to W.H.X. (xiangwh2005@163.com). 15 298 METHODS 299 Site description. This study was carried out in the Dashanchong Forest Park (28°23ʹ58ʺ- 300 28°24ʹ58ʺN, 113°17ʹ46ʺ-113°19ʹ08ʺE) in Changsha County, Hunan Province, China. The 301 altitude ranges from 55 to 17 m a.s.l. The park has a mean annual precipitation of 1416 mm 302 and a mean annual temperature of 17.3 °C. The soil is a well-drained red clayey loam classified 303 as an Alliti-Udic Ferrosol. Details of the site are provided by Jiang et al.42 and Zhu et al.43. 304 No activities of human disturbance, such as firewood collection, have been allowed in the 305 park since the late 1950s. Secondary forests have developed after decades of forest protection, 306 dominated by Pinus massoniana, Choerospondias axillaris, Cyclobalanopsis glauca, 307 Lithocarpus glaber and Loropetalum chinense. A 1-ha permanent plot was established in 2013 308 for each of three secondary forests: P. massoniana ±L. glaber coniferous and evergreen 309 broadleaved mixed forest (PLF), C. axillaris deciduous broadleaved forest (CAF) and L. glaber 310 – C. glauca evergreen broadleaved forest (LGF) at early, middle and late successional stages. 311 Forest ecosystems are highly complex, with many microsites varying in environmental 312 factors34, 35. We established a network of forest plots along gradients of tree species richness 313 within the forests to account for environmental factors34, 44, 45. Each plot was subdivided into a 314 grid of 100 subplots of 10 × 10 m. The locations of trees were mapped within each subplot, 315 and the species, diameter at breast height (DBH) and height (H) of all trees were recorded. A 316 similar experimental design was used to examine the effects of plant functional diversity on 317 forest ecosystem function46. Detailed information of stand characteristics is available in 318 Ouyang et al.35 and Zhu et al.43. 319 320 Sample collection. We selected 31 subplots based on their tree species richness along a 321 diversity gradient from 2 to 9 species in PLF, 31 subplots along a diversity gradient from 1 to 322 12 species in CAF and 32 subplots along a diversity gradient from 1 to 11 species in LGF (Fig. 16 323 S2), for a total of 94 subplots containing 40 species (Extended Data Table 1). We avoided 324 adjacent subplots as much as possible to eliminate edge effects but used a five-point mixed 325 sampling method to eliminate edge effects when not possible. The five sampling points 326 included the center of the subplot and four points equidistant from the center toward the corners 327 of the plots (Extended Data Fig. 9). Samples of organic soil were collected within areas 50 × 328 50 cm at each point after the litter was removed. Samples of mineral soil were then collected 329 from the 0-10 cm soil layer. All mixed soil samples were sieved to pass through a 2-mm mesh 330 and divided into three subsamples. One subsample was air-dried for the determination of soil 331 organic-carbon (SOC) concentration, soil available-P concentration and soil pH; one 332 subsample was stored at 4 °C for measuring the amount of bioavailable P and one subsample 333 was stored at -80 °C for measuring microbial diversity. Fine roots (<2 mm in diameter) were 334 collected from the 0-10 cm soil layer at the five points in each subplot using an auger and were 335 transported to the laboratory for further analysis. 336 337 Chemical analysis. Four fractions of bioavailable P (CaCl2-P, citric-P, enzyme-P and HCl-P) 338 were measured using the extraction method reported by Deluca et al.9. Each P fraction was 339 measured in parallel by shaking 0.5 g of fresh soil with each extract (10 ml) in separate 15-ml 340 centrifuge tubes for 3 h on a reciprocal shaker at 180 rpm. The extracts were then centrifuged 341 (4000 g, 25 °C, 30 min) to obtain supernatants containing the four forms of bioavailable P. 342 CaCl2-P was assessed using a 10 mM CaCl2 solution, citric-P was assessed using a 10 mM 343 citric acid solution, enzyme-P was assessed using a final concentration of 0.02 enzyme units 344 ml-1 solution mixed with phosphatase and phytase and HCl-P was assessed using a 1 M HCl 345 solution. Citric-P extracts were diluted 10-fold, and HCl-P extracts were diluted 20-fold. The 346 CaCl2-P and enzyme-P extracts were not diluted. All extracts were analyzed colorimetrically 347 (630 nm) by the malachite-green method58 using a multiscan spectrum (Tecan Infinite® 200 17 348 Pro). 349 Soil pH was measured at a soil:water (deionized) ratio of 1:2.5 using an FE20 pH meter 350 (Mettler Toledo, Shanghai, China). Air-dried soil was ground and sieved through a 0.25-mm 351 mesh. The SOC concentration was measured using K2Cr2O7-H2SO4 oxidation. Soil available P 352 concentrations were determined using 0.05 mol L-1 HCl and 0.025 mol L-1 (1/2 H2SO4)59. Soil 353 properties are presented in Extended Data Table 2. Fine roots were separated as described by 354 Liu et al.47, and their biomass and length were then quantified. 355 356 Assessment of microbial diversity. DNA was extracted from 0.5 g fresh weight of thawed soil 357 samples using the E.Z.N.A.® soil DNA Isolation Kit (Omega Bio-tek, Norcross, USA) 358 following the manufacturer’s protocol. The diversity of the soil microbial communities was 359 analyzed by DNA sequencing using the Illumina MiSeq platform. Bacterial 16S rDNA genes 360 were amplified using the primer pair 515F (5′-GTGCCAGCMGCCGCGG-3′) and 907R (5′- 361 CCGTCAATTCMTTTRAGTTT-3′)48. Fungal ITS genes were amplified using the primer pair 362 ITS1F (5′-CTTGGTCATTTAGAGGAAGTAA-3′) and ITS2 (2043R) (5′- 363 GCTGCGTTCTTCATCGATGC-3′)49. Raw fastq files were demultiplexed and then quality- 364 filtered using QIIME (version 1.17) with the following criteria. (i) Reads of 300 bp were 365 truncated at sites receiving an average quality score <20 over a 50-bp sliding window, 366 discarding the truncated reads that were <50 bp. (ii) Exact barcode matching, two mismatched 367 primer nucleotides and reads containing ambiguous characters were removed. (iii) Only 368 sequences that overlapped by >10 bp were assembled based on their overlap sequence. Reads 369 that could not be assembled were discarded. Operational taxonomic units (OTUs) were 370 clustered with a cutoff of 97% similarity using UPARSE (version 7.1 371 http://drive5.com/uparse/), and chimeric sequences were identified and removed using 372 UCHIME. The taxonomy of each 16S rRNA gene sequence was analyzed by RDP Classifier 18 373 (http://rdp.cme.msu.edu/) against the SILVA (SSU115) 16S rRNA gene database using a 374 confidence threshold of 70%50. The fungal ITS OTUs were assigned to taxa using the BLAST 375 interface against the UNITE database V6.9.7. (E<10-5)51. The Shannon diversity index, 376 calculated for these rarefied OTU taxonomies using QIIME (version 1.17), was selected for 377 this study because it provides a robust and informative estimate of taxonomic diversity for soil 378 bacterial and fungal communities52. 379 380 Statistical analysis. We first determined the relationships between the four forms of 381 bioavailable P and soil available P using Pearson correlations (Extended Data Table 3). We next 382 assessed the relationships between biodiversity, bioavailable P, tree basal area, fine-root 383 biomass, fine-root length and SOC (Extended Data Figs. 3-6 and Extended Data Table 4) using 384 linear regressions. We then identified the effects of tree species richness, tree basal area, fine- 385 root biomass and length, soil bacterial and fungal diversity and SOC on bioavailable P; 386 individual variables were subjected to multiple regression model selection based on the 387 corrected Akaike information criterion (AIC) (Extended Data Table 5). 388 Structural equation models (SEMs) were used to analyze the direct and indirect 389 relationships between the four forms of bioavailable P and tree species richness, tree basal area, 390 fine-root biomass and length, soil bacterial and fungal diversity and SOC. The first step in an 391 SEM requires establishing an a priori model based on known effects and the relationships 392 among the driving variables (Extended Data Fig. 1 and Extended Data Table 5). In our model, 393 we only considered the bottom-up effect of tree species richness on soil bioavailable P using 394 tree basal area, fine-root biomass and length, soil bacterial and fungal diversity and SOC. Data 395 manipulation was required before modeling. The distributions of endogenous variables were 396 estimated, and their normality was tested. Tree basal area, fine-root biomass and length, citric- 397 P, HCl-P and SOC were log-transformed to satisfy the requirement of normality. The R 19 398 software platform53 and the lavaan54 and lavaan.survey55 packages were used to analyze our 399 SEMs. Each pathway in the final model was evaluated for significant contributions to the model. 400 Indices of model fit were the χ2-test (a lower χ2 indicates a better model), P (traditionally > 401 0.05), the root mean square error (RMSE) of approximation (RMSEA; the model has a good 402 fit when RMSEA <0.05) and the 90% confidence intervals (CI90). Details of the SEMs are 403 shown in the Extended Data Notes. 404 405 42. Jiang, F. et al. Spatial variations in soil organic carbon, nitrogen and phosphorus 406 concentrations related to stand characteristics in subtropical areas. Plant Soil 413, 289- 407 302 (2016). 408 43. Zhu, W. J. et al. Spatial and seasonal variations of leaf area index (LAI) in subtropical 409 secondary forests related to floristic composition and stand characters. Biogeosciences 13, 410 3819-3831 (2016). 411 44. Baeten, L. et al. A novel comparative research platform designed to determine the 412 functional significance of tree species diversity in european forests. Perspect. Plant Ecol. 413 Evol. Syst. 15, 281-291 (2013). 414 45. Bruelheide, H. et al. Designing forest biodiversity experiments: general considerations 415 illustrated by a new large experiment in subtropical china. Methods Ecol. Evol. 5, 74-89 416 (2014). 417 46. Chiang, J. M. et al. Functional composition drives ecosystem function through multiple 418 mechanisms in a broadleaved subtropical forest. Oecologia 182, 829-240 (2016). 419 47. Liu C. et al. Standing fine root mass and production in four Chinese subtropical forests 420 along a succession and species diversity gradient. Plant Soil 376, 445-459 (2014). 421 48. Xiong, J. B. et al. Geographic distance and ph drive bacterial distribution in alkaline lake 422 sediments across Tibetan Plateau. Environ. Microbiol. 14, 2457-2466 (2012). 20 423 49. Bokulich, N. A. & Mills, D. A. Improved selection of internal transcribed spacer-specific 424 primers enables quantitative, ultra-high-throughput profiling of fungal communities. 425 Appl. Environ. Microb. 79, 2519-2526 (2013). 426 50. Amato, K. R. et al. Habitat degradation impacts black howler monkey (Alouatta pigra) 427 gastrointestinal microbiomes. ISME J. l7, 1344-1353 (2013). 428 51. McDonald, D. et al. An improved Greengenes taxonomy with explicit ranks for ecological 429 and evolutionary analyses of bacteria and archaea. ISME J. 6, 610-618 (2012). 430 52. Delgado-Baquerizo, M. et al. microbial diversity drives multifunctionality in terrestrial 431 ecosystems. Nat. Commun. 7, 10541 (2015). 432 53. R Core Team. R: A language and environment for statistical computing. R Foundation for 433 Statistical Computing, Vienna, Austria. URL https://www.R-project.org/ (2016). 434 54. Rosseel, Y., Oberski, D., Byrnes, J., Vanbrabant, L. & Savalei, V. lavaan: latent variable 435 analysis. R Package Version 05-13 (2013). 436 55. Oberski, D., Grün, B., Pebesma, E. & Zeileis, A. Lavaan.survey: an r package for complex 437 survey analysis of structural equation models. J. Stat. Softw. 57, 1-27 (2014). 438 21 439 Figure legends 440 Figure 1 | The correlations of tree species richness with CaCl2-P (a), citric-P (b), enzyme- 441 P (c) and HCl-P (d). The red and blue fitted lines are from linear regression (n=94). Only 442 significant fitted lines are shown on the graphs. Significance indicated by asterisks: * P < 0.05, 443 ** P < 0.01, *** P < 0.001. 444 445 Figure 2 | Structural equation models of tree species richness, tree basal area (tree BA), 446 fine root length, fine root biomass, soil organic carbon (SOC), bacterial diversity and 447 fungal diversity on soil CaCl2-P (a), citric-P (b), enzyme-P (c) and HCl-P (d) in organic 448 soil (n=94). The fit indices of the four models were the same; χ22=1.112, P=0.573; 449 RMSEA=0.000, CI90 (0.000; 0.172). Numbers in the endogenous variable indicate the 450 explained variance (R2). Numbers next to the arrows indicate standardized path coefficients. 451 Arrow width is proportional to the strength of path coefficients. Significance indicated by 452 asterisks: * P < 0.05, ** P < 0.01, *** P < 0.001. 453 454 Figure 3 | The correlations of soil bacterial diversity and fungal diversity with CaCl2-P 455 (a, b), citric-P (c, d), enzyme-P (e, f) and HCl-P (g, h). The red and blue fitted lines are from 456 linear regression (n=94). Only significant fitted lines are shown on the graphs. Significance 457 indicated by asterisks: * P < 0.05, ** P < 0.01, *** P < 0.001. 458 459 Figure 4 | The correlations of tree species richness with soil bacterial diversity (a) and 460 fungal diversity (b). The red and blue fitted lines are from linear regression (n=94). Only 461 significant fitted lines are shown on the graphs. Significance indicated by asterisks: * P < 0.05, 462 ** P < 0.01, *** P < 0.001. 22 463 464 Figure 1 | Correlations of tree species richness with CaCl2-P (a), citric-P (b), enzyme-P (c) 465 and HCl-P (d). The red and blue lines are the fitted regression lines (n=94). Only significant 466 fitted lines are shown. * P < 0.05, ** P < 0.01, *** P < 0.001. 23 467 468 Figure 2 | Structural equation models of the effects of tree species richness, tree basal 469 area (Tree BA), fine-root length, fine-root biomass, soil organic carbon (SOC), bacterial 470 diversity and fungal diversity on soil CaCl2-P (a), citric-P (b), enzyme-P (c) and HCl-P (d) 471 in organic soil (n=94). The fit indices of the four models were the same; χ22=1.112, P=0.573; 472 RMSEA=0.000, CI90 (0.000; 0.172). The numbers for the endogenous variables indicate the 473 explained variance (R2). The numbers on the arrows indicate standardized path coefficients. 474 Arrow width is proportional to the strength of the path coefficients. * P < 0.05, ** P < 0.01, 475 *** P < 0.001. 24 476 477 Figure 3 | Correlations of soil bacterial diversity and fungal diversity with CaCl2-P (a, b), 478 citric-P (c, d), enzyme-P (e, f) and HCl-P (g, h). The red and blue lines are the fitted 479 regression lines (n=94). Only significant fitted lines are shown. * P < 0.05, ** P < 0.01, *** 480 P < 0.001. 25 481 482 483 Figure 4 | Correlations of tree species richness with soil bacterial diversity (a) and fungal 484 diversity (b). The red and blue lines are the fitted regression lines (n=94). Only significant 485 fitted lines are shown. * P < 0.05, ** P < 0.01, *** P < 0.001. 26