The research was conducted in the district of San Martín de Pangoa, Satipo Province, Junín Region, Peru. The area corresponds to the tropical premontane pluvial forest (bp-PMT) life zone, according to the Holdridge World Life Zone Classification System [35]. The site is characterized by maximum and minimum temperatures of 31.6 and 19.3°C, respectively, a relative humidity of 81.5%, and a mean annual precipitation of 1898 mm year⁻¹ [36]. One of the main economic activities in the area is the cultivation of coffee and cacao for both domestic consumption and export markets.

An exploratory observational study was conducted based on purposive field sampling. Seven cacao-growing areas were identified according to the following criteria: plots at the production stage, a minimum area of one-hectare, organic management, manual weed control, and the absence of weeding activities during the two months prior to sampling. To determine soil heavy metal concentrations and identify the area with the highest Cd levels, as well as to characterize its physicochemical properties, soil sampling was carried out following the approach proposed by Havlin et al. [37]. This area was selected to maximize the likelihood of identifying species with metal accumulation capacity, consistent with the exploratory nature of the study. Soil samples were collected at a depth of 30 cm from edaphically homogeneous areas, with three replicates obtained per study site.

Soil samples were analyzed within the Soil, Water, and Foliar Laboratory Network of the National Institute for Agrarian Innovation (LABSAF-INIA). Prior to physicochemical analyses and the determination of Cd and Pb concentrations, samples underwent a pre-treatment consisting of air-drying at temperatures below 40°C and sieving to obtain the <2 mm fraction, in accordance with the procedure established by the International Organization for Standardization [38].

The variables considered for soil characterization included the following parameters and reference methodologies: sand, silt, and clay contents determined by the Bouyoucos hydrometer method [39]. Soil pH measured in a 1:1 soil–water suspension [40]; electrical conductivity (EC) measured in a 1:5 soil–water extract [41]. Organic matter content determined by the Walkley and Black method [39]; total nitrogen measured by the micro-Kjeldahl method [42]. Available phosphorus in neutral and acidic soils determined using the Bray and Kurtz method [39]. Available potassium determined according to Bazán Tapia [43]; exchangeable bases (Ca²⁺, Mg²⁺, K⁺, and Na⁺) extracted with ammonium acetate. Moreover, exchangeable acidity (H⁺, Al³⁺) extracted with potassium chloride [39]. Finally, effective cation exchange capacity (ECEC) was calculated as the sum of exchangeable bases and exchangeable acidity.

Total recoverable Cd concentrations were determined following the USEPA Method 3050B [44]. The resulting solutions were analyzed using flame atomic absorption spectrometry (FAAS). Standard quality assurance and quality control (QA/QC) procedures were applied throughout the analytical process. A calibration coefficient (R²) of 0.9995 or higher was obtained for the standard calibration curve on the flame atomic absorption spectrophotometer prior to sample analysis. Following initial calibration, the calibration curve was verified using an initial calibration blank (ICB) and an initial calibration verification standard (ICV). The ICV was prepared from an independent certified reference material (second source) at or near the mid-range of the calibration curve, with acceptance criteria set at ±10% of the true value. Spike recovery of Cd from a certified reference material (90%–110%) and duplicate analyses performed every 10 samples were also included as part of the QA/QC protocol. The method detection limit for Cd using FAAS was 0.41 mg kg⁻¹.

Soil analysis results indicated that the area known as Ciudad de Dios exhibited the highest Cd concentration (1.65 mg kg⁻¹), exceeding the Peruvian Environmental Quality Standard (EQS) for agricultural soils established by the Ministry of the Environment of Peru [9]. Accordingly, this site was selected for the identification and analysis of spontaneous weed species associated with cacao cultivation.

The selected area comprises an 8-year-old cocoa plantation with a planting density of 3 m × 3 m, arranged in a triangular (staggered) pattern, and previously used for coffee cultivation. The plantation was classified as polyclonal, predominantly composed of CCN-51 (80%), followed by ICS-1 (5%), ICS-6 (5%), TSH-565 (3%), VRAE-15 (2%), VRAE-99 (2%), ICS-39 (1%), ICS-95 (1%), and other minor genotypes (1%). Agronomic management included pruning carried out as required and fertilization based on seabird guano and dolomite (Fertiphos Tropical), resulting in an average productivity of 1000 kg ha⁻¹. The presence of major diseases and pests was observed, including witches’ broom (Moniliophthora perniciosa), frosty pod rot (Moniliophthora roreri), cocoa pod borer (Carmenta foraseminis), and brown spot disease (Phytophthora sp.).

Soil analysis indicated a clay loam texture, a high organic matter content (4.1%), and a moderately acidic pH of 6.10. Electrical conductivity was low (0.09 dS m⁻¹), while total N, available P, and exchangeable K contents were 0.21%, 6.48 mg kg⁻¹ and 77.80 mg kg⁻¹, respectively. The cation exchange capacity (CEC) was 3.50 cmol (+) kg⁻¹, with exchangeable cation concentrations of Ca²⁺ (2.84), Mg²⁺ (0.56), K⁺ (0.08), and Na⁺ (0.02). In addition, to determine heavy metal (Cd and Pb) levels in cocoa plants, samples of leaves, pods, husks, and cotyledons were collected. Sampling followed a zigzag pattern, collecting plant material from ten trees, including five mature leaves with petioles and two mature pods per tree, free of visible disease or pest damage [45].

In the laboratory, leaves were washed with distilled water; cocoa pods were opened, cut into small fragments, and placed in beakers, while cocoa beans underwent a five-day fermentation process followed by shade drying. Leaves, pods, and cocoa beans were subsequently oven-dried at 70°C for 48 h until constant dry weight was achieved [46]. The husk was manually separated from the cotyledons, and each plant component was ground using porcelain mortars (Table 1).

For plant sampling, a systematic survey was conducted across the select plot in which dominant weed species were defined as those exhibiting the highest frequency of occurrence and visual cover within the evaluated plots. Based on this criterion, the selected species were collected, ensuring that sampled individuals per species were spatially distributed across different sectors of the field, thereby avoiding clustering in a single location. Each plant individual constituted an independent experimental unit, allowing the acquisition of biological replicates per species, and was processed individually for subsequent analyses.

Species-specific percentage cover was estimated across the study area, after which dominant weed species were randomly collected with ten replicates at the flowering stage [28]. These ten replicates were obtained to ensure a representative sampling, provide sufficient plant material for laboratory analyses, and support accurate taxonomic identification. From these, three individual plants per species were selected and analyzed independently for heavy metal determination. Samples consisted of both aboveground and root tissues [47]. In addition, targeted rhizospheric soil samples were collected for heavy metal analysis [28]. The ten most representative species throughout the study area were collected and deposited in herbarium presses to facilitate dehydration and prevent the growth of microorganisms, such as fungi [48]. Taxonomic identification was performed according to the APG IV system The Angiosperm Phylogeny Group [49], at the Herbarium of the Department of Biology, Universidad Nacional Agraria La Molina, Lima, Peru.

a. 

Weed biomass

Samples were washed with distilled water to remove adhering soil residues. Subsequently, the aboveground and root portions were separated. Each sample was oven-dried separately at 70°C until a constant weight was achieved [46]. The dried samples were ground using a stainless-steel mill and passed through a 100-mesh sieve [50]. Biomass was expressed on a dry weight basis, including aboveground biomass, root biomass, and total biomass, calculated from three replicates per weed species.

b. 

Bioconcentration factor (BCF) and translocation factor (TF)

The bioconcentration factor (BCF) indicates the ability of plants to remove chemical components from the soil [32]. BCF is defined as the ratio between the total concentration of a metallic element in plant tissues and its concentration in the soil [51]. BCF values < 1 indicate metal-excluding species, whereas BCF values > 1 indicate metal-accumulating species [52,53]. This factor was calculated as follow:

The translocation factor (TF) determines the ability of a plant to transfer a compound from the roots to other plant organs [51]. A plant is considered to exhibit phytostabilization potential when metals are not translocated to other organs and remain immobilized in the root system (TF < 1) [54]. In contrast, TF values > 1 indicate the plant’s capacity to translocate metals from roots to aboveground tissues, such as shoots, and are therefore indicative of bioaccumulation potential [55]. This factor was calculated using the following equation:

Statistical analyses were performed using R software version 4.5 through the RStudio interface (2024 edition). Cd and Pb concentrations, as well as biomass, were modeled using generalized linear models (GLMs) implemented with the glmmTMB package. Model selection was conducted using the MuMIn package based on the corrected Akaike Information Criterion (AICc). Model fit was evaluated using R² as implemented in the performance package. Assumption checking included residual normality, outlier detection, and homogeneity of variances, which were assessed using the DHARMa package. Multiple comparisons of heavy metal concentrations and biomass among weed species were performed using the emmeans package, considering a significance level of α = 0.05 and applying false discovery rate (FDR) correction.

Additionally, the relationships between soil metal concentrations and biomass with metal contents in weed species were modeled using generalized linear models (GLMs) with spline functions (using different degrees of freedom and polynomial orders), implemented through the splines package. Predicted values were obtained using the ggeffects package, and results were visualized using the ggplot2 package.

The weed species recorded in the study area were distributed across eight botanical families. The identified species included Pseudelephantopus spiralis (Less.) Cronquist, Ocimum campechianum Mill., Bidens riparia Kunth, Tripogandra serrulata (Vahl) Handlos, Euphorbia heterophylla L., Talinum paniculatum (Jacq.) Gaertn., Hilleria latifolia (Lam.) H. Walter (HIW), Cyperus aggregatus (Willd.) Endl., Acalypha stachyura Pax, and Malvastrum coromandelianum (L.) Garcke (Table 2). These species are locally known as achicoria, wild basil, chilco, siempre viva, leche-leche, verdolaga, cuerda de indio, pata de gallina, cola de zorro, and pichana, respectively.

The evaluated weed species showed significant differences in biomass production and Cd concentration (p < 0.05) (Table 3). T. paniculatum recorded the highest total biomass (6.15 g) and root biomass (3.72 g), and ranked second in shoot biomass (2.43 g). The rhizospheric soil associated with this species exhibited a Cd concentration of 0.91 mg kg⁻¹, ranking fourth among the evaluated species. In plant tissues, T. paniculatum showed a total Cd concentration of 0.78 mg kg⁻¹, with 0.59 mg kg⁻¹ in shoots, and was notable for presenting the lowest Cd concentration in roots (0.19 mg kg⁻¹). In contrast, M. coromandelianum exhibited the highest shoot biomass (3.30 g); however, it was associated with the third-highest Cd concentration in soil (1.05 mg kg⁻¹) and ranked fourth in shoot Cd concentration (1.04 mg kg⁻¹).

In contrast, T. serrulata exhibited the lowest total biomass (0.57 g) and root biomass (0.10 g), and ranked among the species with the lowest shoot biomass (0.48 g). Nevertheless, this species grew in soils with the third-highest Cd concentration (1.05 mg kg⁻¹) and showed some of the lowest Cd concentrations in plant tissues, including total Cd (0.66 mg kg⁻¹), root Cd (0.42 mg kg⁻¹), and shoot Cd (0.24 mg kg⁻¹). Conversely, C. aggregatus was recorded in soils with the highest Cd concentration in soil (1.49 mg kg⁻¹) and exhibited the highest values of total Cd accumulation in plant (2.68 mg kg⁻¹), as well as the highest root (1.04 mg kg⁻¹) and shoot Cd concentrations (1.64 mg kg⁻¹). However, this species ranked third in total biomass (3.20 g), second in root biomass (1.68 g), and fourth in shoot biomass (1.52 g).

The results indicate a clear relationship between soil Cd content and Cd accumulation in weed species, modeled using generalized linear models with spline functions based on different bases and polynomial degrees. Fig. 1A illustrates the functional relationship between total soil Cd and total Cd in weeds (R² = 0.67, df = 3; polynomial degree = 2). Soils with lower Cd concentrations were associated with O. campechianum and T. paniculatum, whereas T. serrulata; M. coromandelianum exhibited intermediate soil Cd levels. In contrast, the highest soil Cd concentrations were recorded in association with E. heterophylla, C. aggregatus, B. riparia, and P. spiralis, which also showed intermediate to high levels of Cd accumulation in their tissues. Regarding plant metal concentrations, C. aggregatus stood out by exhibiting the highest Cd concentration 2.68 mg kg⁻¹.

Fig. 1B depicts the relationship between total biomass and Cd accumulated in plant tissues (R² = 0.48, df = 4; polynomial degree = 2). Species with lower biomass, such as T. serrulata and O. campechianum, recorded lower Cd contents. M. coromandelianum showed intermediate biomass and Cd accumulation, whereas C. aggregatus displayed intermediate biomass but the highest Cd accumulation among the evaluated species. In contrast, T. paniculatum, despite producing the greatest biomass, exhibited one of the lowest Cd concentrations in plant tissues.

Fig. 1C relates shoot biomass to shoot Cd concentration (R² = 0.34, df = 4; polynomial degree = 2). The species with the lowest shoot biomass, T. serrulata, exhibited the lowest Cd concentrations in aerial tissues. Conversely, E. heterophylla, P. spiralis, despite also presenting low shoot biomass, showed intermediate levels of shoot Cd. C. aggregatus, stood out by exhibiting the highest shoot Cd concentration even under low shoot biomass conditions. Likewise, T. paniculatum, with intermediate shoot biomass, was among the species with the lowest Cd concentrations in aerial tissues; whereas M. coromandelianum displayed the highest shoot biomass accompanied by intermediate Cd storage.

Fig. 1D illustrates the functional relationship between root biomass and root Cd concentration (R² = 0.86, df = 5; polynomial degree = 2). Species with the lowest root biomass, such as T. serrulata and O. campechianum, exhibited the lowest levels of root Cd. In contrast, P. spiralis, B. riparia, M. coromandelianum and E. heterophylla, despite also presenting low root biomass, showed intermediate root Cd concentrations. C. aggregatus displayed intermediate root biomass but the highest root Cd accumulation among the evaluated species. Conversely, T. paniculatum, although exhibiting the greatest root biomass, was among the species with the lowest root Cd concentrations.

The weed species exhibited significant differences in both biomass production and Pb concentration (p < 0.05) (Table 4). T. paniculatum recorded the highest total biomass (6.15 g) and root biomass (3.72 g), while ranking third in shoot biomass (2.43 g). This species was associated with one of the soils that ranked fourth in Pb concentration (12.08 mg kg⁻¹). In addition, T. paniculatum ranked fourth in total Pb accumulation in plant tissues (6.06 mg kg⁻¹), second in root Pb accumulation (3.98 mg kg⁻¹), and sixth in shoot Pb accumulation (2.08 mg kg⁻¹). In contrast, A. stachyura Pax exhibited the highest shoot biomass (4.03 g) and was associated with soils showing the highest Pb concentration (14.90 mg kg⁻¹), while also recording the highest Pb concentrations in aerial tissues (4.15 mg kg⁻¹).

T. serrulata exhibited the lowest total biomass (0.57 g), root biomass (0.10 g), and shoot biomass (0.48 g); however, its rhizospheric soil recorded the fourth highest Pb concentration (12.16 mg kg⁻¹). Regarding Pb levels in plant tissues, this species showed the highest total Pb concentration (11.19 mg kg⁻¹) and root Pb concentration (8.02 mg kg⁻¹), while ranking fourth in shoot Pb concentration (3.17 mg kg⁻¹). In contrast, the rhizospheric soil of C. aggregatus recorded a Pb concentration of 13.08 mg kg⁻¹, ranking second among the evaluated soils. Nevertheless, this species exhibited the lowest total Pb concentration in plant tissues (2.59 mg kg⁻¹) and the lowest shoot Pb concentration (0.56 mg kg⁻¹), while also ranking among the species with the lowest root Pb concentrations (2.03 mg kg⁻¹). Additionally, C. aggregatus ranked fourth in both total biomass (3.20 g) and shoot biomass (1.52 g), and third in root biomass (1.68 g).

Fig. 2 illustrates the overall relationships between soil Pb content and weed biomass, modeled using generalized linear models with spline functions based on different bases and polynomial degrees. Fig. 2A shows the relationship between soil Pb concentration and Pb accumulation in weed species (R² = 0.67, df = 4, degree = 2). At lower soil Pb levels, E. heterophylla, H. latifolia, B. riparia and T. paniculatum were recorded and were associated with lower Pb concentrations; however, T. serrulata, even within this lower range, exhibited the highest Pb concentration among the species. Soils with intermediate Pb concentrations were represented by C. aggregatus, which showed the lowest Pb values, whereas soils with the highest Pb levels were associated with the presence of A. stachyura, which exhibited intermediate Pb concentrations.

Fig. 2B describes the relationship between total biomass and Pb concentration in weed species (R² = 0.28, df = 2, degree = 2). Species with lower biomass, such as E. heterophylla, B. riparia and P. spiralis, exhibited intermediate Pb concentrations; however, T. serrulata, despite belonging to this low-biomass group, recorded the highest Pb. Conversely, C. aggregatus showed intermediate biomass levels but the lowest Pb content, whereas A. stachyura, T. paniculatum, and H. latifolia were characterized by higher biomass and intermediate Pb concentrations. These patterns indicate that metal accumulation is not solely dependent on plant size, but rather on internal processes of uptake, transport, and partitioning associated with species-specific traits and edaphic metal availability, as widely documented in the phytoremediation literature [32].

Fig. 2C shows the relationship between shoot biomass and shoot Pb concentration (R² = 0.66, df = 4, degree = 2). Species with lower shoot biomass included C. aggregatus, which also recorded the lowest shoot Pb concentration, as well as E. heterophylla, T. serrulata, and B. riparia, which exhibited intermediate shoot Pb concentrations. Notably, P. spiralis, despite belonging to this low-shoot-biomass group, showed one of the highest Pb concentrations in aerial tissues. In contrast, T. paniculatum and H. latifolia displayed intermediate shoot biomass values accompanied by intermediate Pb concentrations, whereas A. stachyura, characterized by the highest shoot biomass, ranked among the species with the highest shoot Pb concentration.

Finally, Fig. 2D illustrates the relationship between root biomass and root Pb concentration (R² = 0.36, df = 2, degree = 2). Species with lower root biomass, such as E. heterophylla, B. riparia, and P. spiralis, exhibited the lowest root Pb concentrations. Likewise, species with intermediate root biomass values, including C. aggregatus, A. stachyura, and H. latifolia, as well as T. paniculatum, which exhibited the highest root biomass, were among the species with the lowest root Pb concentrations. However, T. serrulata, despite belonging to this group, showed the highest root Pb concentration.

Fig. 3 shows the functional classification of the evaluated species according to their bioconcentration factor (BCF) and translocation factor (TF) for Cd and Pb, which allowed the identification of their potential roles as phytoextractors, bioaccumulators, or phytostabilizers.

For Cd (Fig. 3A), the evaluated species were distributed across three of the four quadrants defined by the BCF = 1 and TF = 1 thresholds. Species classified as phytoextractors (TF > 1 and BCF < 1) included T. paniculatum (TPG), P. spiralis (PSC), C. aggregatus (CAE), B. riparia (BRK), E. heterophylla (EHL), and M. coromandelianum (MCG). The phytostabilization quadrant (BCF > 1 and TF < 1) was represented by O. campechianum (OCM), indicating a predominant root retention of the metal and limited internal translocation. In contrast, no species were positioned within the bioaccumulator quadrant (BCF > 1 and TF > 1). Finally, T. serrulata (TSH) was located in the non-classified quadrant (TF < 1 and BCF < 1).

In the case of Pb (Fig. 3B), the distribution pattern of the species reflected a more pronounced differentiation of phytoremediation potential compared with Cd. B. riparia (BRK), A. stachyura (ASP), H. latifolia (HLW), P. spiralis (PSC), and E. heterophylla (EHL) were positioned within the phytoextraction quadrant (BCF < 1 and TF > 1). In contrast, no species were classified within the bioaccumulator (BCF > 1 and TF > 1) or phytostabilization (BCF > 1 and TF < 1) quadrants, suggesting a limited capacity of the evaluated species to combine high accumulation or effective root retention of Pb under the conditions of this study. Finally, C. aggregatus (CAE), T. paniculatum (TPG), and T. serrulata (TSH) were located in the non-classified quadrant (BCF < 1 and TF < 1).