Edaphic Factors as Agents of Genetic Diversity and Adaptation of Native Plant Species of Harsh Environments in the Brazilian Savanna

Purposes To analyze the main edaphic factors that differentiate ultramac from typical Cerrado environments and act as agents of the development of biochemical and morphological mechanisms of species adaptation to these harsh environments; and to determine the genetic diversity of three Cerrado’s shrubland native species (Justicia lanstyakii, Euploca salicoides, and Oxalis hirsutissima). Methods We conducted chemical analysis of metal contents in soils, as well as on elemental composition, and analysis of DNA extracted from leaf tissues of the three species found in both environments. In leaves of E. salicoides grown in both environments we evaluated the changes in the levels of non-structural carbohydrates (NSC), and total proteins. Results The accessions obtained in ultramac soils were closer with each other, indicating genetic similarity and major differences in relation to the accessions collected in the Cerrado area. These differentiations probably are related with higher adaptation to soils rich in metals, mainly Ni in ultramac, and Al in the Cerrado environments. The highest levels of NSC were observed in plants grown in ultramac soils, including ranose, which is related to responses to metal detoxication and drought. The allocation of Ni in the trichomes, which does not affect important processes of plant metabolism, is another mechanism developed by E. salicoides to overcome the hyperaccumulation of Ni in their tissues. Conclusions These ndings can help select seed collection sites representative of the genetic diversity of native plant species for restoring degraded areas or for phytoremediation of metals.


Introduction
The Brazilian tropical savanna (Cerrado biome) occupies an area of more than 203 million hectares in the centre of the country, corresponding to more than 23 % of the national territory. The climate is characterized as Koppen's Aw (tropical savanna climate with dry winter), with two well-de ned seasons: a six-month dry season (May to October) and six-month rainy season (November to April) ). This biome is one of the most diverse tropical savannas, with 11,046 species of phanerogams in which 40% are endemic (Mendonça et al. 2008; Myers et al. 2000). The high diversity of plants and the high degree of endemism found in this biome results from the great heterogeneity of habitats. One of the factors that determine the development of this diversi ed ora is the geochemistry of the soils, mainly due to the presence of metallic elements in high concentrations, such as Ni in soils derived from ultrama c rocks, and Al in acidic Ferralsol.
Ni can display high bioavailability in ultrama c soils. Once it is absorbed by plants, it acts as a cofactor for urease for many plant species, an enzyme that hydrolyzes urea into ammonium and CO 2 , and it is also involved in several other physiological processes, that may include transport of nutrients to the seed and movement of Fe into the plant cells (Brown et al. 1987). It is therefore a micronutrient that is essential for plant metabolism (Pennazio 2012), including hyperaccumulator plants (Sellami et al. 2012). However, when absorbed in high concentrations it can be toxic. Ni toxicity causes water and nutrient imbalances, reducing plant growth and metabolism. On the other hand, Al, a major constituent of the continental crust and especially a predominant metal in acid and highly weathered soils of the Cerrado region, is quite toxic for sensitive species, causing severe effects on root development, plant growth, absorption of water and essential elements, sometimes leading to plant death. In order to overcome these toxicities, the vegetation growing on both ultrama c and non-ultrama c Cerrado soils developed speci c mechanisms of metal tolerance that allowed their survival, involving chemical, physical and biotic components present in such harsh environments. It is noteworthy that several hundreds of Ni and Al hyperaccumulator species have developed mechanisms to tolerate these metals in such environments Jansen et al. 2002;Reeves et al. 2007).
Native species in ultrama c environments of Cerrado biome are natural candidates to recover mining areas or to be used in metal phytoextraction processes in urban and agricultural soils contaminated with these elements. During mining operations, vegetation covering ore deposits is completely removed so that in some cases, it is not possible to restore the original environment throughout mining closure process. Although they are not endemic to the ultrama c environment, some species can exhibit high interspeci c diversity that directly affects recovery strategies. In addition, propagules collected in nonultrama c environments may be poorly adapted for use in mining recovery procedures in ultrama c areas.
The objectives of this work are three-fold: i) to analyze the main edaphic factors that differentiate ultrama c from typical Cerrado environments; ii) to analyze the main edaphic factors that act as agents of the development of biochemical and morphological mechanisms of species adaptation to these harsh environments and iii) to determine the genetic diversity of two Ni hyperaccumulator native species Justicia lanstyachii (Acanthaceae) and Euploca salicoides (Boraginaceae) and one intermediary Oxalis hirsutissima (Oxalidaceae) species that occurs in the Cerrado´s shrubland. Although these species are common in Ni-rich areas, they are not endemic of ultrama c soils, growing in soils with different availability of metals. The three species have the potential to be used in the recovery of degraded areas by Ni mining or in the phytoremediation processes of this metal .
Euploca salicoides, is a facultative hyperaccumulating Ni and Al species and is considered as a "plant model adapted to environments rich in metals". Studies were also conducted in order to clarify which type of tolerance mechanisms (biochemical, morphological) occur in this species that allows its development in areas with high bioavailability of either Ni (ultrama c environment) or Al (typical Cerrado environment). Here, we investigated whether the basal metabolism of plants is altered in the presence of greater availability of metals (mainly Ni and Al) in the soils and the location of Ni in cellular structures in the leaves of plants from the ultrama c area.

Materials And Methods
We carried out this work in ultrama c soils with high Ni bioavailability (SAP), located in the municipality of Barro Alto, Goiás State, and in non-ultrama c typical Cerrado environments constituted by Cambisol (CAM) and Ferralsol (LV), with high Al bioavailability, located in Brasília, Federal District (FD) of Brazil.

Study area
The selected study areas are located in the municipality of Barro Alto, Goiás State and in Federal District (FD) of Brazil (Fig. 1). The rst area is located in the Barro Alto ma c-ultrama c complex which is aligned with the intrusions of Niquelândia and Cana Brava for approximately 350 km extension along a NE-SW structural zone (CPRM 2010). The complex is subdivided into Lower Layered Series and Upper Layered Series. Our study area covers the Lower Layered Series which in turn is subdivided into Lower Ma c Zone and Ultrama c Zone. (Baeta Jr. 1986).
The ultrama c massif of Barro Alto occurs in shallow soils, with rocky outcrops, and strong undulating relief. The vegetation is adapted to the high levels of heavy metals and prolonged water de cit and is composed mostly of herbaceous plants, with sparsely distributed shrubs, and rare trees. According to the vegetation classi cation system proposed by Ribeiro and Walter (2008), there is a mosaic of phytophysiognomies composed of Campo Sujo (grassland), Campo Rupestre (rupestrian grassland) and Cerrado Rupestre (rupestrian Cerrado).
The second study area is located in the FD of Brazil, more speci cally, in the ferruginous detritus-lateritic deposit of Tertiary-Quaternary age, which covers a large portion of the study area (CPRM, 2010). This unit is surrounded by different types of Neoproterozoic lithologies from Paranoá, Canastra, and Bambuí Groups. We can notice a large extension of urban areas covering this study site (Fi. 1).

Soil sampling for chemical and agronomic characterization of the study sites
In the ultrama c (SAP and LAT) and typical Cerrado soils (CAM and LV) sites, three composed soil samples, each one was constituted by 10-30 subsamples, were collected at the 0-20 cm soil depth using a "dutch" auger. The samples were air-dried, ground and passed through a 2 mm sieve in the Soil Laboratory of the Embrapa Cerrados. The chemical and agronomic characterization of the soils were carried out through the following analyzes, according to the procedures described by Embrapa (1999): organic matter (O.M.); pH in water (1:2.5); exchangeable Ca 2+ , Mg 2+ , and Al 3+ extracted by KCl 1 N; P and K, extracted with Mehlich-1 solution (0.05 M HCl + 0.0125 M H 2 SO 4 ); potential acidity (H + Al), extracted with 0.01 M calcium acetate at pH 7. Co, Cu, Fe, Mn, Ni and Zn were extracted with DTPA solution (bioavailable contents), according to methodologies described by Baker and Amacher (1982), and with aqua regia (extractable content) (ISO 1995), analyzed by inductively coupled plasma optical emission spectroscopy (ICP-OES, Thermo Fisher Scienti c 7000).
Anglo American Mining Company provided information on the predominant mineralogy of these sites, such as the characteristics of the soil's parent material and total levels of Ni, MgO, Fe, SiO 2 and Al 2 O 3 were kindly provided by, with soil sampling density of 25 mx 25 m (data not shown).  (Table 1). In the botanic surveys mentioned above, at least one individual of those species collected in the area was found to hyperaccumulate Ni in above ground tissues. This also applies to specimens of E. salicoides sampled in ultrama c or typical Cerrado soils in relation to the hyperaccumulation of Al in the tissues.
Collection of plants for DNA, biochemical, and metal content analysis, and for localization of metals in leaf structures In each site described in Table 1, at different times and according to the objective of the study, the shoot (leaves and stems) of various plants of each species, randomly distributed on the ground, was collected. The collected materials were packed in paper or plastic bags, identi ed and taken to the Laboratory of Chemical Analysis of Plants, for elemental characterization, and to the Laboratory of Plant Genetics and Molecular Biology for DNA analysis, both from Embrapa Cerrados. E. salicoides plants were taken to the Laboratory of Plant Physiology of the Department of Botany, University of Brasília (UnB), for analysis of non-structural carbohydrates and total proteins. With the authorization issued by the Brazilian Institute of Environment and Renewable Natural Resources (IBAMA) (Special Authorization No. 02/2008), the plant tissue of E. salicoides (ES) species collected in the SAP5 site were prepared for identi cation, localization and semi-quanti cation of chemical elements at the tissue level, using scanning electron microscopy (SEM) and Energy-dispersion X-ray spectroscopy (EDXS), at the microscopy service (SCMEM) of the Faculty of Science and Technologies of Nancy (Université de Lorraine), Nancy, France. a) Sample processing for DNA extraction In the ultrama c (SAP5, SAP7, and SAP9) and the typical Cerrado (CAM) sites, shoots (leaves and stems) of two to three individuals were collected to form a composite sample of each species of E. salicoides (EU), O. hirsutissima (OH) and J. lanstyakii (JL), totaling 14 botanical accessions, as described in Table 2. The selected plants were in good vegetative condition and were growing more than two meters apart. The collected materials were packed in plastic bags, identi ed, placed in a polystyrene box with ice, until delivery, on the same day, to the Laboratory of Plant Genetics and Molecular Biology, where they were processed and aliquoted for DNA analysis and forwarding the rest of the plant material to the Laboratory of Plant Chemical Analysis.
Samples of genomic DNA from each access were extracted using the modi ed CTAB methodology (Faleiro et al. 2003) and with the use of washing in Sorbitol buffer (0.35 M). The amount of DNA was estimated by spectrophotometry at 260 nm (A260) and the A260/A280 ratio was used to evaluate the purity and quality of the samples (Sambroock et al. 1989). The DNA samples from each accession were diluted to 5 ng µL − 1 . The ampli cation reactions to obtain Inter Simple Sequence Repeats (ISSR) markers were carried out with 4.9 µL of Milli Q water 1.3 µL of buffer; 0.39 µL of 50 mM MgCl 2 ; 0.26 µL of deoxyribonucleotides (dATP, dTTP, dGTP and dCTP) 10 µM; 1.95 µL of a 2 µM primer (Operon Technologies Inc., Alameda, CA, USA); 0.2 µL of the Taq DNA polymerase enzyme (1 unit); and 3 µL of DNA (15 ng) (total volume of 13µL). Initially, 18 ISSR primers were tested (Table 3). From these tests, eight primers were selected and used to obtain ISSR markers that generated a greater number of polymorphic bands and presented better quality of ampli cations.
The ampli cations were performed in a thermocycler, in which the samples were initially denatured at 94°C for 2 min, followed by 37 cycles, starting with 15 seconds at 94°C; then 30 seconds at 47°C and 72°C per 1 min. At the end of all cycles, the process was completed in 7 min at 72°C and cooled to 4°C.
After ampli cation, 3 µl of a mixture of bromophenol blue (0.25 %) and glycerol (60 %) in water was added to each sample. These samples were applied on an agarose gel (1.2%), stained with ethidium bromide, submerged in a TBE buffer (90 mM Tris-Borate, 1 mM EDTA). The electrophoretic separation was approximately four hours, at 90 V. At the end of the race, the gels were photographed under ultraviolet light.

b) Preparation of samples for chemical analysis of leaf tissues
After removing the aliquot for DNA analysis, all remaining tissue material (leaves and stems) from each sample, together with samples of other specimens collected at the same sites, were sent to the Laboratory of Plant Chemistry Analysis for element characterization. The materials were slightly immersed in tap water and then in deionized water, to remove soil particles, and left on a bench until it dried at room temperature. The materials were then taken to a forced air oven at a temperature of 40°C, until reach constant weight. After drying, the leaf tissues were nely crushed in a knife mill and mineralized by moist digestion, in a mixture of perchloric acid and hydrogen peroxide, in the 2:1 proportion (v/v), respectively, to determine the levels of Ca, K, Mg, P, S, Co, Cu, Fe, Ni, Mn, and Zn by ICP-OES. The N was extracted by the micro-Khjeldahl method and determined by colorimetry (Embrapa 1999) or by ow injection analysis (FIA, Lachat Quikchem 6000 system), coupled with UV/VIS, method. In each of the sites mentioned above, in the morning and, in the peak of dry season (June to August), three plants growing apart on the ground were collected to determine the non-structural carbohydrates (soluble sugars, glucose, fructose, sucrose, ra nose, and starch) and the total proteins. Before being sent for destructive analysis, the numbers of in orescences/plant and branches/stem, as well as length of the stems and of the internodes for each plant were obtained using a pachymeter.

d) Determination of total soluble sugars (TSS)
Leaves samples of E. salicoides were lyophilized and ground (10 mg), for soluble sugars determination. The soluble sugars were extracted four times with 80 % ethanol (500 µL), at 80°C (40 min.). After centrifugation (10,000 g, 10 min), the supernatants were combined and depigmented by the modi ed Shannon method (Shanon 1968). In a separating funnel, the ethanolic fraction (2.0 mL), absolute ethanol (0.5 mL), chloroform (3.0 mL), and water (5.5 mL) were added in sequence. Separation occurred after a period of approximately 12 h. The measurement of TSS was performed according to the phenol-sulfuric method (Dubois et al. 1956). For comparison purposes, a standard glucose curve (SIGMA) was used at concentrations of 0, 5, 10, 20, 40, and 80 µg. Supernatants containing TSS were collected and reserved for further analysis. The precipitate was used to remove the other components.
For the analysis of TSS (sucrose, glucose, fructose, and ra nose), the alcoholic fractions were dried, resuspended in water (1 mL), passed through an anionic and cationic exchange column (Dowex), and analyzed by High Performance Ion Exchange Chromatography with Detector of Integrated Amperometric Pulse (HPAEC/IPAD), in CarboPac PA-10 column (Dionex Corporation, Sunnyvale, Ca, USA), using an elution gradient with 200 mM NaOH in water (30 min.). The detector responses were compared with the patterns of glucose, fructose, sucrose, and ra nose at 0.625, 1.25, 2.5, 5.0, 10.0, and 20.0 µM. The standard curve for each sugar was used to calculate the carbohydrate content in the leaves.

e) Starch determination
After removing the TSS, the starch in the precipitate was extracted twice with Bacillus licheniformis thermostable α-amylase (MEGAZYME − 120 U mL − 1 ) (0.5 mL) at 75°C, for 30 min, and twice with amyloglucosidase from the Aspergillus niger fungus (MEGAZYME − 15 U.mL − 1 ) (0.5 mL) at 50°C and for 30 min. At the end of the four incubations with the enzymatic extract (2.0 mL), 100 mL of 0.8 M perchloric acid was added to stop the reaction and to precipitate proteins. After centrifugation (13,000 g, 2 min), the glucose released (50 µL of extract) was dosed by the GODPOD enzymatic method, using 750 µL of the PAPLiquiform Glucose kit (CENTERLAB®), which contains glucose oxidase, peroxidase, 4-aminoantipyrine, and phenol. After incubation at 30°C and, for 15 min, the glucose content was determined in a spectrophotometer at 490 nm. The data obtained were compared with a standard glucose curve (SIGMA), at the concentrations of 0, 5, 10, 20, 30, and 40 µg. The released glucose was calculated and adjusted (-10 %) for the mass of bound glucose that is present in the starch.

f) Determination of total proteins
The proteins were extracted four times with sodium phosphate buffered saline (PBS, 100 mM pH 7.4) (0.5 mL) for 1 h, except the rst, which remained 12 h at 4°C. After centrifugation (13,000 g, 10 min.), the total proteins were dosed using the Bradford method (Bradford 1976). The data obtained were compared with a standard glucose curve (SIGMA), at the concentrations of 0, 5, 10, 15, 20, and 25 µg. The released glucose was calculated and adjusted (-10 %) for the mass of bound glucose that is present in the starch. g) Location of Ni in cell compartments of E. salicoides leaves, native to ultrama c soil, by SEM-EDXS About three E. solicoides (ES-P5) plants were randomly collected, in the SAP5 ultrama c site for use in microscopic analysis for the location, distribution and relative concentration of Ni in the cells of the leaf tissue. Still in the eld, the third pair of leaves that precedes the youngest lea et were cut into small segments, with a stainless-steel blade. The cuttings were made in order to have leaf blade tissue and secondary ribs in the same segment and placed in a small test tube (10 mL) containing 5 mL of 70 % alcohol. In the laboratory, about 10 segments of each sample were cut into thin pieces (1 mm x 3 mm), xed in adhesive tape, placed on carbon cassettes ("stub") and dried at room temperature. In these samples, the location of Ni accumulation in tissues was analyzed by using SEM-EDXS (Hitachi-S4800 SEM), with accelerated power ranging from 5 to 15 KV.

Statistical analysis
Due to the fact that the assumption of normality was not met, the data of metal content in soil and in leaf tissues were processed using non-parametric statistical analysis. The comparison between the medians of treatment was made based on the Kruskal-Wallis test at the signi cance level of p ≤ 0.05. The principal component analysis (PCA) was used for the correlation of the sites according to the soil chemical characteristics and nutrient accumulation in leaf tissues. Understanding communality as the explanatory power of the variable by the factor, the value adopted was > 0.70. In order to evaluate how the soils are grouped according to all chemical attributes together, a cluster analysis was carried out by Ward's method, considering the Euclidean distance (Bussab et al. 1990). Optimum number of clusters was veri ed by the Average silhouette method (Rousseeuw 1987).
In the DNA analysis, the generated ISSR markers were converted into a matrix of binary data, from which the genetic dissimilarity between the different genotypes was estimated, based on the complement of the Nei and Li similarity coe cient (Nei and Li 1979), using the Genes Program (Cruz 2013). The genetic similarity (GS) was given by: Sgij = 2Nij / (Ni + Nj), where: Nij is the number of bands present in both i and j genotypes; and Ni and Nj are the number of bands present in i and j genotypes, respectively. By subtracting the unit's SG value by 1 (1 -SG), we obtained the genetic dissimilarity. The genetic dissimilarity matrix was used to perform cluster analysis, by the Unweighted pair group method arithmetic average (UPGMA) method (Sneath and Sokal 1973) as a clustering criterion, and the graphic dispersion based on multidimensional scales, using the main coordinate method.
To test the effects of treatments (type of soils) on the production of total sugars, proteins and starch by E. salicoides plants growing on different soils, the data were analyzed using a one-way ANOVA, followed by Tukey's test, when signi cance level was < 0.05.

Soil fertility and geochemistry
The agronomic characteristics and bioavailability of metals in the soils from the collection sites of the botanical accesses are described in Table 4. The ultrama c soils in the region of Barro Alto (SAP5, LAT6, SAP7, and SAP9), in addition to the greater bioavailability of Co, Cu, Fe, Mn, Ni, and Zn, presented medium to high levels of cation exchange capacity (CEC) at pH 7 (T), pH in water higher than 6 or even close to neutrality (i. e. SAP9), and higher base saturation (V > 50%) than the Cerrado Soil (CAM) (p < 0.05). The average Mg contribution to the T values (Mg/T, %) observed in SAP soils was higher than 55%, in contrast to less than 4% of Mg saturation in the CAM soil.
The Cerrado soil (CAM) also showed a signi cantly higher exchangeable Al 3+ and its Mg:Ca ratio was 10-34 times lower than the values of the four ultrama c soils with signi cant differences.
Results correspond to the average obtained from three composite samples per site, 0-20 cm depth: one composite sample was constituted by 10-30 subsamples; sd = standard deviation. OM = organic matter; H + Al = titratable acidity; Al, Ca, Mg extractable by 1N KCl; K and P extractable with Mehlich-1 solution; Co, Cu, Fe, Ni, Mn, and Zn extractable with DTPA solution; SB = sum of bases; CEC = Cation exchange capacity at soil pH; T = CEC at pH 7; V = base saturation; m = Al 3+ saturation; means followed by the same letters within the same line are not statistically different, according to the Kruskal-Wallis test (p ≥ 0.05).
In the PCA analysis involving agronomic characteristics without the bioavailability of heavy metals (Ni, Zn, Co, Mn, and Fe), there was a clear separation between the sites of ultrama c soils and typical Cerrado soils (Fig. 2). The rst principal component explained 52.5 % of the variance, associated with the pH, TpH7, BS, Mg, and Mg/Ca which were all statistically different between CAM and ultrama c soils. Therefore, this analysis clearly separated the typical Cerrado soil (CAM), which had negative values, from all the ultrama c soils which presented positive values in the PC1 graph. The soil samples from the FD (CAM) presented a major in uence of Al 3+ attribute. The second principal component explained 35.4 % of the associated variance with a major in uence of the K, P, H+Al, Ca, and OM attributes (communalities for r 2 ≥ 0.7). Despite the correlation of the ultrama c sites shown by the PCA analysis, two distinct subgroups were formed, with a clear separation of the SAP5 soils on the upper and right side of the graph, with higher levels of K, P, H+Al, Ca, and OM, and a group of the other three soils (LAT6, SAP7, and SAP9) on the bottom and right side, presenting signi cantly higher pH, Mg, Mg:Ca, and BS.
Soils of ultrama c origin in the world are easily recognized for having high levels of heavy metals such as Co, Cr, Cu, Fe, Mn, Ni and Zn. However, the typicalCerrado soils are well documented as having low levels of these elements which are, with the exception of Cr, essential for plant development (Correia et al, 2004). Considering only micronutrients in the PCA analysis, the rst principal component explained 69.1% of the variance, with the soil bioavailable content of Co, Ni and Zn as the correlation factors (communalities r 2 ≥ 0.70), where the SAP7 and SAP5 are the soils that presented the highest levels of those metals, respectively (Table 4; Fig. 3a). In contrast, Al 3+ and Fe 3+ were strongly related to CAM soil.
The second principal component explained 19.9% of the variance, indicating higher bioavailability of Mn and Cu in the SAP9 soil (Table 4; Fig. 3a). The LAT6 soil, despite being derived from ultrama c rocks rich in Ni and other micronutrients, has a weathering process that is different from the others, having undergone greater Ni and Mg leaching, and presenting a deeper soil pro le (> 1 m), similar to the CAM soil.
In the PCA analysis of metal contents (AR) in soil (Fig. 3b), the rst principal component explained 58.4 % of variance, which indicated separation in the right side, and in positive axis, SAP7 and SAP9, mainly by Cu, Cr, Fe, and Zn attributes. In negative position, the SAP5 samples were grouped by Co, Mg, Mn, and Ni contents in soils. The second principal component explained 29.7 % of variance and, on the left and positive side, Al was the attribute which gathered the soils samples of CAM. Again, the LAT6 soil was in an intermediary position between the CAM, also a soil with a deep diagnostic horizon, and SAP5, rich in Ni and other metals.
In summary, considering the chemical characteristics of the soils (fertility attributes, semi-total and bioavailable levels of metals), it is evident the formation of two distinct groups (Fig. 4): Group 1, represented by the typical Cerrado soil (CAM) and Group 2, represented by ultrama c soils, which in turn are formed by the Sub-group 2.1 (SAP5), Sub-group 2.2 (LAT6), Sub-group 2.3 (SAP7), and Sub-group 2.4 (SASP9). This formation of clusters con rms previous ndings. It clearly outlines another distinction between LAT6 on one side and SAP7 and SAP9 on the other side. In fact, the latter two show signi cantly higher T, CEC, Mg:Ca, and H+Al, Co, Ni, and Zn values in comparison with those from LAT6, which is more weathered both mineralogically and chemically (lixiviation of elements). The collection of soils under study therefore constitutes a wide range of ultrama c soils with highly varied physical and chemical characteristics that are probably derived from local edaphic conditions. However, all ultrama c soils are clearly distinguishable from typical Cerrado.

Elements characterization of botanical accesses
The plants require many elements (or their ions) to maintain their normal metabolism and growth and they are taken primarily from the soil. This requirement is largely different among plant species, often between plants of the same species. The nutrient concentration in plant shoots varied widely among E. salicoides, J. lanstyakii and O. hirsutissima species and between analyzed samples of the same species growing in the same site, re ecting the spatial variation of the soil (data not shown). The average values of the accumulated elements in the shoot of the plants growing in SAP soils (independent of the site) and typical Cerrado soil (CAM) are shown in Table 5.
Regarding the accumulated elements in the shoots, the rst and second principal components explained only 28.5% and 17.4% of the variance of the data, respectively (Fig. 5). This result indicates that the weight of the variables (concentration of nutrients in the tissues) did not oscillate signi cantly within the species growing in different soils sites, suggesting that the nutritional requirements are characteristic of the species, regardless of the place of growth. However, it re ects the availability of some elements in the soil, which in turn is regulated by geochemical and biological factors such as the content in the parent material, degree of weathering, pH, redox potential, and microbiological activities. These factors will determine the excess and / or de ciency of the element in the root environment. This was the case of Mn and Cu in greater concentration in the tissues of plants growing in the CAM soil, although the bioavailability of these elements, indicated by the DTPA extractor, is much lower than those observed in the SAP soils (Table 4). On the other hand, higher availability of Mg, Co, Ni and, to a certain extent, Zn in ultrama c soils is re ected in higher absorption and accumulation of these nutrients in plant tissues. The Mg is also the dominant cation in the exchangeable sites (T value) of SAP soils (Mg:T > 55 %), resulting in higher Mg:Ca ratios in plant tissues.

Plant diversity
The analysis of the 14 accessions studied using eight primers generated 108 ISSR markers, with an average of 13.5 markers per primer. The high percentage of polymorphic markers and the high average of markers per primer demonstrated the high genetic variability between the accessions of the three species.
This result can be explained by the number of accessions used and because they are from different species, and by the e ciency of the ISSR technique in the quanti cation of genetic diversity between the accessions.
In addition to the divergence between the accessions, the cluster analysis showed the formation of ve similarity groups, considering the genetic distance 0.6 as the cutoff point: Group 1, formed by the ES-P5, ES-P7, ES-P9, and ES-CAM1 accessions (numbers in the graph: 1, 2, 3, and 12, respectively); Group 2, formed by the OH-P5, OH-P7, and OH-P9 accessions (numbers in the graph: 4, 5 and 6, respectively); To understand whether the basal metabolism of plants is altered in the presence of greater availability of heavy metals (Ni and Al) in the soil, analyses of the content of total soluble sugars (TSS), starch and total proteins were carried out in plants growing in ultrama c soils (SAP5 and LAT6) and typical Cerrado´s soil (LV). The results shown in Fig. 9 indicates variations in the TSS levels in plants growing in these three sites. The highest levels of TSS were found in the specimens collected in the SAP5 soil, when compared with the performance of plants in other sites (Fig. 9a). The observed differences suggest a possible effect of the increased availability of Ni in the SAP5 soil and its in uence on photosynthesis. In addition to the greater production of glucose and fructose (reducing sugars), there was an accumulation of sucrose and ra nose (non-reducing sugars) in plants growing in the same environment. In fact, ra nose was not detected in plants growing in LAT6 or in LV sites (Fig. 9c), soils with lower levels of Ni and with a soil deeper than SAP5 (Cambisol), whose greater restriction on root growth exposes plants to water stress during periods of drought in the region.

Ni localization in the tissues of Euploca salicoides
Metal-hyperaccumators plant species possess mechanisms to maintain adequate concentrations of metal ions (essential or not) in different cellular compartments to minimize the damage in important cell organelles Clemens 2001). In our study, using the SEM microscope coupled to an EDXS probe, we searched for Ni in various cell compartments where other metals were reported to accumulate in the leaves (e.g. epidermal cells, guard cells). We observed that the surface architecture on both sides of E. salicoides leaves was characterized by a dense pubescence formed by non-glandular trichomes. The trichome surface was microsculptured with numerous prominent protrusions. Ni allocation was observed only in the protrusions in trichomes of E. salicoides leaves collected in an ultrama c site (Fig. 10). Al 3+ occupying most of the exchange sites in the CEC, which is nevertheless very low due to the lack of OM and negatively-charged minerals (Table 4).

Discussion
Among the four ultrama c soils which are clearly separated from the acidic Ferralsol, there is again a gradient of available Ni and Mg which causes a gradient in the fertility stress to plants. The cluster analysis clearly shows this gradient by separating the soils into three groups. The Ferralsol (LAT6) which corresponds to the less stressful environment, separated from the three Cambisol (SAP5, SAP7 and SAP9). Among the three Cambisol SAP7 and SAP9 were the most hostile for plant growth as the least developed (weathered soils) whereas SAP5 is signi cantly less toxic in terms of available Ni and excess of Mg. For different reasons, both the ultrama c and typical Cerrado soils have very low levels of available Ca, K, and P so they both cause a similar nutritional stress for plant species although the strong acidity of CAM soil contrasts with moderately acid to neutral ultrama c soils.
The phytophysiognomies that occur in the ultrama c zone of Barro Alto are structurally similar to those that occur in the typical Cerrado biome. In the case of Campo Sujo and Campo Rupestre, the dominant component is the herbaceous-sub-shrub stratum growing on shallow soils, and eventually in the presence of rocky outcrops. These characteristics offer some physical restriction to root growth and, consequently, subject to water stress in the driest periods of the year. What differentiates ultrama c environments from other areas of the Cerrado biome, in a very striking way, are the chemical characteristics and, possibly, the microbiota associated with the local ora. Pessoa-Filho et al. (2015) found differences in the bacterial community structures between the non-ultrama c soil and the two ultrama c soils located in the same sites of the present study. They found that the abundance of bacterial lineages between Sites 1 (SAP5) and 2 (LAT6) was different. These authors hypothesized that these phenomena could be attributed to the establishment of different plant communities growing on them, as indicated in the ora survey done in the same area by Aquino et al. (2011), rather than to the differences in the Ni availability in the soil. Indeed, rhizosphere communities are critical to the normal ecophysiological performance of a plant and are an important component of ultrama c environments due to their role in nutrient cycling.
According to Kabata-Pendias and Mukherjee (2007), the levels of Ni and Co in leaf tissues of the plants growing on ultrama c and non-ultrama c soils, as observed in this work, are considered toxic for several species, but they are the levels expected for native species in serpentine soils. In this environment, the three species are considered very tolerant to Ni, including two Ni hyperaccumulators (> 1,000 mg Ni kg − 1 in dry matter), but do not meet the criterion of Co hyperaccumulators, which was recently de ned as 300 salicoides species also showed the capacity for hyperaccumulation of Al (> 1,000 mg kg − 1 in dry matter) in the tissues of some specimens collected on typical Cerrado soils (data not shown). Therefore, this species is considered a facultative metal hyperaccumulator, that is, the capacity of hyperaccumulation of either Ni or Al seems to depend on the availability of the respective element in the soil in which the plant is growing.
Syndrome of ultrama c soils (i.e.'Serpentine syndrome') is related to the combined physical, chemical, and biological factors associated with serpentine soils (Kruckeberg, 2002). It manifests in the competition between Mg, with high availability in the environment, and Ca for exchange sites on the root surfaces, resulting in Mg:Ca ratio > 1 and Ca de ciency. In typical Cerrado soils, as a result of lower Mg availability in the soil, the relationship between these nutrients in plant tissue is, in general, reversed. In our study, E. salicoides plants growing in ultrama c soils absorbed higher levels of Ca and Mg than plants in the Cerrado, however, regardless of the environment, in a greater proportion of Ca, resulting in a similar Mg:Ca ratio (Mg:Ca < 1). There was no effect in the Mg:Ca soil ratio so that the plants had a similar Mg:Ca balance among the three ultrama c Cambisols. One could suspect a genetic adaptation that helps the plant cope with a gradient of intensity of such toxic nutrient status.
Surprisingly, there seems to be no difference in Ni uptake among the three ultrama c Cambisols for any of the three plant species although they display clear differences in Ni availability. One reason could be that the level of available Ni is high in the three soils and none of the three is therefore limiting for Ni accumulation, although O. hirsutissima is not characterized as Ni-hyperaccumulator. Calcium has an important role in the alleviation of growth inhibition by salts and metal stress. For instance, the supply of Ca to the growth medium of rice plants was reported to reduce the Ni translocation factor to the shoots where it could cause physiological disturbances (Aziz et al. 2014). In the ultrama c environment, the Ca:Ni ratio in E. salicoides plants was 8.5-fold higher than that observed in the J. lanstyakii species and 17-fold higher than that of O. hirsutissima. It is possible that the preference for Ca absorption is one of the physiological mechanisms developed by this species to avoid possible toxic effects of Mg and Ni on their above-ground tissues.
The soils re ect quality of parent material, both physical properties and chemical content, which in turn affects plant's life. The results show that the accessions collected in ultrama c sites have a genetic differential, which may be associated with greater adaptation to soils rich in Ni, Cr, Co, Zn, Mn, Fe, and Mg. In fact, the PCA analysis regarding the accumulation of nutrients in the shoots (Fig. 5)  The genetic studies of the populations of three sites in the ultrama c area of Barro Alto showed that the accessions collected were closer with each other, thus indicating genetic similarity among ultrama c populations but major differences with the accessions collected on Cerrado environment. Similar results were obtained by Makela et al. (2016), who, in order to determine the level of genetic variation in populations of Quercus rubra from mining-damaged ecosystems, used the Random Ampli ed Polymorphism DNA (RAPD) marker system to assess the level of gene expression of candidate genes for Ni resistance. These authors reported that the genetic diversity in populations of Q. rubra adapted to metalliferous soils were genetically close, possibly due to the ability of pollen to disperse over long distances, which would favor gene ow. Pereira et al. (2004), also using the RAPD molecular marker technique, evaluated the effects of adapting plants to a stress condition, resulting from the presence of toxic metals in serpentine soils (originated from ultrama c rocks), on the genetic structure of Cerrado native populations of that environment and non-serpentine soils, growing well separated with each other. These authors did not observe signi cant divergences between the two native groups, also attributed to the occurrence of gene ow between each group. However, they detected a tendency for divergence between populations, suggesting that when colonizing the ultrama c environment, the two populations followed different evolutionary path.
Habitat isolation is the reduction in gene ow between populations caused by spatial separation of the habitats to which they are differently adapted (Mayr 1947). In the present study, the geographical distance between the places of collection of the accessions of E. salicoides was signi cant, around 250 km apart, leading to a low probability of gene ow between the native groups in the ultrama c zone of Barro Alto and that of non-ultrama c region, in Brasília, FD. Although there is genetic diversity between them, it is not possible to guarantee whether the groups diverged to the point of being considered different species or whether they should be considered separated ecotypes that developed speci c adaptations to different habitats, with very distinct edaphic characteristics, in which they can explore, in order to survive and reproduce.
Plants growing in ultrama c soils produced more sugars and probably sustained the greatest growth in terms of branches and internodes. Accumulation of soluble sugars has also been found in plants exposed to stress by As and Zn (Ghnaya et al. 2010;Jha and Dubey 2004). The increase in sucrose and ra nose suggests a response to a mechanism of detoxi cation of metals, leading to a decrease in the levels of reducing sugars, which are potentially harmful to cells (Costa and Spitz 1997; Kim et al. 2003).
The production of ra nose in plants is associated with an adaptive response to environmental stress, such as water stress and metal toxicity. In fact, ra nose was detected only in plants growing in SAP5, which is a Cambisol type (soil depth < 0.50 m), where the plants are more susceptible to water stress than those growing in LAT6 and LV Ferrasol (both with soil depth > 1 m) during the dry season that occurs in the region.
Protein content is related to stress (as drought), but the effect of stress depends on the development stage of the plant, species and genotypes, and the intensity and duration of stress (Lazcano and Lowatt 1999). In our study, the total protein content did not differ between the plants growing at the three sites (Fig. 9d). However, E. salicoides presented both morphological and molecular changes, indicating that the lack of differential abundance of proteins in the plant's proteome may be indirectly associated with these phenomena. identi cation process, none of the identi ed species has been con rmed as endemic to the studied environments. They can be found in other areas of Campo Sujo and Campo Rupestre, where plants do not hyperaccumulate metals due to the low availability in soils. Pollard et al. (2014) postulated that, in such species, the hyperaccumulation is an ancestral phylogenetic trait or an anomalous manifestation of physiological mechanisms evolved on "normal" soils, and may or may not have direct adaptive bene ts.
However, we noticed that some species growing in similar phytophysiognomies are unusual in both environments (observation data), indicating ora with distinctive species composition. Goolsby and Mason (2016) suggested that hyperaccumulation evolved as a by-product of selection for physiological processes unrelated to adaptation on metalliferous soils. On the other hand, van der Ent et al. (2015b) reported that extreme natural phenotypes such as hyperaccumulators evolved through positive, directional natural selection, rather than neutral genetic drift.
According to the results obtained in our study, in typical Cerrado soils, the plants of the three species accumulated, on average, less than 30 mg kg − 1 of Ni in their shoots tissues, suggesting that the critical level of Ni for these species is closer to that of normal plants, as stated by Chang et al. (2002). In the ultrama c environment, re ecting the metal's bioavailability, the three species accumulated more than 600 mg kg − 1 of Ni, considered excessive for most species. The phenomenon of hyperaccumulation was observed in J. lanstyakii and E. salicoides, with no irreparable damage occurring at metabolic, physiological and / or morphological levels, as shown in studies with E. salicoides.    Cluster analysis of the sites according to the dissimilarities between the attributes of fertility, metals bioavailability and semi-total metals content in the soils where the botanical access collections were made.   Carbohydrate contents in E. salicoides plants growing in three sites (LV, LAT6, and SAP5), with different edaphic characteristics. a) total soluble sugars (TSS); b) starch; c) content of soluble sugars, as monosaccharides (glucose and fructose) and oligosaccharides (sucrose and ra nose); and d) total soluble proteins. Bars represent standard deviation of the mean. Means followed by same letters are not statistically different (p > 0.05).

Figure 10
SEM images of a) central rib; b) trichomes of E. salicoides in leaf blade, cultivated in soil with high Ni bioavailability (SAP5) and X-ray emission spectra intensities of elements: c) "dust" trapped between trichomes; and d) trichomes -Ni peak is shown only in protrusions of trichomes (kv = 10)

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Tables.docx