Seed Yield, Heritability of Pod Traits, and Antibacterial Potential of Yam Bean (Pachyrhizus spp.) Progenies in the Central Amazon

Abstract

Yam bean (Pachyrhizus spp.) is a promising, yet underutilized, leguminous crop with considerable potential for biopesticide development. However, baseline data on seed yield, the heritability of pod traits, and antibacterial effects against Ralstonia solanacearum (RS) are lacking, limiting informed selection for seed production and biopesticide use. This study evaluated the morpho-agronomic traits of nine Amazonian yam bean progenies and screened their seed extracts for in vitro antibacterial activity against RS. We conducted the experiment on a Spodosol, using a trellis system in a randomized complete block design with four replications and four plants per plot. Seed yield varied from 0.56 to 0.98 t ha⁻¹, with progenies P14, P20, and P23 exhibiting the highest numerical yields. Broad-sense heritability (H²) was high for pod length (0.77), pod width (0.81), and 100-seed mass (0.85), suggesting considerable potential for genetic improvement of these traits. In vitro antibacterial assays revealed concentration-dependent biphasic activity: 0.5% (w/v) extracts initially stimulated bacterial growth at 6 and 12 h but suppressed multiplication by 88–90% at 24 h, whereas 5% (w/v) extracts promoted growth throughout the 24 h period. These findings establish baseline seed yield data for yam bean grown in Amazonian Spodosols and highlight the genetic variability available for breeding programs in the region. Progenies P14, P20, and P23 warrant further investigation as sources of natural compounds for sustainable plant protection, underscoring the need for optimized formulations and extended exposure periods.

1. Introduction

In recent decades, the global agricultural sector has faced increasing pressure to enhance food security while simultaneously reducing reliance on synthetic pesticides owing to environmental concerns, pest resistance, and the impact on human health [1]. This paradigm shift has spurred significant interest in sustainable agricultural practices and the exploration of plant-derived biopesticides as environmentally friendly alternatives [2]. Plant-based biopesticides offer a promising avenue for controlling various pests and pathogens, often with high specificity and rapid biodegradability, contributing to more resilient and sustainable agroecosystems [2,3].

Despite the growing recognition of biopesticides, their large-scale adoption is often hampered by challenges related to the consistent supply of raw materials and the lack of established agronomic practices for cultivating biopesticide-producing plants [4]. Therefore, identifying and developing neglected or underutilized crops with dual potential (food production and bioactive compound sources) is crucial for advancing sustainable agriculture, particularly in biodiversity-rich regions such as the Amazon.

Among these alternatives, plant-derived biopesticides have shown promise in controlling pests while maintaining the beneficial insect populations [5]. Given that crop losses induced by pathogens and pests can range from 8% to 40% globally [6], there is an urgent need for effective pest control strategies that are both environmentally benign and efficacious.

Despite the growing interest in the development of biopesticides, significant challenges remain in large-scale production and efficacy assessment against different phytopathogens. These challenges stem from costly production methods, variable efficacy, and a narrow spectrum of action, among other factors [7,8].

Yam bean (Pachyrhizus spp.), an herbaceous climbing vine legume of the Fabaceae family, is a promising candidate for large-scale biopesticide production. Pachyrhizus comprises five identified species: Pachyrhizus tuberosus (Lam.) Spreng, P. erosus (L.) Urb., P. ahipa (Wedd.) Parodi, P. ferrugineus (Piper) M.Sørensen, and P. panamensis R.T.Clausen [9]. These species are adapted to nutrient-poor soils in the Amazon and Andean regions [10]. They produce edible tuberous roots [11], with yields of up to 108 t ha⁻¹ [10], and their seeds contain toxic compounds [12].

Yam bean seeds have demonstrated potential as biopesticides. They have proven effective against plant fungi, such as Sclerotium rolfsii Sacc. (syn. Agroathelia rolfsii (Sacc.) Redhead & Mullineux) [13,14], Colletotrichum gloeosporioides (Penz.) Penz. & Sacc., Fusarium oxysporum Schltdl., and Rhizopus stolonifer (Ehrenb.) Vuill. [15], bacteria, such as Ralstonia solanacearum (Smith) Yabuuchi et al. [14], and various insect pests across multiple orders (Lepidoptera, Coleoptera, Hemiptera, and Diptera), notably Spodoptera frugiperda (J.E. Smith) [16], Helicoverpa armigera (Hübner) [17], Bemisia tabaci (Gennadius) [18], and Plutella xylostella (Linnaeus) [19].

These toxic properties are attributed to diverse bioactive compounds, including rotenoids (rotenone, erosone, pachyrrhizone, dolineone, and pachyrrhizine), isoflavonoids (neotenone and dehydroneotenone), and other secondary metabolites [12,15,20,21,22], as well as defensins, such as SPE10 protein [23]. Among these, rotenone was used as a commercial pesticide until the 1940s, when dichlorodiphenyltrichloroethane (DDT) became dominant [24].

In recent decades, rotenone has regained importance in agriculture because of its rapid biodegradability [25] and low human toxicity. Its acute oral toxicity is characterized by LD₅₀ values for rats ranging from 60 to 135 mg kg⁻¹ [26], and a fatal case in humans involved an estimated dose of 40 mg kg⁻¹ (“Rotenone [MAK Value Documentation, 2003]” 2012). The half-lives of rotenone in plants vary from 0.89 to 1.55 days in cabbage [27], 1.98 to 2.76 days in soil [27], or 5 to 8 days depending on the soil type [28].

Therefore, the Pachyrhizus breeding program at the National Institute for Amazonian Research (INPA, Brazil) was expanded to include a research line targeting seed toxicity and yield, in addition to the traditional focus on tuberous root yield improvement [14]. Of the 64 INPA progenies, nine showed high efficacy against Sclerotium rolfsii in vitro [13], but have not yet been evaluated for seed yield or screened against other economically important pathogens.

Ralstonia solanacearum (RS) is a critical soil-borne bacterial pathogen that causes bacterial wilt in numerous economically important Amazonian crops, including tomatoes, bananas, and hot peppers [29]. This pathogen is the primary constraint to tomato production in Amazonas State, Brazil, where efficient control methods are unavailable [14].

We hypothesized that some of the nine yam bean progenies selected by Lima et al. [13] would suppress RS multiplication and bear a high seed yield. Therefore, this study aimed to (i) evaluate the seed yield and key morpho-agronomic traits of nine selected Pachyrhizus progenies under trellis cultivation in the Central Amazon and (ii) conduct a preliminary in vitro screening of seed extracts from these progenies for antibacterial activity against the RS.

2. Materials and Methods

2.1. Field Experiment

2.1.1. Location

The study was conducted at the Horticultural Research Station of INPA at Km 14 of Highway AM-010, Municipality of Manaus, Amazonas, Brazil. The soil was classified as Spodosol with a sandy texture. The climatic variables during the experimental period were obtained from INMET (2025) and are shown in Figure 1.

2.1.2. Material

Seeds from nine yam bean progenies (P1, P5, P14, P15, P20, P23, P46, P61, and P64), previously selected for toxicity against Sclerotium rolfsii [13], were obtained from the INPA Vegetable Germplasm Bank in Manaus, Brazil. Owing to within-progeny color variation, seeds were grouped by color: beige (P1, P5, P14, P15, P20, P23, and P64), brown (P46), and red (P61).

2.1.3. Soil Preparation

First, a composite soil sample was collected and analyzed at the Soil Laboratory of the Federal University of Amazonas. The soil exhibited the following characteristics: pH 4.8, P 168 mg dm⁻³, K 22 mg dm⁻³, Ca 2.8 cmolc dm⁻³, Mg 0.4 cmolc dm⁻³, Na 9.0 mg dm⁻³, H + Al 1.82 cmolc dm⁻³, CEC (pH 7) 5.12 cmolc dm⁻³, effective CEC 3.3 cmolc dm⁻³, and base saturation 64.5%. On the day of transplanting, soil preparation was conducted using a heavy tractor (Massey Ferguson 250X, AGCO Corporation, Jundiaí, Brazil) equipped with disc plows, making three passes. Immediately, furrows were created at 1-m intervals using a micro-tractor (Yanmar TC-14, Agritech Lavrale, Indaiatuba, Brazil), and mineral fertilization (N-P-K at 0-40-40 kg ha⁻¹) [30] and dolomitic lime (2 t ha⁻¹) were applied in the furrows and incorporated into the soil using the micro-tractor.

2.1.4. Seedling Preparation and Field Establishment

Seeds were pre-soaked in water for 24 h at ~25 °C and sown in April 2021 in 300 mL cups filled with a commercial substrate (Tropstrato HT Hortaliças, Vida Verde, Mogi Mirim, São Paulo, Brazil). Thirty-day-old seedlings were transplanted to the field in a randomized complete block design with nine treatments (progenies), four replications, and four plants per plot arranged in double rows (1 × 1 m spacing). This minimum plot size was adopted based on recommendations for climbing plants [31].

2.1.5. Cultivation

Plants were trained using 2.5-m gliricídia stakes (Gliricidia sepium (Jack.) Kunth) positioned adjacent to each plant. Stakes were secured with twine to galvanized wire (No. 12) that was stretched longitudinally along the planting rows and supported by bamboo poles at 2-m height. Manual weeding was conducted at 15-day intervals, and irrigation was restricted to the first month following field transplantation.

2.1.6. Harvest and Evaluation

Pods were harvested three times between October and November 2021, upon reaching full maturity and dryness. For each plot, the following agronomic traits were measured: plant stand, pod yield, seed yield (both in t ha⁻¹), 100-seed mass (g), number of pods per plant, and pod dimensions. Pod length and width (cm) were determined using ten randomly selected pods per plot. Yield was calculated using the following formula: yield (t ha⁻¹) = [average mass per plant (kg) × 10,000 plants ha⁻¹] ÷ 1000. Broad-sense heritability (H²) was estimated using Formula (1).

$$H^2={\displaystyle \frac{{\sigma }_g^2}{{\sigma }_g^2+\frac{{\sigma }_e^2}{r}}}$$

(1)

where: ${\sigma }_g^2$ = genetic variance, ${\sigma }_e^2$ = error variance, and r = number of blocks = 4.

2.2. In Vitro Experiment

2.2.1. Material

Seeds from nine Pachyrhizus spp. progenies obtained from the field experiment were separated by color (beige, red, and brown) (Figure 2). Due to limited seed availability, P20 (red) and P64 (red) were excluded, resulting in 16 samples for the antibacterial assay.

The Ralstonia solanacearum phylotype II isolate FIO104B was obtained from the INPA Phytopathogenic Microorganism Collection. This isolate was originally collected from tomatoes in Iranduba, Amazonas, Brazil. The bacterial culture was grown on yeast–peptone–glucose–agar (YPGA) medium in Petri dishes at 28 °C for 48 h.

2.2.2. Preparation of Aqueous Seed Extracts

Seeds from the field experiment were sterilized as follows: (i) immersion in 70% ethanol for five minutes; (ii) immersion in commercial bleach (sodium hypochlorite, 2–2.5% active chlorine) for five minutes; and (iii) ten washes with distilled water. Seeds were blotted dry with paper towels, stored in aluminum foil envelopes, and oven-dried at 50 °C for 48 h. The samples were then ground in a ball mill for five minutes at 15 Hz s⁻¹. The resulting powder was stored in airtight plastic containers, treated with ultraviolet light for 20 min, placed in aluminum envelopes, and kept at −17 °C.

The extract was prepared by macerating the powder from each progeny with sterilized water (1:10, w/v) at 10 °C for 48 h. After maceration, the extract was purified using a 3.0 µm membrane filter, followed by centrifugation at 5500 rpm for 15 min. The supernatant was filtered again through 0.45 µm and 0.22 µm mesh membranes. The purified extract was stored at a concentration of 10% in sterile Erlenmeyer flasks and frozen at −17 °C.

The extracts were then adjusted to 0.5% (w/v) and 5% (w/v) concentrations by mixing them with sterile yeast–peptone–glucose (YPG) medium. The controls consisted of an antibiotic control (tetracycline^® mixed with YPG medium to obtain final concentrations of 200, 240, and 300 mg L⁻¹), bacterial growth control (YPG + bacteria, no extract), and sterility control (YPG without bacteria and extract). All treatments were adjusted to a final volume of 20 mL.

2.2.3. In Vitro Antibacterial Assay

Bacterial inoculation was performed by adding 30 µL of the suspension (1 × 10⁶ cells mL⁻¹) to 20 mL of extract [32], resulting in a final concentration of 1.5 × 10³ cells mL⁻¹. The samples were incubated in a shaker at 150 rpm and 25 °C. Aliquots were collected at 6, 12, and 24 h and stored at −17 °C for subsequent cell counting using a Neubauer chamber.

2.2.4. Statistical Analysis

Data were subjected to analysis of variance (ANOVA), and the means were compared using Duncan’s multiple range test (p < 0.05). All analyses were conducted using the PROC GLM procedure of SAS 9.4 software.

3. Results

3.1. Agronomical Performance and Heritability

Seed yield among progenies ranged from 0.56 to 0.98 t ha⁻¹, with no statistically significant differences (

Table 1. Mean values of seven morpho-agronomic traits of nine yam bean (Pachyrhizus spp.) progenies grown on a Spodosol, Manaus, Brazil, 2021.

Progeny	Stand	Pod Yield (t ha−1)	Pod Length (cm)	Pod Width (cm)	Pods per Plant	100-Seed Mass (g)	Seed Yield (t ha−1)
P1	3.50 †	1.88 †	11.15 a	1.41 abc	52.95 †	21.70 ab	0.80 †
P5	3.50	1.47	10.90 a	1.46 a	41.12	23.20 a	0.56
P14	4.00	1.84	11.11 a	1.37 bcd	55.56	19.63 cd	0.83
P15	3.75	1.50	10.70 a	1.40 abc	47.50	20.93 bc	0.66
P20	4.00	1.98	10.90 a	1.35 cd	56.31	20.90 bc	0.86
P23	3.75	2.21	11.30 a	1.46 a	59.70	22.2 ab	0.98
P46	3.50	1.49	10.80 a	1.40 abc	44.81	18.99 d	0.61
P61	3.75	1.69	10.91 a	1.43 ab	48.33	22.64 a	0.74
P64	3.50	1.57	10.07 b	1.32 d	48.60	20.52 bcd	0.69
Mean	3.70	1.74	10.90	1.40	50.55	21.20	0.80
CV (%)	14.21	32.20	3.10	2.90	30.83	5.00	35.42
R2	0.31	0.36	0.61	0.65	0.35	0.74	0.34
H2	0	0	0.77	0.81	0	0.85	0

Different letters indicate significant differences according to Duncan’s test (p < 0.05). ^† Means are not significantly different according to the F-test (p < 0.05). H² = broad-sense heritability.

). Similarly, stand and pod yields showed no significant variations. However, three morphological traits exhibited significant differences (p < 0.05): mass of 100 seeds, pod length, and pod width (Table 1).

Progenies P14, P20, and P23 exhibited the highest numerical seed yields (0.83–0.98 t ha⁻¹) and pod yields (1.84–2.21 t ha⁻¹) (Table 1). Therefore, these progenies should be regarded as candidates for further testing.

The broad-sense heritability (H²) ranged from 0.77 to 0.85 for pod morphology and seed mass (pod length, pod width, and 100-seed mass) (Table 1). However, H² for stand, pod yield, pods per plant, and seed yield was considered zero because the error mean square exceeded the progeny mean square, precluding estimation of the genetic variance component.

3.2. Effect of 0.5% Yam Bean Extracts

At 6 and 12 h of incubation, Duncan’s test (p < 0.05) showed that the yeast–peptone–glucose (YPG) medium with extracts had higher bacterial cell counts than the bacterial growth control (YPG + bacteria, no extracts) (Figure 3). In contrast, at 24 h, the bacterial growth control exhibited significantly higher cell numbers than those of the extract treatments (Figure 3). These results indicate that the extracts initially promoted bacterial multiplication for up to 12 h, but after 24 h, they exhibited toxic effects on RS. While the bacterial growth control increased exponentially from 7.06 to 10.10 log₁₀ cells mL⁻¹ between 6 and 24 h, the extract-treated samples showed limited growth from ~8.54 to ~9.30 log₁₀ cells mL⁻¹ on average, representing an approximately 6-fold reduction in bacterial multiplication.

Regarding the effect of tetracycline (200, 240, and 300 mg L⁻¹), bacterial cell counts at 6 h were significantly lower than those in the bacterial growth control (Figure 3) and remained stable at approximately 6.60 log₁₀ cells mL⁻¹ at all time points (6, 12, and 24 h). These results indicate that tetracycline exerted rapid antibacterial activity, with inhibition occurring within the first 6 h of incubation and being maintained thereafter.

Among all progenies tested, P14 beige and P23 beige exhibited the highest antibacterial efficacy at 24 h, with cell counts of 9.11 and 9.09 log₁₀ cells mL⁻¹, respectively (p < 0.05). These extracts reduced RS multiplication by approximately 10-fold compared to that of the bacterial growth control. In contrast, P1 red (9.60 log₁₀ cells mL⁻¹) showed the lowest antibacterial activity, limiting bacterial growth by only 3-fold.

3.3. Effect of 5% Yam Bean Extracts

At 6, 12, and 24 h of incubation, Duncan’s test (p < 0.05) showed that the YPG culture medium with 5% extract had higher bacterial cell counts than the bacterial growth control (YPG + bacteria) (Figure 4). This indicates that the 5% extract promoted bacterial multiplication throughout the 24 h incubation period. Figure 4 also shows that the cell numbers in these extracts at 24 h reached 10.82 log₁₀ cells mL⁻¹ and bacterial growth control (10.10 log₁₀ cells mL⁻¹), indicating that the toxic effect on RS remained minimal at 24 h after inoculation.

At 24 h, the progenies differed significantly in cell concentration: P5 beige yielded the lowest values (10.54 log₁₀ cells mL⁻¹) and P1 red and P14 red the highest (10.82 and 10.81 log₁₀ cells mL⁻¹, respectively). This pattern may reflect differences in the extract composition that affect RS growth, including nutrient availability.

Despite their strong inhibition at 0.5% after 24 h, P14 beige and P23 beige only modestly multiplied at 5%, reaching 10.69 and 10.63 log₁₀ cells mL⁻¹, respectively. This suggests that these progenies may still exhibit comparatively greater antibacterial potential after 24 h, even at higher extract concentrations.

4. Discussion

4.1. Seed Yield and Agronomical Traits

Currently, the commercialization of botanical pesticides faces major challenges because it requires large quantities of plant material, standardization, methods for characterizing active substances and impurities, data on safe use and environmental impact, and compliance with very rigorous legislation [33,34]. For these reasons, producing sufficient plant material is essential, which is why this study focused on seed yield as a first step toward developing Pachyrhizus as a biopesticide crop.

Duncan’s test showed no significant differences in seed yield among the nine progenies, whose yields ranged from 0.56 to 0.98 t ha⁻¹ under the evaluated conditions. However, the highest mean yield approached 1 t ha⁻¹, which is greater than the seed yields reported for approximately 36% of other grain legume species [35]. The evaluated progenies had not been specifically bred for seed yield, and genetic variability was observed within progenies, for example, in flower (white and violet) and seed color. These findings indicate that the species has considerable potential for increasing seed yield if breeding methods appropriate for allogamous plants are applied, such as population improvement or the selection of pure lines and hybrids.

The present study also highlighted the advantages of growing Pachyrhizus using a trellis system. The mean pod yield reached 1.74 t ha⁻¹ (Table 1), which was 164% higher than the 0.66 t ha⁻¹ reported by Silva et al. [36] for 20 INPA progenies cultivated at the same site without staking. The use of stakes for vertical plant support was the main difference in management between the experiments. This substantial yield improvement suggests that trellis systems are essential for maximizing pod and seed production in Pachyrhizus spp.

Water availability critically influences seed productivity in Pachyrhizus spp. This study evaluated progenies on sandy Spodosol without supplemental irrigation (except for the first month post-transplanting) during a seven-month growing cycle (April–November) spanning the Central Amazon dry season (June–October). Zanklan et al. demonstrated seed yield reductions under water stress, with P. ahipa declining from 3.04 to 1.10 t ha⁻¹ (64% reduction) and P. erosus from 5.66 to 4.69 t ha⁻¹ (17% reduction). Cultivation on Latosols during the rainy season (beginning November–December) could enhance seed yields, as these soils offer better fertility and structure than sandy Spodosols, and natural precipitation would eliminate the need for supplemental irrigation.

In contrast, tuber yield at harvest was near zero in this study, consistent with a previous report showing that increasing tuber yield requires the removal of flowers [37]. Therefore, these progenies are recommended for seed production tests rather than for dual-purpose use (seed and tuber yield).

4.2. Heritability

Broad-sense heritability (H²) was high for pod length (H² = 0.77), pod width (H² = 0.81), and 100-seed mass (H² = 0.85), suggesting that phenotypic selection for these traits should produce genetic gain.

In contrast, H² for pod and seed yields was zero. This contrast between high and zero heritability values is central to the interpretation of the trial. Experimental design, local control and field management cannot explain the absence of detectable genetic variance for the yield traits, since these same factors applied equally to the dimensional traits, which showed coefficients of variation below 5%. The dichotomy instead points to the sampling unit itself: dimensional traits are averaged over a large number of pods and seeds within each plot, whereas yield traits are derived from only four plants.

Under the classical model for plot-based observations, the variance of a plot mean composed of n plants, V(n), may be decomposed into two additive components:

$$V\left(n\right)={\displaystyle \frac{V_{\mathrm{plant}}}{n}}+V_{\mathrm{env}}$$

(2)

where V_plant is the variance among individual plants within a plot, n is the number of plants per plot and V_env is the environmental variance associated with the plot as a physical position in the field. The essential property of this model is that only the first component is reduced by increasing the number of plants sampled, whereas V_env is invariant with respect to n. The response of residual variance to plot size therefore identifies which component dominates.

To estimate V_plant and V_env, the residuals e_ij of the complete block design were retained and all subsequent computations were performed on them, so that the estimated components describe exclusively the experimental error and are not contaminated by the treatment structure. Plots were then classified by the number of surviving plants. Two classes provided sufficient replication: n = 3 (k = 9 plots; 8 degrees of freedom) and n = 4 (k = 26 plots; 25 degrees of freedom). Within each class the residual variance for n plants per plot ($\widehat{V}\left(n\right)$) was computed as:

$$\widehat{V}\left(n\right)={\displaystyle \frac{\sum {\left(e_{ij}-{\overline{e}}_k\right)}^2}{k-1}}$$

(3)

where e_ij is the experimental error for each observation, and ${\overline{e}}_k$ is the average of e_ij associated with the k plots. The estimates were $\widehat{V}\left(3\right)$ = 0.3945 and $\widehat{V}\left(4\right)=$ 0.1725 for pod yield and $\widehat{V}\left(3\right)$ = 0.0806 and $\widehat{V}\left(4\right)=$ 0.0424 for seed yield. The coefficient of variation can be estimated from these values. For pod yield the residual coefficient of variation fell from 36.11% in three-plant plots (k = 9) to 23.88% in four-plant plots (k = 26), and for seed yield from 37.72% to 27.36%, corresponding to reductions of 33.9% and 27.5%, respectively, achieved by adding a single plant to the observational unit.

V_plant was estimated by subtracting $\widehat{V}\left(3\right)-\widehat{V}\left(4\right)$ to eliminate V_env, and V_env was estimated by difference $\widehat{V}\left(n\right)-{\displaystyle \frac{V_{\mathrm{plant}}}{n}}$. The estimates were V_plant = 2.6642 and V_env = −0.4935 for pod yield and 0.4580 and −0.0721 for seed yield. Assuming V_env = 0, the maximum V(3)/V(4) ratio is 1.3333, but the observed ratios of 2.29 (pod yield) and 1.90 (seed yield) both exceed this ceiling. Therefore, the environmental component is effectively zero, and the experimental error for pod and seed yield is almost entirely of within-plot sampling origin.

The optimum number of plants per plot can be estimated by the modified maximum curvature method of Meier and Lessman [38] applied to Smith’s variance law [39], CV(n) = A n^−b, where CV(n) is the coefficient of variation with n plants per plot, A is the coefficient of variation of the basic unit, and b is the heterogeneity index. The critical plot size was obtained by maximizing the curvature of Smith’s function, yielding Equation (4).

$$X_c={\left[{\displaystyle \frac{A^2{b}^2\left(2b+1\right)}{b+2}}\right]}^{{\displaystyle \frac{1}{2(b+1)}}}$$

(4)

Considering A (n = 4) and b = 0.5, X_c was 11.6 plants for pod yield and 12.9 plants for seed yield; across the interval b = 0.4 to 0.6 the estimates ranged from 9.2 to 13.8 and from 10.3 to 15.3 plants, respectively. Therefore, seed yield consistently required the larger plot, as expected from its higher coefficient of variation. These estimates agree with empirical reports indicating that 14–30 plants per progeny are required to adequately capture the genetic diversity of half-sib families [40,41]. Therefore, to estimate seed yield and H² of seed yield ≥14 plants per plot.

4.3. Effect of Pachyrhizus Extracts

Because the assay did not include colony-forming unit counts, determinations of minimum inhibitory or minimum bactericidal concentration, membrane-integrity assays, or recovery assays after extract removal, the observed responses were interpreted as changes in apparent bacterial multiplication rather than as proof of bactericidal activity.

Antibacterial assays revealed that 0.5% (w/v) yam bean extracts significantly inhibited RS multiplication by 3- to 10-fold at 24 h post-inoculation, whereas 5.0% (w/v) extracts stimulated bacterial growth throughout the evaluation period (6, 12, and 24 h).

After 6 and 12 h of incubation, 0.5% yam bean extract stimulated bacterial proliferation 10-fold and 25-fold, respectively, compared to the bacterial growth control (YPG medium + bacteria, no extract). However, at 24 h, the extracts reduced bacterial growth 10-fold relative to the control. These findings corroborate those of Silva et al. [14], who tested yam bean extracts at various concentrations (0.01%, 0.05%, and 0.5%) against RS and observed that 0.5% extracts provided effective bacterial control at 48 h, confirming the time-dependent antibacterial activity observed in the present study.

This biphasic pattern suggests that water-soluble nutrients in the extract (amino acids, simple sugars, and minerals) initially support bacterial metabolism and growth, while antimicrobial secondary metabolites exert delayed inhibitory effects, possibly due to time-dependent accumulation or metabolic activation.

One of these metabolites is rotenone, which demonstrates sustained cytotoxic activity lasting up to 7–8 days [27]. This extended activity profile suggests that the antibacterial effects may continue to intensify beyond 24 h. Therefore, while 24-h assessments provide valuable initial antibacterial data, extended monitoring at multiple timepoints (48–168 h) is warranted to determine the full spectrum of antimicrobial efficacy of the extract.

These findings indicate that Pachyrhizus extracts could be used to dip the roots of seedlings before transplanting to the field and to treat the plant collar during the first weeks after transplanting, thereby reducing bacterial multiplication.

It was also observed that tetracycline inhibited bacterial multiplication from 6 to 24 h with the same intensity and greater inhibitory capacity than the yam bean extracts. These results confirm the efficiency of tetracycline on RS at doses ranging from 240 to 700 mg L⁻¹ [14,42]. Despite this efficacy, soil application is not recommended because tetracycline significantly disrupts soil microbial communities, with effects observed at concentrations as low as 5 mg kg⁻¹ and increasing substantially at 25–500 mg kg⁻¹ [43].

At 24 h and 0.5% concentration, progenies P23 beige, P14 beige, and P20 beige reduced the bacterial population by up to 90%, 89%, and 88%, respectively, relative to the bacterial growth controls. These findings indicate that these progenies may be promising candidates for the development of biocontrol strategies against RS.

The 5% extract promoted bacterial growth at 6, 12, and 24 h compared with the bacterial growth control (YPG + bacteria, no extract), suggesting that the nutritional effects outweighed the antimicrobial activity at this concentration. Therefore, extended incubation studies (48–192 h) are needed to determine whether antimicrobial effects emerge at later time points, given that rotenone persists for 7 or 8 days [27,28].

4.4. Limitations, Implications and Future Studies

This study has three limitations. (i) The failure to detect genetic variation in seed yield reflects insufficient sampling, as only four plants per experimental unit were evaluated. (ii) The antibacterial activity was assessed against a single RS isolate (phylotype II, FIO104B, from tomato in Iranduba, Amazonas, Brazil). Given the high genetic diversity and recombination within RS, particularly in virulence-related genes [44], these findings require validation against other phylotypes. (iii) The incubation period was limited to 24 h.

These findings indicate that progenies P14, P20, and P23 are promising candidates for breeding programs focused on dual purposes: seed production and biopesticide development. The high genetic variability within progenies suggests that breeding methods for allogamous (cross-pollinating) plants should be applied to Pachyrhizus spp. The 0.5% concentration was more effective than 5%, indicating that lower concentrations can be used in commercial formulations to reduce production costs. The biphasic activity (initial stimulation followed by inhibition) suggests that application timing is critical for maximizing antimicrobial effects.

Pachyrhizus extracts (0.1%) had no significant effect on soil mesofauna after 8 days [45], suggesting a low impact on soil mesofauna, which supports their use by smallholder farmers in the Amazon. However, further testing at other concentrations is needed.

Future studies should estimate the heritability of seed yield using larger plots (≥14 plants per plot) in different Amazonian soils. Further work should also determine the antibacterial efficacy of the extracts over extended periods (48–168 h) to assess the full duration of activity, and evaluate other concentrations with the candidate progenies to model concentration effects. Finally, studies should evaluate the effects of Pachyrhizus extracts applied to seedlings and under field conditions on RS, mesofauna, and plant growth, and quantify rotenoids and other compounds by liquid chromatography–mass spectrometry (LC-MS) and gas chromatography–mass spectrometry (GC-MS).

5. Conclusions

Progenies P14, P20, and P23 are promising candidates for seed production and the development of biopesticides targeting Ralstonia solanacearum. These progenies should be evaluated in experimental designs with larger plot sizes (≥14 plants per plot) for seed yield, following breeding methods for half-sib families.

Broad-sense heritability for pod traits and seed size was high, suggesting that genetic gains can be achieved through selection for these traits.

The 0.5% (w/v) extracts exhibited time-dependent activity, stimulating bacterial multiplication at earlier time points but reducing R. solanacearum growth at 24 h after treatment. Conversely, the 5% (w/v) extracts promoted bacterial growth at 24 h, highlighting the need for assessments beyond 24 h to evaluate delayed inhibitory effects.