Proteomic Profiling and Rhizosphere-Associated Microbial Communities Reveal Adaptive Mechanisms of Dioclea apurensis Kunth in Eastern Amazon’s Rehabilitating Minelands

Dioclea apurensis Kunth is native to ferruginous rocky outcrops (known as canga) in the eastern Amazon. Native cangas are considered hotspots of biological diversity and have one of the largest iron ore deposits in the world. There, D. apurensis can grow in post-mining areas where molecular mechanisms and rhizospheric interactions with soil microorganisms are expected to contribute to their establishment in rehabilitating minelands (RM). In this study, we compare the root proteomic profile and rhizosphere-associated bacterial and fungal communities of D. apurensis growing in canga and RM to characterize the main mechanisms that allow the growth and establishment in post-mining areas. The results showed that proteins involved in response to oxidative stress, drought, excess of iron, and phosphorus deficiency showed higher levels in canga and, therefore, helped explain its high establishment rates in RM. Rhizospheric selectivity of microorganisms was more evident in canga. The microbial community structure was mostly different between the two habitats, denoting that despite having its preferences, D. apurensis can associate with beneficial soil microorganisms without specificity. Therefore, its good performance in RM can also be improved or attributed to its ability to cope with beneficial soil-borne microorganisms. Native plants with such adaptations must be used to enhance the rehabilitation process.


Introduction
Iron ore extraction in the eastern Amazon occurs mainly in ferruginous mountain outcrops surrounded by rainforests in the Carajás mineral province, covering some of the largest iron reservoirs of the world [1]. This ecosystem is characterized by a high plant diversity that forms different phytophysiognomies known as canga [2]. The plant communities that grow in canga are subjected to adverse environmental conditions such as intense ultraviolet (UV) radiation, high temperatures (soil and air), which, combined with constant winds, increase evapotranspiration processes [3]. At the same time, water acquisition is hampered by poor water-holding capacity influenced by soil composition and shallow soil formation, which tends to intensify the effects of water deficit [2,3]. Additionally, the oxidic soils from canga are characterized by low availability of soluble phosphorus (P) caused by high adsorption of P to iron and aluminum oxides [4]. Plants require several adaptations to grow and colonize such ecosystems. A robust adaptive

Physical and Chemical Properties of Canga and Mining Area Soil Substrates
The soil substrates of RM showed a greater proportion of clay, while canga soils showed higher proportions of sand ( Table 1). The soils from canga were more acidic than RM soils (Table 1). Canga soils showed more organic matter content than soils from RM (Table 1). Total nitrogen (N) was higher in canga soils, whereas available P was higher in RM soil substrates ( Table 1). The concentrations of metals differs between environments; copper and iron were higher in canga soils, whereas manganese and zinc were higher in RM soils (Table 1). Table 1. Physical and chemical characteristics of soils associated with Dioclea apurensis growing in canga and rehabilitating minelands (RM). Soil results are mean ± standard deviation for n = 5.

Protein Profiles, Annotation, and Functional Enrichment
A total of 1401 proteins were successfully identified and quantified, with 396 showing significant differential abundance (p < 0.05) between the roots from both environments (Dataset S1), being more abundant in roots from canga. Only two proteins were exclusively detected in the roots from RM, while 19 were identified only in the roots from canga (Dataset S1). Within the total proteins identified, 291 showed higher levels in roots from canga, while 105 were showed higher levels in roots from RM. The principal component analysis (PCA) of the 1401 proteins ( Figure 1a) and 396 filtered proteins (Figure 1b) showed the separation between root samples from canga and roots from RM.
The enrichment analysis showed that the most accumulated proteins in canga act in 44 main processes, while the most accumulated proteins in RM participate in at least 20 processes (Figure 1c). In the roots from both sampling sites, the most prominent categories included proteins involved in response to stimuli, mainly related to abiotic stresses, and proteins involved in growth and reproduction, including proteins involved in carbon metabolism and biosynthesis of secondary metabolites (Figure 1c). Proteins involved in these pathways, especially in secondary metabolite biosynthesis, including amino acid biosynthesis, and carbon metabolism, were more accumulated in canga. Figure S1a,b show the 40 main biological processes from the functional annotation of the most accumulated proteins in the roots from canga and RM, respectively.

Protein-Protein Interactions (PPI)
The analysis of PPI showed 178 proteins with higher levels in the roots from canga ( Figure 2, Dataset S2) and 69 proteins in the roots from RM ( Figure 2, Dataset S2). The enrichment and PPI analysis showed an essential co-occurrence of a large part of the proteins accumulated in plants of each environment. These proteins were grouped into three highly interacting protein clusters, with a PPI enrichment p-value of 1 × 10 −16 in canga plants and p-value < 1.4 × 10 −6 in RM plants ( Figure 2, Dataset S2).  In canga plants, the functional enrichment analysis showed that the cluster in red includes 63 proteins mainly involved in energy metabolism and secondary metabolite biosynthesis, including amino acid biosynthesis. The green cluster comprised 69 proteins related to the cell cycle and response, especially abiotic stresses such as osmotic stress. The blue cluster contains 46 proteins, mainly including binding proteins such as nucleic acid binding proteins, structural components of ribosomes, and protein folding. In RM plants, the red cluster was composed of 19 proteins related to abiotic stimuli, especially temperature. The cluster in green contains 24 proteins categorized by the functional enrichment analysis of the term "cellular anatomical entity", which includes proteins involved in carbon metabolism and biosynthesis of secondary metabolites. The 26 proteins included in the blue cluster of plants on RM were categorized mainly into cellular processes such as gene expression and protein metabolic process ( Figure 2, Dataset S2).

Microbial Diversity
The ITS2 sequencing produced 3,948,832 raw reads across 16 input libraries. After quality filtering, 2,760,666 amplicon sequences were selected. The number of fungal OTUs in each sample ranged from 359 to 789 and was on average 563 (Table S1). Canga soils showed more fungal sequences (1,404,745; comprising 521 ± 132 OTUs) than RM (1,355,921; comprising 605 ± 97 OTUs). Additionally, rhizosphere soil samples presented more sequences (1,531,425; comprising 634 ± 116 OTUs) than the bulk soil samples (1,229,241; comprising 492 ± 78 OTUs). The Shannon and Simpson diversity indexes of fungal sequences were higher in RM, with higher values in the bulk soil substrates of RM samples ( Figure 3). The 16S sequencing produced 4,208,259 raw reads across 16 input libraries. After quality filtering, 1,361,859 sequences were considered. The number of bacterial OTUs ranged from 205 to 996 and was on average 579 (Table S2). The soils from RM showed a higher number of sequences (769,122; comprising 723 ± 240 OTUs) than canga (592,737; comprising 435 ± 190 OTUs), whereas more bacterial sequences were detected in the bulk soils (750,740; comprising 601 ± 366) than in the rhizosphere soils (611,119; comprising 556 ± 76 OTUs). The Shannon and Simpson diversity indexes were higher in the RM samples, with higher values in the rhizospheric substrate samples ( Figure 3).
In canga, the analysis of fungal sequences showed Ascomycota and Basidiomycota as the most abundant phyla in the rhizosphere and bulk soils ( Figures S2 and S3, Table S3). Glomeromycota was also detected in bulk soil samples ( Figures S2 and S3, Table S3). Regarding bacterial sequences, Acidobacteria, Proteobacteria, and Actinobacteria were the three most abundant phyla in the rhizosphere and bulk soils ( Figures S2 and S3, Table S3).
In RM, the analysis of fungal sequences revealed Asc omycota, Basidiomycota, and Glomeromycota as the most dominant phyla in the rhizosphere and bulk soil substrate ( Figures S2 and S3, Table S3). Proteobacteria, Acidobacteria, and Actinobacteria were the more abundant phyla in both the rhizosphere and bulk soil substrates (Figures S2 and S3,  Table S3).
A heatmap based on OTU abundance indicates differences between the structure of the microbial communities associated with plants in both ecosystems (RM and native canga) (Figure 4a,c). PCoA and cluster analyses of microbial community structure among the different samples showed that microbial communities were distinct between RM and native canga. Microbial communities also differed when comparing rhizospheric and bulk soil from canga (considering both fungal and bacterial populations) but not in RM (Figure 4b,d). Differences in the composition of microbial communities were estimated by calculating LEfSe scores at the family and genus levels. A total of 25 distinct fungal taxa were identified in plants established in canga or RM ( Figure 5a). Most of these taxa were related to Ascomycota and Basidiomycota. Whereas a total of 29 preferential bacterial taxa were identified in plants growing in canga or RM, most of which were related to Proteobacteria and Actinobacteria (Figure 5b).
Seven trophic modes were predicted in the fungal OTUs, being the categories pathotroph_ symbiotroph most abundant in the soils from canga, whereas the categories pathotroph_ saprotroph_symbiotroph were most abundant in the soils from RM ( Figure 6). Similarly, 36 predicted functions were identified in the bacterial OTUs, being the function nitrogen fixation and nitrate reduction more abundant in both soils but higher in the RM ( Figure 6).

Discussion
In this study, the proteomic approach has revealed the differential abundances of proteins related to metabolic pathways involved in response to the abiotic stress conditions of both native canga and RM. Together, the functional annotation, enrichment, and PPI analyzes showed the co-occurrence of a large part of the essential proteins accumulated in plants growing in each ecosystem and include those related to responses to oxidative stress, drought, P deprivation, excess iron, and symbiosis. Thus, the readiness of a molecular machinery joined to specific rhizosphere-associated fungal and bacterial communities inhabiting the rhizosphere can be considered crucial for establishing this native plant in RM in the eastern Amazon.

Physical and Chemical Properties of Canga and Mining Area Soil Substrates
Alteration in the normal levels of reactive oxygen species (ROS) is one of the main metabolic responses observed in plants growing under stressful environmental conditions [19]. In order to mediate the balance of ROS, the removal system relies on the action of enzymatic and non-enzymatic antioxidants [20]. In this study, proteins involved in synthesizing enzymatic antioxidants such as superoxide dismutase, peroxidases, catalases, glutathione reductase, and monodehydroascorbate reductase showed more abundance in the roots from canga (Dataset S1). Additionally, the non-enzymatic antioxidant systems proline-rich receptor protein kinase and betaine aldehyde dehydrogenase (BADH) were more accumulated in the roots from canga (Dataset S1). Such results reveal the abundances of proteins involved in the antioxidant system, especially in the roots growing under the severe conditions of canga, which also are identified in plants growing in RM.

Proteins Involved in the Response to Water Deficit
Two types of aquaporins showed more abundance in plants from canga, including aquaporin PIP1-1 and water stress-induced tonoplast intrinsic protein (Dataset S1), which play a regulatory role in the cellular transport of water in plasma membranes and tonoplasts, respectively [21]. These proteins have been described as essential for improving plant growth, water deficit tolerance, and osmotic balance [22,23]. Roots from canga showed higher levels of proteins involved in abscisic acid (ABA) transport and signaling (i.e., phosphatase 2C, ABC transporter, ATP-binding protein, ABC transporter B, and ABC transporter I), as well as proteins involved in signaling by serine/threonine kinase (i.e., phospholipase D alpha 1, proline-rich receptor-like protein kinase and receptor-like protein kinase S) than roots from RM (Dataset S1). ABA has been classified as a versatile phytohormone involved in plant signaling in response to drought stress [24]. Under conditions of water deficit (a common characteristic of canga ecosystems), there is also a high levels of phosphatase 2C (PPC2)-type proteins, which acts on the dephosphorylation of SnRK2, maintaining optimal ABA levels during water deficit [25], which agree with the results obtained in the roots from plants growing in canga.
High levels of osmolytes such as glycine betaine (GB), which is synthesized in a pathway in which BADH acts as a critical enzyme, has been recently described in plants growing under abiotic stress, including the water deficit [26], contributing to a higher relative water content, the integrity of cell membranes, stabilization of proteins, and detoxification [27]. The relationship of high levels of BADH biosynthesis with the increase in GB levels and greater tolerance to water deficit has already been observed in plants such as Arabidopsis thaliana [27], Sesuvium portulacastrum [28], and Zea mays [29]. In addition to BADH, roots from canga showed higher levels of proline-rich receptor-like protein kinase and spermidine synthase (Dataset S1). These proteins have been considered essential components of the cell membrane, playing essential roles in signal transduction in response to drought and stomata activity to prevent water loss [30,31]. The greater abundance of these water deficit-responsive proteins in roots from canga indicates that D. apurensis have developed adaptive mechanisms to resist extensive drought periods in canga and help resist the drought events in RM.

Proteins Involved in the Response to Metal Stress
Metal stress directly influences the establishment and growth of plants in mining areas. In fact, soils from canga showed higher levels of Fe (Table 1). High levels of metals in the soil solution can activate distinct signaling pathways such as cadmium-dependent, mitogen-activated protein kinase, ROS, and phytohormones, enhancing the expression of transcription factors or stress-responsive genes [32]. Most of these proteins were more accumulated in roots from canga (Dataset S1). Protein kinases, calmodulins, and calciumdependent protein kinases have been described as receptors capable of activating signaling networks in the tolerance to the excess of metals [33,34].
Additionally, Glutathione-related proteins are directly involved in the balance of metals in the intracellular medium, a process that has been described in Cucumis sativus, Triticum aestivum, and Beta vulgaris [35,36]. These results also agree with Jiang et al. [37], who reported a direct role of heat shock proteins (HSP70) in the contribution of tolerance to metal stress in Rosa hybrida. Such proteins were more accumulated in roots from canga (Dataset S1). Similarly, proteins related to the biosynthesis of compounds commonly detected in plants growing under metal stress, such as phenylalanine ammonia-lyase and nicotianamine synthase, were detected in this study (Dataset S1) [38][39][40]. Additionally, ferritin was also more accumulated in roots from canga, and their role in cytoplasmatic sequestration of soluble iron [38,39] can contribute to the diminishing of the negative effects of the metal in D. apurensis.

Proteins Involved in the Response to P-Starvation
According to this study, P levels in RM were higher than in canga ( Table 1). The higher concentration of P in RM can be explained by hydroseeding containing NPK fertilization in these sites. Phosphorus deficiency in canga can be related to the formation of complexes between P and iron oxides, limiting plant uptake availability [41]. Roots from plants sampled in canga showed more proteins related to P depletion, such as monogalactosyldiacylglycerol synthase 2, phospholipase, alcohol dehydrogenase, extracellular purple acid phosphatases, and antioxidant system proteins such as catalases and monodehydroascorbate reductases [42][43][44]. During Pi deprivation, glycolipids are transferred to extraplastid membranes, where they replace degraded phospholipids to meet the need for Pi essential for various biological processes in the cell. This membrane remodeling occurs in a process dependent on the activation of monogalactosyldiacylglycerol synthase induced by a low Pi level. Studies have already observed this pattern in leaves and roots with different plant species under Pi depletion, including Sesamum indicum, Zea mays, and Arabidopsis thaliana [42,45,46].
Additionally, three phospholipase D alpha-1 proteins were also more accumulated in roots from canga (Dataset S1). Phospholipase induction has been observed in studies with plants during P starvation [43]. These proteins are involved in P storage and induction of root growth to tolerate P scarcity conditions [43], such as in canga. In this sense, D. apurensis from canga synthesize proteins that support the growth under low P levels or induce mineralization from the soil substrate in canga. This study showed that alcohol dehydrogenase (ADH) was more accumulated in roots from canga (Dataset S1). The increase in ADH levels is in line with the results observed in studies with different plant species growing under P shortage. The high levels of this protein is observed together with increases in glycolysis and fermentation intermediaries, possibly related to cell expansion in Pi-limitation [47,48]. In acidic and Fe-rich soil substrates, such as the canga ecosystems, P can form complexes with iron oxides in an inaccessible form for absorption by plant roots [44]. This study identified extracellular purple acid phosphatases in roots from canga (Dataset S1). These proteins belong to a group of hydrolases induced by P deficiency, acting in recycling P from esters and anhydrides [44]. Additionally, P scarcity in canga soil substrates can induce ROS formation, which was also observed in the roots from canga (Dataset S1) participating in the signaling of cellular responses and protein biosynthesis involved in adaptations to P starvation [49]. Under both conditions, the roots expressed transcription factors commonly reported under P shortage, including MYB, WRKY, and HLH (Dataset S1), which regulate the expression of target genes and define metabolic responses to P starvation [50,51]. The set of proteins accumulated in plants growing in canga should contribute to the adaptation of D. apurensis in this ecosystem, although there are still no impediments to its development in environments with optimal P levels as in RM.

Proteins Involved in Symbiosis
Under abiotic stress, plants depend on beneficial soil microorganisms to become established in harmful environments such as canga. In this study, the analyses have detected several proteins involved in the establishment of symbiosis with bacteria and fungi, including nodulins [52], monodehydroascorbate ascorbate reductase [53], and enzymes involved in the ascorbate-glutathione cycle [54]. These enzymes were detected in the roots from both ecosystems, with higher levels in the roots from canga (Dataset S1). The nodulation in D. apurensis has been observed in recent studies with depletion of nutrients such as N [7]. Considering the successful establishment of D. apurensis in canga, these species contribute to maintaining essential soil processes such as N fixation during mineland rehabilitation [7,55]. The higher levels of symbiosis-related proteins in the roots from canga than from RM is explained by the higher diversity of plant species found in canga ecosystems and their corresponding effect on soil microbial communities.

Rhizosphere-Associated Microbial Communities
The rhizosphere-associated fungal and bacterial community analysis showed that specific taxa are enriched in the rhizosphere of D. apurensis. Proteobacteria and Ascomycota were the most abundant bacterial and fungal phyla, respectively, in both rhizosphere and bulk soil substrates ( Figure 5). Recent studies have reported similar results in plants growing in ecosystems affected by mining activities, where the soil microbial communities and specific mechanisms of abiotic stress tolerance can be considered key to promoting the phytostabilization towards rehabilitation of ecosystems services [8,56]. Among the preferential microorganisms identified in this study, several beneficial saprophytic, freeliving, symbiotic, and endophytic taxa were directly identified in the rhizospheric soil, especially in plants from canga, where specific fungal and bacterial taxa belonging to plant growth-promoting microorganisms were detected (e.g., Paraconiothyrium, Rasamsonia, Scytalidium, Rhodoplanes, Bradyrhizobium, Rhizobium, Roseiarcus, and Actinotalea) (Table S3; Figure 5). Such results agree with recent studies analyzing the diversity of microbial communities associated with plants growing under stressful environmental conditions, where specific rhizosphere-associated microorganisms have been described as essential to promote plant establishment [57,58]. This study has also detected a higher microbial diversity in plants from RM (especially fungal taxa), which can be related to the presence of widespread soil microorganisms inhabiting the RM soil substrates, with competitionmediated co-existence, latent soil microbes, and low influence of plant species on the soil microbiota (compared to native canga). However, the predicted functions of microorganisms involved in N-fixation were more abundant in RM than canga (Figure 6), showing that low specificity for microbial taxa can also be a characteristic supporting the establishment of this species in RM.
The rhizosphere-associated fungal and bacterial communities play an essential role in establishing native species in RM [59,60]. Additionally, several beneficial bacterial and fungal genera with key roles in nutrient solubilization, plant growth promotion, and defense against phytopathogens were detected (e.g., Glomeromycetes, Sphingomonas, Actinotalea, Rhizobium, Rasamsonia, Paraconiothyrium) [61][62][63]. Despite the higher diversity of rhizosphere-associated microorganisms identified in plants growing in RM, they were mostly different from those detected in canga, denoting that D. apurensis can associate with beneficial soil microorganisms inhabiting soil substrates in RM without an apparent specificity, similar to what has been described in Mimosa acutistipula [8]. Accordingly, the specific rhizosphere-associated microbial communities of D. apurensis, which is related to better performance under abiotic stress conditions in canga [64], must be considered in mineland rehabilitation projects to preserve essential microbe-mediated processes that can contribute to later successional stages of rehabilitation in Amazonian canga.

Soil Substrate Sampling and Chemical Analyses
Samples were collected at the end of the wet season (May 2018) in a native metalliferous savanna in Serra dos Carajás, Pará, northern Brazil. The sampling sites included: (i) a native herbaceous shrub canga ecosystem with minimal anthropogenic intervention (6 • 00 41.0 S 50 • 17 45.0 W); and (ii) a RM in waste piles of iron mining (6 • 02 32.0 S 50 • 07 04.0 W), where a revegetation program started in 2014. The application of D. apurensis seeds was carried out by hydroseeding containing NPK fertilization with 04-14-08 (8 kg for each 8 × 12 m of soil substrate).
The region's climate is tropical warm with a rainy season from November to March and a dry season from May to September, an average annual rainfall of 2033 mm, and temperatures varying between 25.1 • C and 26.3 • C [9,65]. Bulk soil samples (n = 5) were collected near D. apurensis roots naturally growing in canga and from D. apurensis growing in an RM project, at a depth of 10 cm, 5 m away from each plant considering zones without other plants. Rhizospheric soil (n = 5) was taken by gently shaking the soil substrate adhered to the root system from plants growing in canga and RM. The samples of bulk soil were submitted to chemical and physical analyses. After being air-dried, the samples were sieved using a 2 mm mesh. The pH was determined in a 1:2.5 soil substrate to water ratio, and the organic carbon content was determined using the potassium dichromate (K 2 Cr 2 O 7 ) method. The available P, K, B, Zn, Fe, Mn, and Cu were determined using the Mehlich-1 method (0.05 mol L −1 HCl + 0.0125 mol L −1 H 2 SO 4 ), S-SO 4 −2 by calcium phosphate monobasic at 0.01 M, and the total N content was determined using the Kjeldahl method [66]. The soil texture was determined as described by Kettler et al. [67].

Root Sampling and Protein Isolation
Roots from five D. apurensis individual plants were collected at each sampling site, kept in a cold phenol/SDS (sodium dodecyl sulfate) buffer, and transported to the laboratory for further processing. Protein isolation, determination of protein concentrations, and further processing were performed according to the protocol proposed by do Nascimento et al. [68], with minor modifications (Table S4).

Proteome Analysis
The identification and quantification of proteins were performed in a nanoACQUITY UPLC ® ultra-performance liquid chromatography (Milford, MA, USA), configured for fractionation in two dimensions as reported in Herrera et al. [69]. Five micrograms of the peptides were analyzed with five analytical replicates. The first dimension used a 5 µm XBridge BEH130 C18 (300 µm × 50 mm) and a Symmetry C18 5 µm (180 µm × 20 mm) trapping column at a flow rate of 2000 µL min −1 . The second dimension used a 1.7 µm BEH130 C18 1.8 µm (100 µm × 100 mm) analytical column, at a flow rate of 400 µL min −1 . The samples were separated in five fractions with a gradient of 10.8, 14.0 16.7, 20.4, and 65.0% acetonitrile. The chromatograph was coupled to a NanoLock ESI-Q-ToF SYNAPT G2-S (Waters) mass spectrometer. The acquisition ranged from 50 to 2000 Da, in MS E mode (data independent analysis) at a scan rate of 0.5 s and an interscan delay of 0.1 s.
The data were processed using the Progenesis QI software (Waters) for identification and quantification, using the Viridiplantae database from UniProt (UniProtKB/swissprot, uniprot.org, accessed on 10 October 2021). Protein identification was accepted if the probability of identifying peptides was greater than 90% and proteins with 95%. The significance levels of the differential abundances of proteins were determined by applying the ANOVA test (p < 0.05). To compare the proteome of D. apurensis grown in canga or RM, a PCA of the proteins with differential abundances and with p < 0.05 were produced in the R software v3.6.3 (R Core Team 2018; https://www.R-project.org, accessed on 10 October 2021), using packages FactoMineR, Factoshiny, and Factoextra. The functional annotation of proteins was performed using the OmicsBox v1.2.4 (bioBam) and Uniprot (UniProtKB/swiss-prot, uniprot.org, accessed on 10 October 2021). Kyoto Encyclopedia of Gene and Genome (KEGG) pathway enrichment analysis for proteins was performed using the KOBAS software v3.0. [70] (kobas.cbi.pku.edu.cn, accessed on 10 October 2021) with Arabidopsis thaliana as the background species for the analysis. The R package ggplot2 was used for the enriched KEGG pathways visualization. The PPI networks were predicted based on functional analysis using the software STRING (research tool for recovering genes/proteins in interaction) v11.0 (http://string-db.org/, accessed on 10 October 2021), using homologous proteins from the Arabidopsis thaliana as the background species.

Microbial Diversity
Total DNA was extracted from 0.25 g of soil using the PowerSoil DNA Isolation Kit (QIAGEN, Hilden, Germany), according to the manufacturer's recommendations. The DNA concentration was determined with the Qubit fluorometer (Thermo Fisher Scientific, Waltham, MA, USA), and the quality was verified in a 1 % electrophoresis agarose gel.
The amplicon libraries for bacteria and fungi were prepared according to Costa et al. [8], with minor modifications. The 16S rRNA gene was amplified by PCR using the bacterial primer set S-D-Bact-0341-b-S-17-N (5 -TCGTCGGCAGCGTCAGATGTGTATAAGAGAC AGCCTACGGGNGGCWGCAG-3 ) and S-D-Bact-0785-a-A-21-N (5 -GTCTCGTGGGCTCG GAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC-3 ). After a hot start at 95 • C for 3 min, 35 PCR amplification cycles at 95 • C for 30 seg, 55 • C for 30 seg, and 72 • C for 30 seg were performed, followed by a final extension step at 72 • C for 5 min. The ITS region of the 18S rRNA gene was amplified by PCR using the primer set fITS7i (5 -TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGGTGARTCATCGAATCTTTG-3 ) and ITS4i (5 -GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGTCCTCCGCTTATTG ATATGC-3 ). After a hot start at 94 • C for 2 min, 35 PCR amplification cycles at 94 • C for 30 seg, 56 • C for 1 min, and 72 • C for 30 seg were performed, followed by a final extension step at 72 • C for 7 min.
The size and quality of the PCR fragments were estimated on an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) using a DNA 1000 chip. The libraries were purified with the AMPure XP purification kit (Beckman Coulter, Brea, CA, USA) and further processed with the Nextera XT kit (Illumina, San Diego, CA, USA). The gene libraries were sequenced in a Miseq-Illumina platform using a MiSeq V3 reagent kit (600 cycles; Illumina) in the human and medical genetics laboratory at Universidade Federal do Pará (Belém, PA, Brazil).
The ITS and 16S sequences were analyzed using the Pipeline for MetaBarcoding Analysis (PIMBA), which allows the analysis of metabarcodes based on the pipeline QIIME [71]. The low-quality sequences were filtered and trimmed using PRINSEQ v0.20.4, and forward and reverse sequences were merged using PEAR v0.9.19 [72]. Reads were dereplicated, singletons removed, and the sequences were truncated to 200 for fungi and 240 for bacteria. Chimeras were filtered, and the sequences were grouped into operative taxonomic units (OTUs) using VSEARCH v2.8.2. The taxonomic assignment was developed using the UNITE database for fungi and the Ribosomal Database Project for bacteria [73,74]. Graphs were constructed considering the alpha and beta diversity in R software using the ggplot2 and vegan packages. Beta diversity was calculated, and principal coordinate analysis (PCoA) graphs were constructed using the "weighted UniFrac distances" in R software using the phyloseq package. Heatmaps were constructed with the total abundance of OTUs using R software (packages pheatmap and phyloseq). Alpha diversity was estimated using the vegan package with the Shannon and Simpson diversity indices. Additionally, clustering analyses of the data were performed using an "hclust" function with the options "methods = ward.D2" and "method.dist = correlation" from the pvclust package in the R software. Permutational multivariate analysis of variance was applied using the function "adonis" (vegan package). Linear discriminant analysis (LDA) of effect size (LEfSe) was performed with the Kruskal-Wallis test, and the effect size was estimated with a logarithmic score of 2.0 in the LDA. Finally, the predicted ecological roles of the identified microbial taxa were assigned using FUNGuild v1.1 [75] and FAPROTAX v1.2.4 [76] in Python v3.8.2. The data were plotted in the R software using the viridis, dplyr, and scales packages.

Conclusions
This study showed that D. apurensis growing in native canga have a set of proteins involved in the response to environmental stress to cope with the abiotic stress challenges. Its ability to increase the levels of a wide range of proteins in response to the challenging conditions of post-mining areas enhances D. apurensis establishment in rehabilitating minelands. Among them, the identification of proteins involved in the antioxidant system, response to water deficit, excess of metals, and deficiency of P. Additionally, our results confirm that D. apurensis establish interactions with beneficial microbial taxa without specificity. High levels of specific proteins involved in response to severe environmental conditions and interaction with key microbes at the rhizosphere are characteristics that can be identified in native species to select and diversify the plant species used for mineland rehabilitation.  Table S1: Fungal 18S rRNA sequences obtained in rhizospheric and bulk substrates samples from Dioclea apurensis growing in canga (canga) or rehabilitating minelands (RM); Table S2: Bacterial 16S rRNA sequences obtained in rhizospheric and bulk substrates samples from Dioclea apurensis growing in canga (canga) or rehabilitating minelands (RM); Table S3: Most abundant fungal and bacterial taxa identified in Dioclea apurensis soil substrates in plants from canga (canga) and rehabilitating minelands (RM). Numbers between parentheses are the relative abundance percentage (RA %) of each identified taxa; Table S4: Steps and procedures of protein extraction; Dataset S1: Protein report of 396 proteins with differential abundances in canga and RM plants filtered based on p < 0.05 and fold change ≥ 1.5; Dataset S2: PPI enrichment of Dioclea apurensis root proteins in STRING.

Data Availability Statement:
The sequences obtained in this study were deposited in the NCBI Sequence Read Archive (https://www.ncbi.nlm.nih.gov/sra/PRJNA690164, accessed on 1 June 2021) under the accession number PRJNA690164. The proteomic data was submitted to Massive repository under the accession MSV000087423 (https://massive.ucsd.edu/, accessed on 1 June 2021).

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.