Variation in Morphological and Chemical Traits of Mediterranean Tree Roots: Linkage with Leaf Traits and Soil Conditions

1  Variation in morphological and chemical traits of Mediterranean tree 1 roots: linkage with leaf traits and soil conditions 2 3 Teodoro Marañón1*, Carmen M. Navarro-Fernández1, Marta Gil-Martínez1, María T. 4 Domínguez2, Paula Madejón1 and Rafael Villar3 5 6 1 IRNAS, CSIC, Avenida Reina Mercedes 10, 41012 Seville, Spain 7 2 Departamento de Mineralogía, Cristalografía y Química Agrícola, Universidad de Sevilla, 8 Prof. García González s/n, 41012, Seville, Spain 9 3 Área de Ecología, Facultad de Ciencias, Universidad de Córdoba, Campus de Rabanales, 10 14071 Córdoba, Spain 11 12 * Author for correspondence ([email protected]) 13 14 15 Author contribution statement 16 TM conceived the study, TM, CMNF and MTD conducted fieldwork, CMNF and PM 17 measured morphological and chemical traits, TM, CMNF and MGM analysed the data, TM 18 wrote the first draft, TM, CMNF, MGM, MTD, PM and RV participated in the interpretation, 19 discussion and preparation of the final version. 20 21 Postprint of: Plant and Soil (2020) https://doi.org/10.1007/s11104-020-04485-5 2  Abstract 22 Aims: Root functions are multiple and essential for the growth and survival of terrestrial 23 plants. The aim of this work was to analyse the main trends in the variation of root traits, their 24 coordination with leaf traits and their relationships with soil conditions. 25 Methods: We measured the variation of 27 fine root traits (five morphological, 20 chemical 26 and two isotopic signatures) in trees of seven species of a mixed plantation in a metal-27 contaminated and remediated site of Southern Spain. 28 Results: We found evidences supporting the existence of a root economics spectrum (RES). 29 However, other dimensions were identified as being independent of the main RES: mainly the 30 variation in the carbon concentration, the accumulation of trace elements associated with 31 tolerance of metal-rich soils, and the fractionation of δ15N as a time-integrated trait of 32 mycorrhizal-mediated nutrition. In general, roots and leaves were functionally coordinated, 33 although most of the trace elements showed strong root-leaf discordance. The soil conditions 34 interacted with the fine root traits in feedback processes. The ability of tree roots to 35 accumulate trace elements and to reduce their translocation to leaves is a desirable trait for the 36 phytoremediation of metal-contaminated soils. 37 Conclusions: Roots are multifunctional. Understanding the variations in the root traits of trees 38 will help us to predict both the responses of forests to global changes, including soil 39 contamination, and the provision of soil-based ecosystem services. 40 41 Keywords 42 Root traits, Root economics spectrum, Root-leaf coordination, Trace-element contamination, 43 Root multifunctionality 44 45 Introduction 46 3  Advances in trait-based plant ecology are focused on the analysis of functional traits 47 across individuals and species, to predict emergent properties of communities and ecosystems 48 (Garnier et al. 2016; Laliberté 2017). There are evolutionary and biophysical constraints 49 limiting the existing spectrum of plant traits (Reich 2014). In a global view of the functional 50 diversity of vascular plants on Earth (analysis of key traits for more than 46,000 plant 51 species), most of the trait variation was concentrated in a two-dimensional spectrum of plant 52 form and function: plant size and leaf economics spectrum (LES) (Díaz et al. 2016). The LES 53 reflects the trade-off between resource acquisition and resource conservation. At one end of 54 the spectrum, there are species with high photosynthetic and respiration rates, high nitrogen 55 (N) and phosphorus (P) concentrations, low leaf mass per area (LMA) and low leaf longevity. 56 At the other end of the spectrum, there are species with the opposite traits (Wright et al. 57 2004). 58 Although roots are essential organs for terrestrial plants, no root traits were considered 59 in the global study of Díaz et al. (2016). Root functions are multiple and essential for plant 60 growth and survival: they include nutrient and water acquisition, anchorage, resource storage 61 and support of symbiotic soil microbes. At the ecosystem level, they contribute to soil 62 structure and to the carbon and nutrient cycles (Erktan et al. 2018). Therefore, an 63 understanding of how root traits vary is fundamental to the comprehension of plant functional 64 ecology. 65 Some studies support the existence of a “root economics spectrum” (RES), analogous 66 to the LES, with a trade-off between resource acquisition and conservation (Reich 2014; 67 Roumet et al. 2016; de la Riva et al. 2018a). Thus, plants growing in favourable environments 68 would develop lighter roots with a lower dry matter content and higher specific root length to 69 maximize resource acquisition. By contrast, plants growing in adverse or limiting 70 environments would exhibit a resource conservation strategy, developing denser roots with a 71 4  higher dry matter content and lower specific root length. However, the RES hypothesis was 72 challenged by Weemstra et al. (2016) who argued that root traits are constrained not only by 73 resource uptake, but also by other drivers (like soil texture and chemistry), and that the RES 74 hypothesis does not incorporate soil heterogeneity and mycorrhizal symbiosis. Moreover, 75 several studies (Kramer-Walter et al. 2016; Kong et al. 2019) have found that high SRL roots 76 can be constructed with any density, indicating exceptions in the RES. 77 Resource acquisition is coupled and linked among plant organs. Thus, fast acquisition 78 and processing of water and nutrients by roots would require fast acquisition and processing 79 of carbon (C) by leaves (Reich 2014). A strong coordination between root morphology and 80 aboveground traits was found for a set of 80 woody species (de la Riva et al. 2018a). Root 81 traits that achieve nutrient conservation favour tissue longevity and slower growth rates, and 82 in consequence diminish nutrient requirements and amortization of the construction costs 83 (Poorter and Villar 1997; Villar et al. 2006; de la Riva et al. 2016b, 2018a). However, other 84 studies did not find correlations between leaf and root traits, suggesting that trade-offs in 85 different organs operate independently and that the leaf-root coordination may depend on 86 specific limiting factors in each habitat (Tjoelker et al. 2005; Kembel and Cahill 2011; 87 Fortunel et al. 2012). 88 Contrasting leaf habits in trees - that is, evergreen versus deciduous - are usually 89 associated with different functional traits. For example, deciduous species are characterised 90 by acquisitive traits such as lower LMA, higher rates of photosynthesis and respiration, and 91 higher nutrient concentrations, in comparison to evergreen species, which tend to exhibit 92 more-conservative traits (Wright et al. 2004; Villar et al. 2006; de la Riva et al. 2018b). 93 However, few studies have investigated the differences in root traits between evergreen and 94 deciduous trees; for example, Martinez et al. (2002) did not find differences in root C or N 95 concentrations between deciduous and evergreen species of Quercus. 96 5  Most tree roots are intimately associated with mycorrhizal fungi in a symbiosis that is 97 crucial for nutrient acquisition and tolerance of diverse stresses (drought, heavy metals or 98 pathogens), while the fungus obtains carbon compounds from the plant (Smith and Read 99 2008). In fact, that combination of root and fungus (mycorrhiza) can be considered as the 100 functional absorptive trait, in which fungal tissues may represent up to 54% of the “root” N 101 concentration (Ouimette et al. 2013). The degree of root colonisation by mycorrhizal fungi 102 has proved to be a useful plant trait to understand ecosystem processes (Soudzilovskaia et al. 103 2015; Navarro-Fernández et al. 2016; Laliberté 2017). Fungal traits, like the type of hyphal 104 exploration, add more complexity to the soil-fungus-plant relationships and resource 105 acquisition strategies (Chagnon et al. 2013; Gil-Martínez et al. 2018; López-García et al. 106 2018). 107 Besides unravelling the RES, another research challenge is to understand how 108 different drivers of global change impact a suite of root traits, and to predict their cascading 109 effects on soil-based ecosystem processes (Bardgett et al. 2014). Root traits are plastic and 110 respond to physical soil limitations, the heterogeneous distribution of soil water and nutrients 111 and biotic interactions (Bardgett et al. 2014). For example, soil compaction limits the 112 formation and penetration of thin roots, inducing a lower specific root length (SRL) in tree 113 seedlings (Alameda and Villar 2012). Under dry soil conditions, plants tend to develop 114 thinner roots, with greater SRL and increased root hair density, to improve water acquisition 115 (Comas et al. 2013; Olmo et al. 2014). Nitrogen deposition decreases fine root biomass, C:N 116 ratio and fungal colonization, while increasing root respiration (Li et al. 2015). 117 As a global change driver, soil pollution may also promote the adjustment of root 118 traits in plants. For instance, a high concentration of trace elements in soil often reduces root 119 elongation and alters root architecture (Kahle 1993). Although, at a global scale, soil pollution 120 is one of the main threats to soils and the ecosystems services provided by them (Rodríguez-121 6  Eugenio et al. 2018), its effects on plant functional traits have not been fully addressed. 122 Among the different soil pollutants, heavy metals are relevant stressors, altering the plant-soil 123 interactions (Krumins et al. 2015). 124 In this study we analysed the variation in morphological and chemical root traits in 125 seven Mediterranean tree species, and explored the linkages with aboveground traits and soil 126 conditions in a heterogeneously-polluted environment (Guadiamar Green Corridor, in SW 127 Spain). This area is a large-scale example of the phytoremediation of land contaminated by a 128 mine-spill, with high concentrations of metals. A mixed plantation of native trees and shrubs 129 was set up after cleaning and remediating the soil (Domínguez et al. 2008; Madejón et al. 130 2018a, b). This large-scale experiment is an opportunity to explore how the soil conditions (in 131 this case, the concentration of heavy metals) affect the root traits of different tree species 132 coexisting in a similar environment. We addressed the following hypotheses: 133 1) Roots of different tree species differ in their functional traits in accordance with the 134 root economics spectrum (RES). However, there are other root dimensions 135 (independent of RES) that reflect root multifunctionality (Weemstra et al. 2016). 136 2) Root and leaf traits are coordinated in accordance with the plant economics spectrum. 137 Fast plant growth depends on the coordination of roots and leaves, with roots ensuring 138 a water and nutrient supply sufficient to maintain acquisitive leaves with high 139 photosynthetic rates and high evaporative demand (Reich 2014). 140 3) Soil conditions and metal contamination affect root traits. Root traits are plastic and 141 respond to physical soil limitations, the heterogeneous distribution of soil water and 142 nutrients, biotic interactions (Bardgett et al. 2014) and soil pollution (Kahle 1993). 143 144 Material and methods 145 Study area 146 7  The study area is the Guadiamar Green Corridor (Seville, Spain). The climate is 147 Mediterranean with mild, rainy winters and hot, dry summers. The average annual rainfall is 148 450 mm and the mean annual temperature is 17 ºC, with a maximum of 33 ºC (in July) and a 149 minimum of 5 ºC (in January). For more details, see the area description in Domínguez et al. 150 (2008) and Madejón et al. (2018a). 151 The study area was affected by a mine-spill (in April 1998) that polluted the soil with 152 trace elements. After the spill, the soil was cleaned up, remediated and afforested with native 153 species of shrubs and trees in mixed patterns to simulate a diverse forest (Madejón et al. 154 2018b). In a plot of about 14 ha (37º 23.165´ N, 6º 13.668´ W) within the remediated area we 155 randomly selected five replicates of seven tree species (35 tree samples in total), with an 156 average distance of more than 100 metres between replicates of the same species, resembling 157 a “common-garden experiment”. The area was afforested in 2000, using seedlings (1-2 years 158 old) grown in a nearby nursery. The tree species were selected for this study according to 159 their contrasting leaf habits: deciduous species (Populus alba L., Celtis australis L. and 160 Fraxinus angustifolia Vahl) and evergreen species (Quercus ilex subsp. ballota (Desf.) 161 Samp., Olea europaea subsp. europaea var. sylvestris (Mill.) Lehr, Ceratonia siliqua L. and 162 Pinus pinea L.); hereafter we use only the genus name for simplicity. 163 The soil in the plot is of the Fluvisol type, being acidic (pH below 5) and nutrient-164 poor, with a loamy texture (about 20% sand). In the spill-affected and remediated soils the 165 residual contamination by trace elements such as As, Cd, Cu, Pb and Zn was still high during 166 the study (16 years after the spill). However, there was a low transfer rate of trace elements to 167 the aboveground parts of the woody plants (Domínguez et al. 2008; Madejón et al. 2018a, 168 2018b). 169 170 Trait measurements 171 8  Roots were sampled (in late autumn, December 2014) at the individual tree level, by 172 excavating the first 20–30 cm of the soil adjacent to the tree trunk base. We selected the fine 173 roots (< 2 mm in diameter) for the trait analysis. Although fine roots are composed by 174 absorptive and transport roots (McCormack et al. 2015), we assume that, given the small 175 diameter of the selected roots (range from 0.35 to 0.53 mm), most of them should be 176 absorptive. 177 In the selected fine roots we measured the following morphological traits: specific root 178 length (SRL, root length per unit of root dry mass, m g-1), root mass per area (RMA, root 179 mass per unit of root area, g m-2), root mean diameter (RDI, mm), root dry matter content 180 (RDMC, root dry mass per unit of water-saturated fresh mass, mg g-1) and root tissue mass 181 density (RTD, root dry mass per unit of root volume, mg cm-3). We followed methods in 182 Pérez-Harguindeguy et al. (2013) and de la Riva et al. (2016a) to characterise these variables. 183 However, in this study, we adopted the trait RMA as an analogue of leaf LMA, being the key 184 functional trait for roots (see the arguments in favour of using LMA in Poorter et al. 2009 and 185 de la Riva et al. 2018b). The roots were scanned with an EPSON Perfection V700 photo 186 scanner at 1200 dpi. The length, diameter, area and volume of the roots were obtained by 187 analysing the scanned root samples with WinRHIZO 2009 software (Regent Instruments Inc., 188 Quebec, Canada). The mycorrhizal type associated with each tree species was assigned 189 according to several sources (Maremmani et al. 2003; Manaut et al. 2015; Navarro-Fernández 190 et al. 2016). We assigned the “ectomycorrhizal type” (ECM) to those tree species (Pinus, 191 Populus and Quercus) that form symbiotic associations predominantly with ECM fungi, 192 although they can also associate with arbuscular mycorrhizal (AM) fungi. In contrast, trees of 193 the “arbuscular mycorrhizal type” (AM) are exclusively associated with AM fungi (see Table 194 S1). 195 9  In the selected trees, morphological traits of fully-expanded leaves were recorded 196 following the methods of Pérez-Harguindeguy et al. (2013). Young, fully expanded leaves 197 still attached to a portion of stem of the previous year were collected from each individual 198 tree. These stems with leaves were stored in plastic bags to prevent water loss and transported 199 to the laboratory, where they were maintained with the basal portion of the stem submerged in 200 water at 10 ºC for 24 h, in darkness, to allow complete re-hydration (de la Riva et al. 2016a). 201 They were sampled in early autumn (October 2014), when we expect them to have their 202 maximum concentrations of chemical elements (Madejón et al. 2004, 2006; Domínguez et al. 203 2008). We measured the leaf mass per area (LMA, leaf dry mass per unit of area, g m−2) and 204 leaf dry matter content (LDMC, dry mass per unit of water-saturated fresh mass; mg g-1). We 205 also measured the stem dry matter content (SDMC, dry mass per unit of water-saturated fresh 206 mass; mg g-1) and stem wood density (SWD, dry mass divided by the stem fresh volume; mg 207 cm-3; based on the Archimedes principle, measuring the volume of water displaced by 208 immersion of the stem) in the sampled branches and twigs. One of the Fraxinus trees suffered 209 summer defoliation and only had young leaves; therefore, it was excluded from the leaf traits 210 dataset (n=34). 211 A subsample of the roots and leaves collected from each tree was dried and then 212 ground using a stainless steel mill, for chemical analyses. The N and C concentrations and the 213 isotopic ratios of nitrogen (δ15N) and carbon (δ13C) were determined, in leaf and root samples 214 combusted at 1020 ºC, using a continuous flow isotope-ratio mass spectrometry system. This 215 involved a Flash HT Plus elemental analyser coupled to a Delta-V Advantage isotope-ratio 216 mass spectrometer via a CONFLO IV interface (Thermo Fisher Scientific, Bremen, 217 Germany); the analytical measurement errors were ± 0.2‰ for δ15N and ± 0.1‰ for δ13C. The 218 concentrations of macro- (P, K, Ca, Mg, S) and micronutrients (B, Co, Cu, Fe, Mn, Na, Ni, 219 Zn), as well as non-essential elements (As, Ba, Cd, Pb, Sr), were determined after wet 220 16  sylvestris that thicker roots have low densities due to a thicker cortex, thus making them less 371 costly to construct and more suitable to association with mycorrhizal fungi, and enhancing 372 nutrient acquisition. Also, Kong et al. (2019) found nonlinear root trait relationships between 373 RDI, RTD and SRL, which can explain why SRL does not necessarily conform to the RTD-374 related plant economics spectrum in woody species. 375 The key morphological traits RMA and SRL, which are indicators of the root uptake 376 potential, correlated with other morphological traits and with some major nutrients (P, K, Ca 377 and Mg), which supports the uptake function of these traits. Phosphorus and K reach the roots 378 mainly by diffusion from the bulk soil to the root surface (Lambers et al. 2008), and therefore 379 a negative correlation between their uptake and RMA (but a positive one with SRL) would be 380 expected. The correlation between RMA and root nutrients (other than N) has been 381 overlooked previously, and this study is a relevant contribution to support the RES. Despite 382 the fact that there are contradictory results concerning the SRL – root N relationship (see 383 reviews in Reich 2014 and Weemstra et al. 2016), based on our data we can concur with 384 Reich (2014) that the RES exists, although not as uniformly and strongly coordinated as the 385 LES. 386 This evidence supporting the RES as a main root dimension does not preclude the 387 existence of other root dimensions representing the multifunctionality of roots (Weemstra et 388 al. 2016; Laliberté 2017). 389 Firstly, in this study, root C concentration was not correlated with the morphological 390 root traits, but it was negatively correlated with the concentration of root N and 12 other 391 chemical elements. Root C concentration, together with root branching traits (not measured 392 here), defined the second dimension in the PCA of 14 root traits of 96 woody species from 393 subtropical forests in China (Kong et al. 2014). 394 17  Secondly, soil resources are multiple (water and nutrients) and plant roots differ in 395 their uptake strategies, such as mycorrhizas, N2-fixing symbioses, and P-absorbing cluster 396 roots (Lambers et al. 2008). The micronutrients Cu, Mn and Ni and the non-essential element 397 Pb exhibited a trend that was orthogonal of that of the RES (Fig. 2). In particular, the Pb 398 concentration in roots was correlated negatively with RMA and root C, but positively with 13 399 other elements. Among the root traits which confer improved tolerance to elevated metal 400 concentrations in soils is the ability to bind trace elements to root cell walls and accumulate 401 them belowground; in this way, roots may be barriers impeding the uptake of potentially toxic 402 elements and their translocation to the leaves (Lambers et al. 2008, Zhao et al. 2016). This 403 additional root dimension that confers metal tolerance may be very important for plant fitness 404 in metal-rich environments, like the study site. 405 Thirdly, the N isotope composition (δ15N) has been used to infer symbiotic uptake of 406 N by mycorrhizal fungi and its transfer to plants, due to the discrimination against heavier 15N 407 in these processes (Hobbie and Hobbie 2008; Hobbie and Högberg 2012). In this study, root 408 δ15N was a relatively-independent trait; it was correlated negatively with the root C 409 concentration and δ13C but there was no relationship with morphological or chemical traits 410 (with the exception of root Co and Mn). Recently, Laliberté (2017) has suggested the use of 411 the N isotope composition in plants as a time-integrated trait showing the mycorrhizal 412 influence on N acquisition. Thus, we would expect lower δ15N values in the roots of trees 413 associated with ECM fungi, which discriminate against 15N and preferentially transfer 14N to 414 their host plants; while no or only slight depletion of 15N is expected for AM plants (Hobbie 415 and Högberg 2012; Craine et al. 2015). However, we found that the δ15N values in roots of 416 ECM trees (mean=1.61, n=15) were not different from those in AM type trees (mean=1.40, 417 n=20). The natural abundance of 15N in plants is not easy to interpret because it is a single 418 18  response variable with multiple drivers (i.e. climate, mycorrhizal fungi, and microbial 419 processing; Craine et al. 2015). 420 421 Coordination of root and leaf traits 422 Fast plant growth depends on the coordination of roots and leaves, with the former 423 providing enough water and nutrients supply to maintain acquisitive leaves with high 424 photosynthetic rates and high evaporative demand (Reich 2014). In general, we found that the 425 main root variation trend (PCA axis 1) was significantly correlated with the corresponding 426 leaf variation trend (Fig. 4), supporting the existence of a plant economics spectrum (Pérez-427 Ramos et al. 2012, de la Riva et al. 2016b, 2018a). 428 In particular, we found significant correlations between morphological root traits 429 (RMA and RDMC) and the analogous leaf traits (LMA and LMDC), supporting such root-430 leaf coordination, as reported in other studies (Holdaway et al. 2011; de la Riva et al. 2018a). 431 However, there are exceptions: for instance, Larix decidua trees display acquisitive leaf traits, 432 typical of deciduous trees, but conservative root traits, typical of conifers (Withington et al. 433 2006; Weemstra et al. 2016). 434 In this study, the C concentrations in roots and leaves were positively correlated. 435 Villar et al. (2006) also found a positive correlation between root and leaf C in 16 woody 436 species. The C concentration is normally high for species with strong structural defences 437 (such as lignin or cellulose) (Poorter and Villar 1997) and, therefore, with a conservative 438 strategy (de la Riva et al. 2016b). On the other hand, there was not a significant relationship 439 between RMA and the root C concentration, although RMA was positively correlated with the 440 C:N ratio (r=0.55, p=0.001). The C:N ratio reflects the relative investments in structure 441 (mainly carbon) respect to cell metabolism (indicated by nitrogen). Thus, plant organs with a 442 19  higher C:N ratio represent a conservative strategy (Villar et al. 2006; de la Riva et al. 2016c, 443 2018b). 444 Our results do not fit the previously-reported global trend of root and leaf N 445 concentrations, which are highly correlated in woody species (n=89, r=0.58, p<0.001; 446 Valverde-Barrantes et al. 2017). Trees with N-rich leaves and high photosynthetic rates are 447 expected to have N-rich and exploitative fine roots (Reich 2014). However, this trend can be 448 influenced by the specific symbiosis (type of mycorrhiza) present. In this study, we found a 449 significant correlation only when analysing the subset of ECM tree species (r=0.72, p=0.002). 450 However, tree species in a symbiosis with AM fungi did not show this root-leaf N 451 relationship. The two types of mycorrhizal trees have different nutrient economies: ECM trees 452 are able to acquire N from the soil organic matter due to the greater enzymatic capabilities of 453 ECM fungi, while AM trees depend mostly on inorganic N (Phillips et al. 2013). In the N-454 limiting conditions of the study site, we expect ECM trees to be more efficient at taking up 455 soil N, through the root-fungi symbiosis, and translocating it to their leaves. Kong et al. 456 (2019) also found different relationships of root traits depending of mycorrhizal types (ECM 457 versus AM). Thus, for ECM species, thin roots were related with higher root N concentration, 458 but the contrary for AM species. This could explain the positive relationship found in our 459 study between root N and leaf N only for ECM species. 460 The C isotope composition in leaf tissues is widely used as a functional trait 461 representing the time-integrated measurement of water-use efficiency. It is based on the 462 discrimination by photosynthetic enzymes against the heavier isotope 13C during 463 photosynthesis, and depends on the ratio between the internal and air CO2 concentrations, in 464 turn regulated by stomatal opening (Seibdt et al. 2008, Pérez-Harguindeguy et al. 2013). The 465 δ13C values in roots should reflect the isotopic signature of the carbohydrates synthesised in 466 the leaves, although during the leaf-root translocation some 13C enrichment in roots (relative 467 20  to leaves) has been observed (Cernusak et al. 2009). In this study, the δ13C values in roots and 468 leaves were positively correlated (Table 1), indicating root-leaf coordination. In general, long-469 lived tissues are associated with a more-conservative use of resources and a higher efficiency 470 in water-use, usually reflected in their higher δ13C values (Reich 2014, de la Riva et al. 471 2016b). However, in this case, the δ13C values in roots of deciduous and evergreen trees were 472 not different (Table S4). 473 The plant N isotope composition reflects mainly the soil source of N, and also any 474 isotope fractionation and N pool mixing (Robinson 2001). Although the root values for δ15N 475 did not show significant differences among species, when analysing the intra-plant 476 fractionation (i.e. δ15Nroot – δ 15Nleaf) there were significant differences among species and 477 mycorrhizal types (Fig. S5). The depletion of 15N in ECM trees may be related to the 478 preferential retention of 15N by the ECM fungal biomass (but not by that of AM fungi) and 479 the consequent transfer of 15N-depleted N to the host trees (Craine et al. 2015). 480 Nutritional differences among tree species result from the functional diversity in 481 mechanisms of nutrient uptake from soil, nutrient requirements and long-term nutrient use 482 efficiency (Lambers et al. 2008). The coordinated variability in P, Ca and Mg concentrations 483 between roots and leaves indicates that these nutrients are under biological control, due to 484 their importance for plant growth (Newman and Hart 2006; Geng et al. 2014; Zhao et al. 485 2016). 486 In contrast, most of the trace elements had a strong discordance between their 487 concentrations in roots and leaves. The excess uptake of non-limiting elements seems poorly 488 regulated by plants, and therefore they exhibit high variability (Ladanai et al. 2010). Plants 489 tend to accumulate trace elements in roots, binding them to cell walls as a detoxification 490 mechanism (Domínguez et al. 2009; Kabata-Pendias 2011; Zhao et al. 2016). However, some 491 tree species have a selective uptake and transport of certain trace elements, accumulating 492 21  them in leaf tissues. Notable examples are the accumulation of Cd and Zn in Populus leaves 493 (Madejón et al. 2004) and the accumulation of Mn in Quercus leaves (Madejón et al. 2006), 494 but not in their roots (Fig. S4). In soils contaminated by trace elements, the adequate selection 495 of plant species for phytostabilisation is essential. One of the main criteria is that the selected 496 tree species control the mobility of the trace elements, keeping their root to shoot 497 translocation factors as low as possible, to avoid toxicity risks in the trophic web (Mendez and 498 Maier 2008; Bolan et al. 2011; Madejón et al. 2018b). 499 500 Root traits and soil conditions 501 There are reciprocal interactions and feedbacks between roots and soil. The soil 502 conditions influence root traits and plasticity (Bardgett et al. 2014). In turn, roots modify the 503 rhizospheric soil; for example through root exudates to increase nutrient uptake (Dakora and 504 Phillips 2002). Here, soil pH was significantly related to the root morphological traits. On the 505 one hand, this indicates that roots with traits indicative of lower exploration (higher RMA or 506 lower SRL, as in Pinus and Quercus) could compensate with higher production of acid 507 exudates to promote nutrient uptake, decreasing the soil pH (Dakora and Phillips 2002). On 508 the other hand, this relationship between soil acidity and a conservative root strategy could be 509 linked to the effects of the litter compounds of these species (with a high C:N ratio and high 510 LMA) on soil. The accumulation of litter with a high C:N ratio, such as that of coniferous 511 species, tends to have an acidifying effect on soil (Augusto et al. 1998; Sariyildiz et al. 2005; 512 Alameda et al. 2012). As a consequence, soil pH usually decreases after the afforestation of 513 grasslands or former agricultural lands with coniferous species (Jug et al. 1999; Sauer et al. 514 2007; Berthrong et al. 2012). The analysis of the amount and quality of root exudates would 515 be needed to elucidate the causes behind the observed relationship between soil pH and RMA, 516 besides the indirect effects of litter traits on soil chemistry. In any case, as suggested by 517 22  Laliberté (2017), it would be worth including root exudation as a physiological trait to 518 advance in trait-based plant ecology. 519 In metal-rich soils low pH usually leads to a higher solubility of these elements and 520 therefore to a high availability to roots. In this trace-element polluted site, we observed some 521 significant relationships between soil metal content and some root traits; in particular, RDMC 522 was positively correlated to the soil content of Mn, Ni and Cd. One of the first symptoms of 523 plant toxicity to soil metals is the inhibition of root elongation (Kahle 1993; Wisniewski and 524 Dickinson 2003). Other responses to metal toxicity are: collapsing of root hairs, increments of 525 suberification and lignification, decrease of vessel diameter and structural alterations of 526 hypodermis and endodermis (Arduini et al. 1994; Barceló and Poschenrieder 2004). 527 Experimental exposure of Quercus ilex roots to Cd resulted in a decline in fine root 528 production and in a reduction in the length of taproots (Domínguez et al. 2009), linked to a 529 high capacity to retain Cd at the root level, likely by binding Cd to cell wall pectins. Thus, the 530 links between pH, soil metal content, and RDMC observed in this study could also indicate a 531 trend towards a more conservative strategy at the root level to promote the immobilization of 532 these metals in the rhizosphere, avoiding their translocation to the aboveground biomass. 533 The root chemical traits were related to the availability in the soil for some nutrients 534 (K, Mn, Na and Zn), as expected. However, the concentrations of many other elements in the 535 roots were relatively independent of the soil conditions; this weak coupling between the soil 536 and plant concentrations of chemical elements has been found in other studies (Ladanai et al. 537 2010; Zhao et al. 2016). The uptake and accumulation of nutrients in roots is a complex 538 process which depends on numerous factors - such as the relative allocation within the plant, 539 the developmental stage, the plant species and the environmental conditions (Lambers et al. 540 2008). More research is needed to understand how those factors affect the transfer of trace 541 elements from the soil to the roots. 542 23  A particularly-interesting root physiological trait is the potential to reduce metal 543 availability in soil, by several mechanisms like precipitation of metals, their complexation 544 with organic products, their sorption onto root surfaces or their accumulation inside root 545 tissues (Mendez and Maier 2008). The planting of tree species with higher phytostabilisation 546 potential would improve and remediate metal-contaminated soils (Madejón et al. 2018b). In 547 fact, one of the criteria used to select the best-suited tree species is to have a high 548 bioconcentration factor (root:soil ratio) for different metals, in particular for those with 549 harmful effects (i.e. Cd and Pb) (Madejón et al. 2018b). 550 551 Conclusion 552 There is increasing interest in advancing our knowledge about root traits because of their 553 often-overlooked but essential contribution to plant functional ecology. Our results reinforce 554 the existence of a root economics spectrum (RES) as the main determinant of fine root traits 555 in Mediterranean trees, even in soil contaminated by heavy metals. However, this study also 556 supports the idea of root multifunctionality and the importance of fine root dimensions 557 independent of the RES; namely, root carbon concentration, fractionation of nitrogen isotopes 558 as a time-integrated trait of mycorrhizal-mediated nutrition, and the ability to bind trace 559 elements in root cells (associated with tolerance of high levels of metals in soils). We found 560 that roots and leaves were functionally coordinated; however, most of trace elements showed 561 strong root-leaf discordance. We also found links between soil pH, soil metal content, and 562 root traits (RDMC) promoting the immobilization of metals in the rhizosphere. In summary, 563 the rhizosphere is a complex environment where soil, roots and microorganisms interact in 564 feedback processes. An understanding of the multifunctionality of root traits would help us to 565 predict the forest responses to global changes and the provision of soil-based ecosystem 566 services. 567 24  568 Acknowledgements 569 This work was financially supported by the European Union Seventh Framework 570 Programme (FP7/2007–2013) (Grant No. 603498RECARE), the Spanish Ministry of 571 Science, Innovation and Universities (Grants No. CGL2014-52858-R-RESTECO, CGL2017-572 82254-R-INTARSU, and CGL2014-53236-RECO-MEDIT), and European FEDER funds. 573 MG-M was supported by the Spanish Ministry of Economy and Competitiveness (Grant No. 574 BES-2015-073882), and MTD by the Universidad de Sevilla (Contrato de Acceso, V Plan 575 Propio de Investigación). We thank J. M. Murillo and J.M. Alegre for their help in the field 576 work, the IRNAS Analytical Service for multielement analyses of plants and soil, and the 577 EBD-CSIC Laboratory of Stable Isotopes for determinations of δ15N and δ13C. 578 579 Conflict of interest 580 The authors declare that they have no conflict of interest. 581 582 References 583 Alameda D, Villar R (2012) Linking root traits to plant physiology and growth in Fraxinus 584 angustifolia Vahl. seedlings under soil compaction conditions. Environ Exper Bot 585 79:49-57. 586 Alameda D, Villar R, Iriondo JM (2012) Spatial pattern of soil compaction: Trees´ footprint 587 on physical properties. For Ecol Manage 283:128–137. 588 Arduini I, Godbold DL, Onnis A (1994) Cadmium and copper change root growth and 589 morphology of Pinus pinea and Pinus pinaster seedlings. Physiol Plantarum 92:675–590 680. 591 25  Augusto L, Bonnaud P, Ranger J (1998) Impact of tree species on forest soil acidification. For 592 Ecol Manage 105:67–78. 593 Barceló J, Poschenrieder C (2004) Structural and ultrastructural changes in heavy metal 594 exposed plants. In MNV Prasad (Ed.) Heavy metal stress in plants, 2nd edition (pp. 223-595 248). Springer, Berlin, Heidelberg. 596 Bardgett RD, Mommer L, de Vries FT (2014) Going underground: root traits as drivers of 597 ecosystem processes. Trends Ecol and Evol 29:692-699. 598 Berthrong ST, Piñeiro G, Jobbagy EG, Jackson RB (2012) Soil C and N changes with 599 afforestation of grasslands across gradients of precipitation and plantation age. Ecol 600 Appl 22:76–86. 601 Bolan NS, Park JH, Robinson B, Naidu R, Huh KY (2011) Phytostabilization. A green 602 approach to contaminant containment. Adv Agron 112:145–204. 603 Cernusak LA, Tcherkez G, Keitel C, Cornwell WK, Santiago LS, Knohl A, Barbour MM, 604 Williams DG, Reich PB, Ellsworth DS, Dawson TE, Griffiths HG, Farquhar GD, 605 Wright IJ (2009) Why are non-photosynthetic tissues generally 13C enriched compared 606 with leaves in C3 plants? Review and synthesis of current hypotheses. Funct Plant Biol 607 36:199-213. 608 Chagnon PL, Bradley RL, Maherali H, Klironomos JN (2013) A trait-based framework to 609 understand life history of mycorrhizal fungi. Trends Plant Sci 18:484-491. 610 Comas L, Becker S, Cruz VMV, Byrne PF, Dierig DA (2013) Root traits contributing to plant 611 productivity under drought. Front Plant Sci 4:442. 612 Craine JM, Brookshire ENJ, Cramer MD, Hasselquist NJ, Koba K, Marin-Spiotta E, Wang L 613 (2015) Ecological interpretations of nitrogen isotope ratios of terrestrial plants and soils. 614 Plant Soil 396:1-26. 615 32  R Development Core Team (2018) R: A Language and Environment for Statistical 764 Computing. 765 Reich PB (2014) The world-wide ‘fast–slow’plant economics spectrum: a traits manifesto. J 766 Ecol 102:275-301. 767 Robinson D (2001) δ15N as an integrator of the nitrogen cycle. Trends Ecol Evol 16:153-162. 768 Rodríguez-Eugenio N, McLaughlin M, Pennock D (2018) Soil Pollution: a hidden reality. 769 FAO, Rome. 770 Roumet C, Birouste M, Picon-Cochard C, Ghestem M, Osman N, Vrignon-Brenas S, Cao K, 771 Stokes A (2016) Root structure–function relationships in 74 species: evidence of a root 772 economics spectrum related to carbon economy. New Phytol 210:815-826. 773 Sariyildiz T, Anderson JM, Kucuk M (2005) Effects of tree species and topography on soil 774 chemistry, litter quality, and decomposition in Northeast Turkey. Soil Biol Biochem 775 37:1695-1706. 776 Sauer TA, Cambardella C, Brandle J (2007) Soil carbon and tree litter dynamics in a red 777 cedar-scotch pine shelterbelt. Agroforest Syst 71:163–174. 778 Seibt U, Rajabi A, Griffiths H, Berry J A (2008) Carbon isotopes and water use efficiency: 779 sense and sensitivity. Oecologia 155:441-454. 780 Smith SE, Read DJ (2008) Mycorrhizal symbiosis, 3rd edition. Academic Press, London. 781 Soudzilovskaia NA, van der Heijden MG, Cornelissen JHC, Makarov MI, Onipchenko VG, 782 Maslov MN, Akhmetzhanova AA, van Bodegom PM (2015) Quantitative assessment of 783 the differential impacts of arbuscular and ectomycorrhiza on soil carbon cycling. New 784 Phytol 208:280-293. 785 Tjoelker MG, Craine JM, Wedin D, Reich PB, Tilman D (2005) Linking leaf and root trait 786 syndromes among 39 grassland and savannah species. New Phytol 167:493–508. 787 33  Tripathi DK, Singh S, Gaur S, Singh S, Yadav V, Liu S, Singh VP, Sharma S, Srivastava P, 788 Prasad SM, Dubey NK, Chauhan DK, Sahi S (2018) Acquisition and homeostasis of 789 iron in higher plants and their probable role in abiotic stress tolerance. Front Environ 790 Sci 5:86. 791 Valverde-Barrantes OJ, Freschet GT, Roumet C, Blackwood CB (2017) A worldview of root 792 traits: the influence of ancestry, growth form, climate and mycorrhizal association on 793 the functional trait variation of fine-root tissues in seed plants. New Phytol 215:1562-794 1573. 795 Villar R, Ruiz-Robleto J, De Jong Y, Poorter H (2006) Differences in construction costs and 796 chemical composition between deciduous and evergreen woody species are small as 797 compared to differences among families. Plant Cell Environ 29:1629–1643. 798 Weemstra M, Mommer L, Visser EJ, Ruijven J, Kuyper TW, Mohren GM, Sterck FJ (2016) 799 Towards a multidimensional root trait framework: a tree root review. New Phytol 800 211:1159-1169. 801 Wisniewski L, Dickinson NM (2003) Toxicity of copper to Quercus robur (English Oak) 802 seedlings from a copper-rich soil. Environ Exp Bot 50:99–107. 803 Withington JM, Reich PB, Oleksyn J, Eissenstat DM (2006) Comparisons of structure and life 804 span in roots and leaves among temperate trees. Ecol Monogr 76:381-397. 805 Wright IJ, Reich PB, Westoby M, Ackerly DD, Baruch Z, Bongers F, Cavender-Bares J, 806 Chapin T, Cornelissen JHC, Diemer M, Flexas J, Garnier E, Groom PK, Gulias J, 807 Hikosaka K, Lamont BB, Lee T, Lee W, Lusk C, Midgley JJ, Navas ML, Niinemets U, 808 Oleksyn J, Osada N, Poorter H, Poot P, Prior L, Pyankov VI, Roumet C, Thomas SC, 809 Tjoelker MG, Veneklaas EJ, Villar R (2004). The worldwide leaf economics spectrum. 810 Nature 428:821-827. 811 34  Zadworny M, McCormack ML, Żytkowiak R, Karolewski P, Mucha J, Oleksyn J (2017) 812 Patterns of structural and defense investments in fine roots of Scots pine (Pinus 813 sylvestris L.) across a strong temperature and latitudinal gradient in Europe. Global 814 Change Biol 23:1218-1231. 815 Zhao N, Yu G, He N, Wang Q, Guo D, Zhang X, Wang R, Xu Z, Jiao C, Li N, Jia Y (2016) 816 Coordinated pattern of multi-element variability in leaves and roots across Chinese 817 forest biomes. Global Ecol Biogeogr 25:359-367. 818 35  Table 1. Correlations between root traits and analogous leaf traits (n=34), and between root chemical traits and soil availability of elements (n=35). Soil availability (CaCl2 extracted) of As and Pb were below detectable limits. Pearson´s test coefficient r and p values are indicated; significant values (p<0.05) are in bold. RMA: root mass per area; LMA: leaf mass per area; LDMC: leaf dry matter content; RDMC: root dry matter content. Root-leaf Root-soil Trait r p r p RMA/LMA 0.548 0.001 - - LDMC/RDMC 0.371 0.031 - - C 0.553 0.001 -0.110 0.528 N 0.018 0.920 -0.020 0.910 P 0.486 0.004 -0.111 0.526 K 0.289 0.098 0.379 0.025 Ca 0.826 <0.001 0.166 0.340 M g 0.393 0.021 0.163 0.348 B 0.075 0.675 0.025 0.887 Co -0.250 0.154 0.380 0.024 Cu 0.087 0.623 -0.002 0.990 Fe 0.131 0.461 -0.073 0.678 Mn -0.023 0.897 0.357 0.035 Na 0.051 0.775 0.596 <0.001 Ni -0.070 0.695 -0.031 0.861 S 0.023 0.899 0.134 0.441 Zn -0.093 0.602 0.346 0.042 As -0.001 0.998 - - Ba 0.516 0.002 -0.028 0.872 C d 0.019 0.914 0.071 0.686 Pb 0.331 0.056 - - S r 0.774 <0.001 0.381 0.024 δ13C 0.391 0.022 -0.174 0.317 δ15N 0.165 0.351 0.437 0.009 36  Figure legends Figure 1. Variation among tree species in root mass per area (RMA, g m-2). Mean and SE (n=5) bars are shown; letters indicate significant differences between the tree species (Tukey´s post-hoc test). Deciduous species are marked in grey and evergreen species in black. Figure 2. Results of the principal component analysis of 27 root traits of seven tree species (n=35). Scores of trait variables and tree samples are represented in the plane defined by first (PC1) and second (PC2) axes. Abbreviations of root traits are: RMA: root mass per area; RDMC: root dry matter content, RDI: root mean diameter; RTD: root tissue density; SRL: specific root length; and for tree species names are: C.a.: Celtis australis; C.s.: Ceratonia siliqua; F.a.: Fraxinus angustifolia; O.e.: Olea europaea; P.a.: Populus alba; P.p.: Pinus pinea; Q.i.: Quercus ilex. Symbol fills are in grey for deciduous and in black for evergreen species. Figure 3. Correlogram across morphological and chemical root traits, ordered according to their correlation coefficients. The strength and direction of the correlations are indicated by the circle size and the colour, shown in the right side scale. RTD: root tissue density; RDMC: root dry matter content; RMA: root mass per area; RDI: root mean diameter; SRL: specific root length. Figure 4. Correlation analysis between the main variation trends (PCA axis 1 scores) in leaf and root, comparing 24 analogous traits (r=0.59, p=0.0003). Abbreviations of tree species names are: C.a.: Celtis australis; C.s.: Ceratonia siliqua; F.a.: Fraxinus angustifolia; O.e.: Olea europaea; P.a.: Populus alba; P.p.: Pinus pinea; Q.i.: Quercus ilex. Symbol fills are in grey for deciduous and in black for evergreen species. 37  Figure 1 38  Figure 2 39  Figure 3 40  Figure 4 41  7  Table S2. Results of the principal component analyses (PCA) for the six key root traits (see ordination in Figure S2) and for all 27 traits (see Figure 2), indicating variance explained by the three main axes and standardized factor loading of each root trait. The highest scores (in absolute value) for each axis are marked in bold. See main text for abbreviations of trait names. 6 key traits PCA All-traits PCA Traits Axis 1 (57.3%) Axis 2 (23.4%) Axis 3 (14.7%) Axis 1 (37.6%) Axis 2 (18.2%) Axis 3 (11.0%) RMA 0.97 -0.06 0.05 0.61 0.61 0.13 SRL -0.87 -0.42 -0.04 -0.49 -0.58 -0.23 RDMC 0.87 -0.28 0.23 0.56 0.71 -0.01 RTD 0.60 -0.76 0.16 0.43 0.42 -0.31 RDI 0.68 0.71 -0.06 0.36 0.40 0.45 N -0.38 0.25 0.89 -0.65 0.15 -0.26 C 0.74 -0.53 0.34 Ca -0.62 0.07 -0.62 Mg -0.88 -0.23 0.09 K -0.35 -0.70 0.17 S -0.70 0.14 -0.05 P -0.45 -0.51 -0.11 As -0.86 0.31 0.04 B -0.31 -0.72 -0.12 Ba -0.69 0.24 -0.09 Cd -0.28 -0.31 0.69 Co -0.70 0.38 0.50 Cu -0.82 0.21 0.30 Fe -0.88 0.29 0.06 Mn -0.66 0.32 0.37 Na -0.27 -0.62 0.51 Ni -0.73 0.45 -0.08 Pb -0.88 0.29 -0.04 Sr -0.57 -0.01 -0.61 Zn -0.67 -0.10 0.40 δ13C -0.14 -0.35 -0.41 δ15N -0.17 0.54 0.21 8  Table S3. Mean values and SE (n=5) of aboveground traits of the studied tree species (except n=4 for leaf traits of Fraxinus). F-statistics from one-way ANOVA test or Chi-square-value from Kruskal Wallis test (marked with Ksupersript) are shown, depending on data normality and homoscedasticity. Significant level is p < 0.05 (in bold). LDMC: leaf dry matter content; LMA: leaf mass per area; RDI: root mean diameter; SRL: specific root length; SDMC: stem dry matter content; SWD: stem wood density; HEI: tree height; CRP: crown projection area; LITT: litter accumulation on soil surface.  9  Table S4. Comparison between root traits of trees, according to their leaf habit (deciduous or evergreen); trait units are like in Table S1. They have been ranked by the significance level (ANOVA´s F and p) marking the difference between leaf habits. Significant level is p < 0.05 (in bold). SRL: specific root length; RDI: root mean diameter; RMA: root mass area; RDMC: root dry matter content; RTD: root tissue density. Root trait Deciduous (n=15) Evergreen (n=20) ANOVA statistics Mean SE Mean SE F p B 20.21 1.29 11.77 0.58 42.4 <0.001 P 1054.3 53.2 680.5 32.2 40.0 <0.001 SRL 21.76 2.05 10.62 0.65 33.7 <0.001 RDI 0.37 0.02 0.50 0.01 27.9 <0.001 RMA 43.20 2.08 63.89 3.34 23.2 <0.001 RDMC 267.3 16.6 354.6 13.2 17.4 <0.001 Ca 18329.2 2619.7 7613.8 1391.3 14.9 <0.001 Mg 1835.4 99.7 1224.5 121.6 13.7 <0.001 Sr 34.15 4.10 15.29 3.17 13.7 <0.001 K 8583.7 865.9 5983.3 384.9 8.9 0.005 Zn 163.4 17.3 110.1 9.80 8.1 0.008 Pb 42.28 8.72 20.10 3.36 6.9 0.013 As 13.15 2.97 6.62 1.10 5.2 0.029 S 2419.9 237.4 1652.9 236.7 5.0 0.032 Fe 3739.8 696.0 2208.6 339.8 4.6 0.040 Ba 13.57 1.50 10.17 0.92 4.1 0.051 Ni 9.11 2.14 5.49 0.57 3.4 0.073 N 1.20 0.14 0.92 0.09 3.0 0.092 C 42.9 1.17 44.6 0.36 2.7 0.111 Cu 112.2 12.4 85.8 11.9 2.3 0.141 Mn 102.8 15.3 78.3 8.4 2.2 0.143 Cd 1.55 0.23 1.16 0.14 2.2 0.148 Na 786.5 172.5 520.5 93.0 2.1 0.157 RTD 474.3 25.3 511.5 24.4 1.1 0.305 δ13C -26.2 0.19 -26.6 0.25 1.0 0.314 Co 2.35 0.4 1.99 0.5 0.6 0.444 δ15N 1.40 0.2 1.55 0.2 0.2 0.664 10  Table S5. Mean values and SE (n=5) of topsoil parameters (0-10cm depth) associated to the studied tree species, and adjacent open sites, for comparison; pH, organic C, total N, available concentrations of nutrients and trace elements, and C and N isotope ratios. F-statistics from one-way ANOVA test or Chi-square-value from Kruskal Wallis test (marked with Ksuperscript) are indicated, depending on data normality and homoscedasticity (only soil samples under trees were compared). Significant level is p < 0.05 (in bold). 11  Table S6. Correlations of root morphological traits with soil chemical variables (pH, organic C, total N, available concentrations of nutrients and trace elements, and C and N isotope ratios). Significant values (p<0.05) are in bold. RMA: root mass per area; SRL: specific root length; RDI: root mean diameter; RDMC: root dry matter content; RTD: root tissue density. Soil property Morphofunctional root trait RMA SRL RDI RDMC RTD r p r p r p r p r p pH -0.51 0.002 0.35 0.040 -0.26 0.130 -0.57 <0.001 -0.42 0.013 C -0.33 0.052 0.22 0.210 -0.07 0.690 -0.35 0.040 -0.35 0.041 N -0.27 0.114 0.22 0.205 -0.07 0.704 -0.32 0.059 -0.30 0.076 P -0.16 0.361 0.18 0.303 -0.22 0.194 0.09 0.597 0.03 0.856 K -0.28 0.101 0.12 0.490 0.01 0.954 -0.41 0.014 -0.35 0.037 Ca -0.32 0.061 0.32 0.061 -0.24 0.161 -0.20 0.252 -0.20 0.253 Mg -0.40 0.016 0.27 0.123 -0.12 0.494 -0.29 0.095 -0.37 0.027 S 0.08 0.637 0.05 0.782 -0.14 0.420 0.32 0.058 0.25 0.142 B -0.21 0.235 0.11 0.537 -0.13 0.473 -0.12 0.484 -0.09 0.600 Ba 0.13 0.451 -0.22 0.197 0.25 0.147 0.12 0.485 -0.04 0.834 Cd 0.26 0.127 -0.11 0.519 0.07 0.691 0.36 0.036 0.30 0.081 Co 0.13 0.459 -0.04 0.799 0.003 0.985 0.32 0.061 0.17 0.325 Cu 0.04 0.814 0.06 0.727 -0.10 0.552 0.33 0.052 0.16 0.354 Fe -0.11 0.526 0.15 0.384 -0.22 0.211 0.19 0.261 0.08 0.632 Mn 0.28 0.100 -0.15 0.396 0.09 0.600 0.43 0.009 0.29 0.090 Na -0.17 0.337 -0.01 0.987 0.16 0.359 -0.41 0.015 -0.38 0.026 Ni 0.32 0.060 -0.15 0.389 0.10 0.587 0.45 0.006 0.32 0.057 Sr -0.38 0.023 0.23 0.178 -0.15 0.379 -0.36 0.036 -0.33 0.056 Zn 0.08 0.661 0.01 0.934 -0.07 0.683 0.31 0.067 0.19 0.271 δ13C 0.07 0.688 -0.04 0.825 -0.004 0.982 0.21 0.235 0.06 0.749 δ15N -0.39 0.020 0.20 0.262 -0.11 0.545 -0.19 0.271 -0.37 0.027 12  Figure S1. Rank of relative variation in root traits and their analogue leaf traits, measured as coefficient of variation (CV in %), of seven species (n=35). R/LDMC: root or leaf dry matter content; R/LMA: root or leaf mass per area; 13C: δ13C and 15N: δ15N. 13  Figure S2. Results of the principal component analysis of six key root traits in trees of seven species (n=35). Abbreviations of root traits are: RDI: root mean diameter; RMA: root mass per area; RDMC: root dry matter content; RTD: root tissue density; and SRL: specific root length; species names are: C.a.: Celtis australis; C.s.: Ceratonia siliqua; F.a.: Fraxinus angustifolia; O.e.: Olea europaea; P.a.: Populus alba; P.p.: Pinus pinea; Q.i.: Quercus ilex. 14  Figure S3. Results of the principal component analysis of 29 aboveground traits in trees of seven species (n=34). Traits and trees are ordered in the plane defined by PCA first and second axes. Abbreviation names for tree species are: C.a.: Celtis australis; C.s.: Ceratonia siliqua; F.a.: Fraxinus angustifolia; O.e.: Olea europaea; P.a.: Populus alba; P.p.: Pinus pinea; Q.i.: Quercus ilex. 15  Figure S4. Coordination between root and leaf traits in the three figures of left column, compared with discordant traits in the right column (n=34). Correlations between N in roots and in leaves are significant for Pinus (red diamond and solid line, n=5) and Populus (blue triangles and dashed line, n=5) but not for the other species and for all data (black circles). Species-specific accumulation of Cd in leaves for Populus (blue triangles) and Mn in Quercus (green squares) are shown. The values of correlation coefficient and significance are shown in Table 1. LMA: leaf mass per area; RMA: root mass per area. 16  Figure S5. Intra-plant fractionation of N isotope (difference between δ15Nroot and δ15Nleaf, in ‰), separating ectomycorrhizal (white) and arbuscular mycorrhizal (grey) tree species. Mean and SE (n=5, with the exception of Fraxinus n=4) are shown; different letters mean significant difference between species by post-hoc Tukey test. There are significant differences between mycorrhizal types, t=-4.4, p<0.0001.