Heavy metal-contaminated tailings were sourced from the AngloGold Ashanti mine tailings dam at Pompora, Obuasi (1°3,925”–1°38’24” W and 6°13′11”–6°13’19” N), Ghana. The tailings are old residues from gold extraction activities that used cyanide leaching and flotation methods. Uncontaminated soil was sourced from the botanical garden of the Kwame Nkrumah University of Science and Technology, Kumasi, representing typical tropical agricultural soil with no history of mining.

The experimental design comprised three treatments: garden soil only (GS), 1:1 [mixture of garden soil and tailings (w/v), and pure tailings (PT)]. All soil samples were sieved through a 1 mm sand filter before being weighed into respective pots, each weighing 5 kg. Before the field experiment, the soils were watered to field capacity and initially kept in a greenhouse with natural light, with daytime temperatures ranging from 27 °C to 32 °C.

Four harvest timepoints were established: initial setup after transplanting (H0) and three successive harvests at three-month intervals (H1, H2, H3). Triplicate pots were set up randomly for each treatment (GS, 1:1, and PT) per harvest and arranged in a completely randomised design, totalling 36 experimental samples.

Seeds of L. leucocephala were collected from the Pompora tailings site and stored in a dry container. Seeds were initially germinated in nursery pots containing only garden soil watered to field capacity and kept in a greenhouse. Seeds germinated at an average of 3 days and thinned to one seedling per pot. The seedlings were grown for 2 weeks after germination and then introduced into the experimental treatment pots, which were kept under natural light and temperature conditions in an open field.

Rhizosphere samples were obtained by uprooting plants from treatment pots, gently shaking them to remove bulk soil, and placing them into sterile bags. Rhizosphere samples were collected by vigorously shaking the tightly bound soil attached to the roots, placed in sterile sampling bags. Triplicates of each sample were kept separately for physicochemical and heavy metal analyses. For molecular analysis, triplicates were pooled per treatment to obtain a composite sample representing each treatment-harvest combination, which was stored at −80 °C for 1 month before DNA extraction.

Plant samples were thoroughly washed under slow-running water to remove particles and air-dried for 24 h. Both soil and plant samples designated for heavy metal analysis were oven-dried at 50 °C to a constant weight.

The results of the physicochemical parameters and heavy metal concentrations were analysed using SPSS version 22. The statistical differences in the mean values of the measures were distinguished using the Tukey HSD test at a 95% confidence interval. Correlations between bacterial communities and environmental variables were analysed using canonical correspondence analysis (CCA) in PAST 4.0.

The bioaccumulation factor (BAF) of heavy metals in plant tissues was estimated as follows: BAF = Xt/ Xs, where Xt and Xs represent the concentration of heavy metals in plant tissues and soil, respectively.

The translocation factor (TF) of heavy metals in plant tissues was estimated as follows: TF = X_shoot / X_root, where X_shoot and X_root represent the concentration of heavy metals in shoot and root tissues, respectively.

Rhizospheric physicochemical properties of the rhizosphere changed significantly during phytoremediation (p < 0.05; Table 1). The pH remained acidic across all treatments, with values ranging from 4.57 ± 0.10 to 0.97 ± 0.03, and reduced gradually over time (H0–H4). Specifically, pure tailings showed the most acidic conditions, while garden soil maintained relatively higher pH values. Electrical conductivity was highest at the onset (H0) across all treatments and decreased progressively over time, ranging between 102.50 ± 3.53 μS/cm and 291.00 ± 0.14 μS/cm. Nutrient levels varied significantly among treatments and over time (p < 0.05; Table 1). Available phosphorus ranged from 28.73 ± 1.03 mg/kg to 65.89 ± 1.25 mg/kg, with garden soil consistently showing higher levels than tailing treatments. Total nitrogen content was lowest in pure tailings at the onset and first harvest (0.12 ± 0.03 mg/kg, H0 and H1) and highest in garden soil treatments (0.27 ± 0.002–0.35 ± 0.05). Interestingly, sulphate levels were elevated in tailing treatments (37 ± 0.00 mg/kg–47 ± 1.40 mg/kg) compared to garden soil (14.50 ± 2.12 mg/kg–21.50 ± 2.12 mg/kg), reflecting the sulphur-rich mineralogy of the mine tailings. The cation exchange capacity of the rhizosphere was relatively high in garden soil, with values increasing to the second harvest, which recorded the highest value (26.32 ± 1.47 cmol/Kg), followed by a decline to 16.24 ± 0.34 cmol/Kg. No significant changes in cation exchange capacity were observed between H0–H2 for the 1:1 treatment, with a significant increase at H3 (16.20 ± 0.33). The percentage of organic matter in the rhizosphere reduced at first harvest but increased gradually subsequently across different treatments, with values ranging 1.94 ± 0.03–2.24 ± 0.07, in garden soil, followed by 1:1 (1.24 ± 0.02–1.82 ± 0.04) and pure tailing (0.41 ± 0.02–1.03 ± 0.02) in descending order.

The concentration of iron (Fe) was the highest among all metals studied, ranging from 126.00 ± 1.40 to 205,142.00 ± 2.82 mg/kg (Table 2). Metal concentrations in the rhizosphere followed a contamination gradient consistently (PT > 1:1 > GS) and decreased progressively from the onset to the final harvest (H0 > H1 > H2 > H3) across all treatments. Manganese levels were the second highest concentrations (32.50 ± 0.72–10,014.50 ± 1.70 mg/kg), while the levels of Cu, Zn, As, and Cd ranged between (31.70 ± 0.42–788.30 ± 2.44 mg/kg), Zn levels (5.00 ± 0.28–298.00 ± 1.41 mg/kg), As levels (0.30 ± 0.05–110.61 ± 0.56), and Cd levels (0.39 ± 0.01–2.45 ± 0.01 mg/kg), respectively.

Leucaena leucocephala (Lam.) de Wit demonstrated notable capacity for metal tolerance and uptake across all treatments. Distinguished accumulation of Fe was observed in shoot tissues (14,605 ± 1.40 mg/kg) in pure tailings at the final harvest (H3), which is above the hyperaccumulation threshold of 10,000 mg/kg (Supplementary Table 2). Conversely, manganese showed preferential accumulation in roots, with maximum concentrations of 12,279.20 ± 1.13 mg/kg occurring in the 1:1 treatment (Supplementary Table 1). Similarly, superior accumulation in root tissues was observed for arsenic and cadmium, indicating limited translocation to aboveground or shoot tissues. Alternatively, the accumulation of Zn was higher in shoot tissues (24.35 ± 0.21–4,699.10 ± 1.40 mg/kg) along an increasing contamination gradient (GS < 1:1 < PT), which is typical of micronutrients.

Correlation Heatmap analysis unveiled positive associations between physicochemical parameters (pH, phosphorus, nitrogen, CEC, organic matter and sulphate levels) and bacterial diversity (Supplementary Figure 10). Additionally, metal levels in the rhizosphere correlated negatively with species richness or abundance (ACE, Chao1, Observed ASVs).

Network analysis revealed three major modules comprising 81 nodes and 678 edges (Supplementary Tables 4, 5). The network showed high connectivity and moderate transitivity, indicating a well-connected community with substantial redundancy. Three bacterial genera emerged as keystone species based on network centrality metrics: Arthrobacter (highest degree and eigenvector centrality), Skermanella (high betweenness centrality) and Chelatococcus (high connectivity within modules). Arthrobacter showed the highest degree (40) and eigenvector centrality (0.235), confirming its essential role in community interactions. Despite the low abundance of Skermanella, it exhibited the maximum betweenness centrality (210.45), indicating its vital role as a connector between different community modules. Actinobacteria, Proteobacteria and Acidobacteria preferentially sustained more functions within networks as connectors and module hubs compared to minor taxa, which served as peripheral nodes within and between ecological networks (Supplementary Figure 11).

Strong bioaccumulation of Fe and Zn occurred in roots, whereas As and Cd showed strong bioaccumulation in the shoot. Bioaccumulation of heavy metals in roots was relatively high in pure tailings and later harvests (H2–H3; Supplementary Table 6a). Zinc bioaccumulation was superior to other metals, with values ranging from 2.5–757.43, and this was followed by Mn (0.12–140.27), Fe (1.01–64.62), As (0.17–10.85), Cd (0.03–3.37), and Cu (0.51–11.51) in descending order. The bioaccumulation of metals in the shoot was relatively high in 1:1 and later harvests (H2–H3; Supplementary Table 6b). The bioaccumulation of Zn was superior to other metals, with values ranging 0.38–300.49, followed by Mn (0.77–208.12), Fe (0.09–48.82), As (0.11–34.21), Cu (0.34–12.02) and Cd (0.02–7.42) in descending order.

The translocation of heavy metals in L. leucocephala was higher in garden soil treatments than 1:1 and pure tailings (Supplementary Table 7). In metallifoerus treatments (1:1 and pure tailings), Zn and Fe exhibited higher translocation, whereas moderate translocation was observed for As, Mn and Cu.

Abiotic factors (pH, conductivity, nutrients, and heavy metals) affect the soil’s bacterial community composition and structure (Huang et al., 2021; Liu et al., 2023; Sun et al., 2018). Across all treatments, pH remained consistently acidic (4.57–5.97), with the most extreme acidification in pure tailings. The observed acidification of the rhizosphere represents a key adaptive mechanism that enhances metal solubility and bioavailability for uptake into plant tissues (Park et al., 2023; Wan et al., 2023; Yang et al., 2022). The electrical conductivity (EC) within the rhizosphere increased at the first harvest but decreased gradually across all treatments, and coupled with acidification, suggests active modification of the rhizosphere chemical environment through root exudates, microbial metabolites and organic acid production (Amin et al., 2012; Mulati et al., 2023; Ni et al., 2019).

Progressive nutrient depletion, especially nitrogen and phosphorus, likely reflects intensive utilisation for metal transport proteins, stress response mechanisms, and cellular repair processes, which has been recognised in other research outcomes (Li Q, et al., 2022; Bertrand et al., 2021; Skuza et al., 2022; Qian et al., 2023; Asare et al., 2023). Although gold mine tailings are associated with low organic matter and nutrient levels, they contain high levels of sulphur in the mineralogical sources (pyrite, galena and chalcopyrite), providing opportunities for sulphur metabolising bacteria and challenges for other bacterial communities (Huang et al., 2012; Lottermoser, 2011; Mensah et al., 2020). Organic matter levels and CEC were higher in garden soil and 1:1 treatments than in pure tailings, although they were within normal (low–moderate) ranges characteristic of Ghanaian soils (Attiogbé et al., 2026; Doe et al., 2022). Moreover, these levels showed low–moderate associations with bacterial community diversity, highlighting the challenges faced by rhizosphere microbial communities, especially in pure tailings, which have limited carbon sources and poor nutrient retention capacity (Hu et al., 2021).

Leuceana leucocephala demonstrated remarkable potential in the phytoextraction of metal ions in mine tailing, achieving hyperaccumulation of Fe (10,000 mg/kg) in shoot tissues (Doku et al., 2024; Kahangwa et al., 2021; Kang et al., 2018; Petelka et al., 2019). The aggressive uptake of Fe is linked to the sequestrating activities of plant growth-promoting rhizobacteria and siderophores, which serve as a mechanism for overcoming heavy metal-induced stress (Bomfim et al., 2021; Liu et al., 2024; Yan et al., 2020). Moreover, they have been known to secrete siderophores, phenolics, and organic acids that target the sequestration of divalent ions (Cu²⁺, Zn²⁺, Mn²⁺ and Ca²⁺), chiefly Fe, to maintain vital metabolic processes (Lee et al., 2023; Narayanan and Ma, 2023; Qu et al., 2020; Yang et al., 2022).

The high rhizospheric concentrations of Mn (<3,000 mg/kg) in pure tailing and 1:1 treatments, coupled with an acidic rhizosphere, favoured preferential initial accumulation in root tissues despite the competitive limitation imposed by elevated Fe levels (Abubakari et al., 2022; Garcia et al., 2020; Dey et al., 2023; Yu et al., 2019; Yang et al., 2008). Despite the high mobility of As and Cd in plant tissues, they showed preferential retention in the root, which conflicts with other reports (Kahangwa et al., 2021; Muehe et al., 2015; Saldarriaga et al., 2023; Yan et al., 2022). The minimal background concentrations, vacuolar compartmentalisation in root tissues, and mechanisms which limit translocation to protect other organs and photosynthetic tissues could account for poor shoot accumulation of Cd and As (Liu et al., 2023; Luo and Zhang, 2021; Geng et al., 2023; Khan et al., 2021; Nabi et al., 2021; Schneider et al., 2017). Metal accumulation generally increased over time (H0 < H1 < H2 < H3), indicating a progressive uptake capacity as plants matured and root systems expanded.

The results show changing dynamics in the bacterial abundance and diversity; an initial decrease (H0–H1) followed by an increase (H1–H2) and a decrease (H2–H3) demonstrates a deterministic community assembly of bacterial communities in the rhizosphere of L. leucocephala under metal stress and oligotrophic conditions during phytoremediation (Khalid et al., 2023; Muratova et al., 2023; Park et al., 2023; Sun et al., 2024). Specifically, the initial community disruption (H0–H1) was followed by selective enrichment of metal-tolerant taxa (Nocardioides and Streptomyces), representing a notable ecological response to extreme environmental filtering. It is worth noting that the overall bacterial community abundance and diversity are lower compared to a previous report of tailings and rhizosphere of L. leucocephala, limiting the possibilities of remediating tropical gold tailings (Gagnon et al., 2020; Doku et al., 2024; Qian et al., 2023; Trovão et al., 2024). The collective adverse impacts of elevated metal levels, acidification, and depleted nutrient levels, as shown by the correlation heatmap, favour deterministic assembly processes that sustain microbial survival and metal sequestration over time (Lei et al., 2024; Maretto et al., 2022; Romero et al., 2021; Sun et al., 2022). Again, the recovery of bacterial diversity at the later stages of experimentation (H2–H3) in tailings was concomitant with the depletion of metal ions in the rhizosphere and subsequent uptake in root tissues, thus reflecting the successful adaptation and niche differentiation among surviving taxa. More so, the temporal succession pattern suggests that established phytoremediation systems may achieve greater microbial stability and functional capacity over time. Beta diversity analysis revealed significant differences among treatments, suggesting that contamination intensity drives shifts in bacterial community assembly (Guo et al., 2019; Huang et al., 2021; Saldarriaga et al., 2023; Yu et al., 2022).

The findings of this research are relevant to improving microbiome-enhanced phytoremediation strategies, including the development of bacterial inoculants using Arthrobacter, Gaiella, Streptomyces, and Nocardioides. Additionally, improving soil nutrient levels and pH promotes the selective growth of beneficial microbiomes and the utilisation of microbial dynamics to monitor the progress of phytoremediation efforts.

While the study provides valuable insights into bacterial community diversity during the phytoremediation of tailings, initial nursing in the greenhouse may not accurately capture field conditions, such as weather variability and soil heterogeneity. Moreover, using composite samples for metagemoic profiling limit the robustness of statistical analysis of bacterial communities, whereas predicted functional roles may not accurately capture the exact contributions provided by culture-based methods. Again, the three-month sampling period intervals targeted long-term dynamics but may have missed shorter-term dynamics important for understanding community assembly.