Soil samples were collected from a VTM tailings area located in Panzhihua, Sichuan Province, China (101°58′10.89″E, 26°36′59.47 N) for compositional analysis and a P. pinnata pot experiment. The seeds of P. pinnata were collected in a mangrove forest in Wenchang, Hainan Province, China (110°47′E, 19°37’N). The mature seeds of P. pinnata were taken to laboratory and planted in pots containing VTM tailings. The pots were kept in a greenhouse with a day temperature of 25°C for 16 h and a night temperature of 17°C for 8 h. The potted trees were irrigated using tap water when needed. Three months later, when there were some big and pink nodules on the roots of P. pinnata, the plants were uprooted.

Total nitrogen (N) and available N of the soil samples were determined using the Kjeldahl method and alkali N-proliferation method, respectively (Wang et al., 2016; Spargo and Alley, 2018). Soil pH, organic matter, available phosphorus (P) and potassium (K), were determined using the ASI method. Tailings samples were digested with a mixture of HNO₃:HF:HCl (3:1:1 by vol) for the measurement of total metals, and the available metals in tailings were extracted by using 1 M C₂H₇NO₂ and 0.2 M ethylenediaminetetraacetic acid (EDTA; Saadani et al., 2016). A soil sample known available metal content was designed in all the steps of available metal extraction and measurement process as quality control. Then, the concentration of total metals and available metals of iron (Fe), titanium (Ti), vanadium (V), chromium (Cr), manganese (Mn), zinc (Zn), copper (Cu), nickel (Ni), lead (Pb), and cadmium (Cd) were quantitated using inductively coupled plasma atomic emission spectrometry (ICP- AES) (IRIS IntrepidII, Thermo Electron Corporation, USA; Zhang et al., 2010).

The fresh, big and pink nodules removed from the roots of P. pinnata were sterilized by soaking in 95% ethanol for 1 min, then in 0.1% HgCl₂ for 3 min, and washed several times with sterile water. Under aseptic conditions, the surface-sterilized nodules were crushed in a sterilized EP tube, and a small aliquot of the nodule suspension was taken for steaking on Congo red-containing yeast mannitol agar (YMA, 10.0 g/L mannitol, 1.0 g/L yeast extract, 0.5 g/L K₂HPO₄·3H₂O, 0.2 g/LMgSO₄·7H₂O, 0.1 g/L NaCl, and 1.0 g/L CaCO₃ and 0.04 g/L Congo red, pH 7.0–7.2; Zevenhuizen et al., 1986; Sierra et al., 1999). Mucoid and white colonies were selected for repeated re-streaking until single colonies with uniform colony characteristics were observed (Brenner et al., 2005). Purification of rhizobia was confirmed by Gram staining and microscopic examination, and Gram-negative strains were kept for molecular identification (Brooks et al., 2016).

Total DNA was extracted from purified isolates using the phenol-chloroform method (Chang et al., 2011), and 16S ribosomal RNA (rRNA), BOXA1R and four housekeeping genes (atpD, recA, rpoB, glnII) were amplified by PCR as described (Vinuesa et al., 2008; Menna et al., 2009). The PCR products were sequenced by Sangon Biotech (Shanghai, China), and the sequences were submitted to GenBank for assignment of accession numbers. Similarity analysis was performed by searching using BLAST in GenBank. Neighbor-joining (NJ) trees of the 16S rRNA and housekeeping genes were constructed using MEGA7.0 software with bootstrap values of 1,000 replicates (Menna et al., 2009). Sequence data for the four housekeeping genes were concatenated into a single atpD-recA-rpoB-glnII sequence for multilocus sequence analysis (MLSA; Vinuesa et al., 2008; Menna et al., 2009).

The resistance of the 20 isolated rhizobia to Ni, Cd, Mn, and Cu was assayed by measuring their growth in YMA liquid medium (3.0 g/L yeast extract, 5.0 g/L tryptone, 0.7 g/L CaCl₂·2H₂O, pH 7.0) containing different concentrations of metal ions by adding the salts of NiCl₂, CdCl₂, MnSO₄, and CuSO₄, respectively. The bacterial suspensions (50 μL, 10⁸ cells/mL) were inoculated into 5 mL YMA liquid medium. The medium without heavy metal was used as the control. The minimum inhibitory concentration (MIC) and lethal concentration (MLC) were obtained by measuring the optical density (OD_600nm) of the bacterial cultures with a spectrophotometer (UV-3300; Shanghai MAPADA, Shanghai, China) after incubation in an orbital shaker (28°C, 150 rpm) for seven days (Mao et al., 2020). MLC was defined as the lowest concentration of metal ion in solid medium where the isolate growth was not observed, while MIC as the lowest concentration of metal ion in the solid medium where the isolate growth was weaker than that in the heavy metal-free control (Yu et al., 2014). The bacterial cultures were repeated three times for each treatment, and OD_600nm readings were taken in triplicate.

After the amplification of nifH gene of the isolates, the PCR products were sequenced by Sangon Biotech (Shanghai, China), and the sequences were submitted to GenBank for assignment of accession numbers. Similarity analysis was performed using BLAST in GenBank. Neighbor-joining (NJ) trees of the nifH gene were constructed using MEGA7.0 software with a bootstrap value of 1,000 replicates (Menna et al., 2009).

To test the symbiotic nitrogen fixation capacity of the rhizobia isolates, some representative strains of different genera rhizobia were selected for rhizobia-P. pinnata pot experiment by using the potted experimental equipment (Figure 1). Leonard jars, which consisted of two parts, i.e., an upper bottle and a lower jar, were assembled as the apparatus for testing nitrogen fixation activities of rhizobia (Yu et al., 2017a). The lower jar contained nitrogen-free nutrient solution, while the upper bottle was filled with vermiculite as the substrate. The assembled Leonard jars were autoclaved (100 KPa, 121°C) for 30 min after covering the upper bottles with air-filtering films. Some mature and plump seeds of P. pinnata were selected for surface sterilizing using diluted NaClO₃ and ethanol. The seeds were sowed in the upper bottles after germination, and then the air-filtering films were used to cover the upper bottles again. Approximately 3 × 10⁸ of fresh rhizobia cells were inoculated around the rhizosphere of a P. pinnata seedling, and then the sterilized silica sand was used to cover the vermiculite to avoid contamination. P. pinnata trees in the non-inoculation pots was designed as the control (CK), and each treatment repeated three times. The pots were kept in a greenhouse with a day temperature of 25°C for 16 h and a night temperature of 17°C for 8 h. After 6 months, P. pinnata plants were uprooted, and the nodule numbers, plant height, root length, biomass (dry weight) and nitrogen content were measured using the previous methods (Yu et al., 2017a).

The experimental data were averaged out of at least three independent replicates for each treatment. Microsoft Excel 2016 was used to calculate means and standard deviations. IBM SPSS Statistics 26.0 was used to perform Tukey’s test at p < 0.05.

To establish the basic characteristics of the VTM tailings, the physicochemical properties and metal contents were measured for the collected soil samples (Table 1). Soil pH was slightly acidic with a value of 5.77 ± 0.15. The contents of soil organic matter and total nitrogen were very low in the VTM tailings. The available N, P, and K in the VTM tailings accounted for 23.2, 15.2, and 24.9% of the total contents, respectively. The N, P, and K content in the VTM tailings was far lower than the cultivated soil nitrogen content, so the VTM tailings is very barren. As three main components of the VTM tailings, the concentrations of total Ti and Fe were very high as expected at 108 g/kg and 104 g/kg, respectively, while the content of total V (952.3 mg/kg) was relatively low. Interestingly, the concentration of total Mn reached 3,239.20 mg/kg, and the contents of total Ni, Zn, and Cu were more than 100 mg/kg. Only the amounts of total Cr, Pb, and Cd were relatively small in the VTM tailings. The available Ti, Cd, and Cr in tailings were not detected, the content of available V and Zn was very low, and the available Fe was highest, following by Cu, Mn and Ni.

The results showed that P. pinnata can grow well in VTM tailings. A total of 57 rhizobia strains with the characteristic white mucoid colonies, Gram-negative and rod-shape features were isolated from the nodules of P. pinnata growing in the VTM tailings. The similarities among the 57 isolates ranged from 0.45 to 1.00 in the BOX A1R-PCR fingerprint dendrogram, including 49 distinct fingerprint patterns (Figure 2). These strains were clustered into two groups at 45% similarity level, three groups at 49% similarity level, 8 groups at 61% similarity, 18 groups at 74% similarity, and 41 groups at 91% similarity level. There were also some strains on the same branch with 100% similarity, such as PP31, PP75, PP81, PP82, and PP109.

According to the results of BOXA1R-PCR fingerprints analysis (Figure 2), eight isolates with 100% similarity were deleted, and 49 strains were kept for the determination of heavy metal tolerance. Among them, 20 strains showed different levels of tolerance against four heavy metals including Cu, Ni, Mn, and Zn. After culturing in YMA medium with Cu²⁺ for 7 days, the survival rates were 85% at 100 mg/kg, 20% at 200 mg/kg, 10% at 300 mg/kg, and 5% at 400 mg/kg. For nickel (Ni²⁺), they were 45% at 100 mg/kg, 20% at 300 mg/kg, 10% at 500 mg/kg, and 5% at 700 mg/kg. For cadmium (Cd²⁺), they were 55% at 200 mg/kg, 30% at 400 mg/kg, 15% at 600 mg/kg, and 10% at 800 mg/kg. For manganese (Mn²⁺), it was 75% at 500 mg/kg, 65% at 1,300 mg/kg, 20% at 2100 mg/kg, and 10% at 2900 mg/kg.

Only five rhizobia strains (PP1, PP7, PP14, PP69, and PP76) showed relatively higher tolerance to the four heavy metals (Table 2; Figure 3). PP76 tolerated against Ni, Cd, and Mn with an MIC at 100, 200, 300 mg/L, respectively, and with an MLC at 600, 800, 3,200 mg/L, respectively. PP1 showed high tolerance to Cd and Mn with an MIC at 200, 300 mg/L, respectively, and an MLC at 900, 3100 mg/L, respectively. Strains PP7 and PP14 only showed tolerance against Cu with an MIC at 100 mg/L, and an MLC at 400 and 350 mg/L, respectively. Only PP69 showed high tolerance to Cd with an MIC at 100 mg/L and an MLC at 700 mg/L.

The 20 strains with heavy metal tolerance were selected for molecular identification by 16S rRNA gene sequencing. The sequences of 16S rRNA genes were obtained and compared using BLAST in the National Center for Biotechnology Information (NCBI) database. Similarity analysis of 16S rRNA genes showed that the 20 rhizobia strains were classified into three different genera: Bradyrhizobium (12 strains), Ochrobactrum (4 strains) and Rhizobium (4 strains). The phylogenetic tree of the 16S rRNA gene also divided the 20 rhizobia strains into three different branches (Figure 4). Among them, 12 Bradyrhizobium isolates (PP29, PP40, PP47, PP56, PP57, PP69, PP76, PP80, PP90, PP98, PP99, and PP111) were closest to Bradyrhizobium pachyrhizi PAC 48^T with 100% similarity under the same branch (Figure 4). Three Rhizobium isolates (PP1, PP6 and PP18) were 99.12% similar with Rhizobium nepotum Pulawska 39/7^T, whereas Rhizobium sp. PP15 was most closely related to Rhizobium leguminosarum NBRC14778^T with 99.47% similarity. Other four isolates (PP7, PP14, PP20, and PP49) were classified into Ochrobactrum, which was 100% similar with Ochrobactrum lupini NBRC 102587^T under the same branch (Figure 4).

To accurately determine phylogenetic status of the 20 rhizobia, amplified the four house-keeping genes (atpD, recA, rpoB and glnII) of the isolates and built a phylogenetic tree for each genus. For 12 Bradyrhizobium isolates, sequences of the house-keeping genes [atpD (407 bp), recA (380 bp), rpoB (522 bp), and glnII (474 bp)] were used to perform the multi-locus sequence analysis (MLSA) by constructing a longer housekeeping gene fragment (1,783 bp). Then, the phylogenetic tree of these 12 Bradyrhizobium isolates was built using neighbor-joining method (Figure 5C). For 4 Rhizobium isolates, sequences of the house-keeping genes [atpD (298 bp), recA (246 bp), rpoB (554 bp) and glnII (339 bp)] were subjected to multi-locus sequence analysis (MLSA) by constructing a longer housekeeping gene fragment (1,437 bp). Then, the phylogenetic tree of these 4 Rhizobium isolates was built using neighbor-joining method (Figure 5B). For 4 Ochrobactrum isolates, sequences of the four house-keeping genes [atpD (387 bp), recA (385 bp), rpoB (412 bp) and glnII (447 bp)] were subjected to multi-locus sequence analysis (MLSA) by constructing an longer housekeeping gene fragment (1,631 bp) (Figure 5A).

The MLSA phylogenetic tree of the four housekeeping genes at genus level was basically the same as that of the 16S rRNA gene, but there were some differences at species level. Twelve Bradyrhizobium isolates were most closely related to Bradyrhizobium huanghuaihaiense CCBAU 23303^T with 99.24% similarity, and Bradyrhizobium pachyrhizi PAC 48^T with 98.70% similarity. The isolate Rhizobium sp. PP15 was closest to Rhizobium fabae CCBAU 33202^T with 99.35% similarity, and Rhizobium pisi DSM 30132^T with 98.55% similarity. Other three Rhizobium isolates (PP1, PP6, and PP18) were closest to Agrobacterium deltaense YIC4121^T and Agrobacterium tumefaciens NCPPB 2437^T with 98.95 and 98.44% similarity, respectively. Four Ochrobactrum strains were closest to Ochrobactrum anthropi T16R-87^Tand Ochrobactrum lupini LUP21^T with 99.74 and 97.83% similarity, respectively.

We amplified the nifH gene of the isolates and built a phylogenetic tree using neighbor-joining method (Figure 6). As was similar with the trees of 16S rRNA genes and the house-keeping genes, the same genus was clustered together. Four Bradyrhizobium isolates were most closely related to Bradyrhizobium ferriligni CCBAU51502^T, two Rhizobium isolates were most closely related to Rhizobium pusense VLa18^T, and two Ochrobactrum isolates were most closely related to Ochrobactrum anthropi ATCC 49188^T.

The 20 representative rhizobia with heavy metal tolerance were selected to determine their capacity of symbiotic nitrogen fixation using P. pinnata pot experiment. When the 20 rhizobia were, respectively, inoculated around the P. pinnata rhizosphere, only 11 strains built symbiotic relationships with P. pinnata. Moreover, the nodule numbers of P. pinnata inoculated with different rhizobia strains were fully diverse (Table 3). Plants inoculated with Rhizobium produced the highest number of roots nodules, suggesting Rhizobium had the strongest nodulating capability for P. pinnata, Bradyrhizobium was the next, and Ochrobactrum was the weakest at nodulation. In all treatments, more nitrogen content in P. pinnata was found in the aboveground parts than in the roots. The nitrogen content in the inoculated plants was significantly (p < 0.05) higher than that in the non-inoculated control. Except for Bradyrhizobium sp. PP76 and PP69, the trend of other rhizobia strains’ nitrogen fixation capacity was similar to that of symbiotic nodule number with the following order: Rhizobium > Bradyrhizobium > Ochrobactrum. Among them, Rhizobium sp. PP1 showed the strongest nitrogen fixation activity, and the nitrogen content of the plants’ aboveground parts and roots was 2.4–1.8 times that of the non-inoculation control.

Except for nitrogen fixation capacity, these rhizobia strains showed different levels of plant growth promoting activities for P. pinnata. Among them, the plant height, root length and biomass of the non-inoculated control were the lowest. The plant height of P. pinnata inoculated with Ochrobactrum sp. PP14 was the highest, whereas the root length was the lowest, just as in the non-inoculated control. Compared with the non-inoculated control plants, the improved height of the inoculated plants ranged from 47 to 195%, while the increase of root length of the inoculated plants ranged from 119 to 320%. Because of the promoting effect of rhizobia, the biomass of the inoculated P. pinnata plants was significantly (p < 0.05) higher than that of the control plants. Compared with the non-inoculated plants, the biomass of the inoculated treatments was improved by 1.75–4.27 times. The biomass improvement of P. pinnata inoculated Rhizobium sp. PP15 was the highest among all of them.

The quality of the soil depends on its physicochemical properties. Soil pH affects soil microbial diversity and function (Blum et al., 2018), and the pH of VTM tailings in Sichuan Province, China, was found to be quite acidic, which is similar to the soil near a zinc blende mine north of Spain (Pérez-Esteban et al., 2014). Thus, the microbial communities in the VTM tailings must be more adaptable to an acidic environment. The available N, P, K and organic matter were as low as in other mine tailings (e.g., in Mexico), which indicated that the VTM tailings was not very fertile (Armienta et al., 2019). The Ti concentration was up to 38 times higher than that in the similar soil near a Ti mining site in Kenya, and the V concentration was up to 7.5 times higher than that in Cuban soils on average (Alfaro et al., 2014; Maina et al., 2016). The concentration of Fe was approximately 3,000 mg/kg, which is higher than the soil near a steel plant in India (Kaur et al., 2019). Compared to the soil near a coal mine in China with severe Cu, Zn, and Cr pollution, the concentration of these elements in Sichuan VTM tailings was 6.70, 3.69, and 1.54 times higher, respectively; but Pb was lower at 1.45 mg/kg (Liu et al., 2020). The concentration of Mn and Ni in VTM tailings was approximately 5 and 3 times higher, respectively, than in the magnetite tailings after growth of Imperata cylindrica (Yuan et al., 2018). The Cd concentration already exceeded the minimum inhibitory concentration for plant growth (Zhang F. et al., 2019). Consequently, the reason why plants grown in the VTM soil were infertile may be due to the high heavy metal contents and low available N, P, K, and organic matter. Therefore, when carrying out ecological restoration, attention must be paid to reducing the concentration of heavy metals in the soil and increasing the content of nutrients.

As an biofules resource, P. pinnata is a fast-growing leguminous tree with the potential for high oil seed production and can grow on marginal land (Scott et al., 2008). Only two genera of Rhizobium genera including R. pongamiae, R. miluonense, and Bradyrhizobium genera including B. liaoningense, B. elkanii, B. yuanmingense were found to be symbiotic nitrogen fixation with P. pinnata in India and Australia (Rasul et al., 2012; Arpiwi et al., 2013; Kesari et al., 2013). However, three genera rhizobia of Rhizobium, Bradyrhizobium and Ochrobactrum symbiotic with P. pinnata were isolated from the VTM tailings, which revealed there were abundant rhizobia in the VTM tailings. These rhizobia included B. pachyrhizi, R. nepotum, R. nepotum, and O. lupini, indicating P. pinnata rhizobia isolated from the VTM tailings were different form previous reported others. So, the VTM tailings was a resource pool including abundant functional microbiology. Although it was the first time that Ochrobactrum was found to have symbiotic nodulation with P. pinnata, their symbiotic nitrogen fixation efficiency were not high (Table 2).

The distribution of the Rhizobium population can easily be changed by the influence of different environmental factors (Stefan et al., 2018). Because of multiple heavy metal pollution and barren environmental factors, rhizobia symbiotic with P. pinnata for the VTM tailings were different from others. The proportion of Bradyrhizobium strains was highest among three rhizobia genera in the VTM tailings, probably because of stronger resistance of Bradyrhizobium to heavy metals. Compared with R. pongamiae VKLR-01 isolated from root nodules of P. pinnata, their genetic similarity is not high (Kesari et al., 2013). Some of the strains had specific genetic traits which helped to enhance their adaptability to the toxic environment of heavy metal ions. These special rhizobia from the VTM tailings also proved that microbial composition in a rhizobial system has host-specific and bio-geographical distribution characteristics (Zhang et al., 2011).

Although there are multiple heavy metal pollutants and very poor nutrition in there VTM tailings, abundant heavy metal resistant and plant growth promoting bacteria survival in the extreme environment (Yu et al., 2014). Because of the extreme heavy metal environment, P. innata rhizobia from the VTM tailings also showed heavy metal resistance. Environmental factors following the order: soil pH > heavy metals > nitrogen > soil texture had distinct impacts on microbial community (Deng et al., 2018), indicating that heavy metals were very important affection factor for soil microbe. Only 40% rhizobia from the VTM tailings showed tolerance against Cu, Ni, Mn, and Zn. The tolerance concentrations of these metals (except for Mn) for these strains were higher than those in the VTM tailings, and different isolates had different level of tolerance to heavy metals, indicating soil heavy metals are not the only factor affecting strain resistance. Some microbes metabolize and transform heavy metal into a less hazardous form for surviving in such harsh environments, resulting in the formation of heavy-metal-resistant microbes (Prabhakaran et al., 2016), so these microbes had their own unique resistance characteristics. From BOXA1R-PCR fingerprints and Phylogenetic characteristic, rhizobia form the VTM tailings had different genotype, so their tolerance to heavy metals was different, and even some isolates did not showed tolerance to the four tested heavy metals.

In other research on rhizobial systems, most bacteria were only tolerant to a single heavy metal and only a few were resistant to multiple types of heavy metals (Stan et al., 2011; Fan et al., 2018), which was consistent with the tolerance to heavy metals of rhizobia from the VTM tailings. Bradyrhizobium sp. PP76 was resistant to Ni, Cd, Mn of the four metal ions and Rhizobium sp. PP1 was resistant to Cd and Mn, which makes them the best choices for the establishment of symbiotic systems of leguminous plants and rhizobia in heavy metal-contaminated soils.

As a typical function of rhizobia, symbiotic nitrogen fixation is relation to nod, nif and fix genes, such as nifH named as dinitrogenase reductase (Shamseldin, 2013). The nifH gene is the biomarker most widely used to study the ecology and evolution of nitrogen-fixing bacteria (Gaby and Buckley, 2014), so the amplified nifH gene was an initial evidence of the nitrogen-fixing capability of the rhizobia isolates from the VTM tailing. From the phylogenetic tree, the nifH genes of three genera rhizobia were also consistent with the 16S rRNA and house-keeping genes with high similarity among the same genus rhizobia. Because symbiotic nitrogen fixation was decided by series of nif genes, e.g., S. meliloti and R. leguminosarum bv. viciae have a restricted set of 9 and 8 nif genes, respectively (Masson-Boivin et al., 2009). Therefore, although the nifH genes of same genus with high genotype similarity was on the same branch, these rhizbia showed different nitrogen fixation and plant growth capacity. These rhizobia had symbiotic N-fixation ability with the leguminous host plant of P. pinnata and were facilitative in promoting plant growth. Almost of rhizobia showed consistence between symbiotic nitrogen fixation and plant biomass, except for Ochrobactrum sp. PP7 and PP14. Although Ochrobactrum sp. PP7 and PP14 showed lowest symbiotic nitrogen fixation efficiency among the eleven isolates, their plant-growth promoting activity was stronger than some Bradyrhizboium sp. isolates, indicating that Ochrobactrum sp. PP7 and PP14 might had some of other plant-growth promoting capacity (Yu et al., 2017b).

The excessive metal concentrations cause undeniable damage to rhizobia, legumes and their symbiosis to affect efficiency of symbiotic nitrogen fixation (Zahran, 1999; Ahmad et al., 2012), which does not hinder that rhizobia increase phytoremediation by nitrogen fixation and production of plant growth-promoting factors and phytohormones (Pajuelo et al., 2011). So, legume–rhizobium symbioses has been considered as a tool for bioremediation of heavy metal polluted soils (Pajuelo et al., 2011). However, rhizobium should have heavy-metal resistance to improve legume–rhizobium symbiosis in bioremediation of heavy metal polluted soil (El-Tahlawy and Ali, 2021). P. pinnata rhizobia from the VTM tailings did not only show nitrogen fixation capacity but also heavy metal tolerances, so these rhizobia can be used to build symbiosis bioremediation system for heavy metals. P. pinnata inoculated with B. liaoningense PZHK1 was proved to show huge potential for phytoremediation of mine tailings, which had applied for soil and ecological remediation at the VTM tailings (Yu et al., 2017a, 2019). These rhizobium isolates with nitrogen fixation capacity and heavy metal resistance provided excellent microbial resources for bioremediation of the VTM tailings to other heavy metal polluted soil.