The study was carried out with the tropical peanut crop (Arachis hypogaea L.). The cultivar IAC 505 (Grupo Virgínia) was used as a model, which is characterized as being high oleic, generating plants with a spreading habit and a cycle of more than 130 days. The seeds used in the study were COPERCANA (https://copercana.com.br/), located in Sertãozinho in the western region of the state of São Paulo, Brazil.

The data was analyzed using a completely randomized design with five development stages (R5, R6, R7, R8, and R9) as the variation factor, with five repetitions (n=25). The data was subjected to ANOVA after checking the assumptions of normality and homoscedasticity using the Shapiro-Wilk and Bartlett tests. The averages were compared using the Tukey’s test with a significance level of 5% ( Supplementary Table 5 ). The longevity data (p50) was obtained from the generalized linear model with the Cauchy function at 5% significance level. R software was used for the analyses (R Core Team, 2024).

The data observed for the seed physiological and health quality variables presented so far were subjected to principal components analysis (PCA). Furthermore, Permutational Multivariate Analysis of Variance - PERMANOVA One-way (Canoco 5 software) was used to test the similarity of data behavior (Bray-Curtis similarity index) in two groups. The groups were: i) early seed stages (R5 and R6); and ii) late seed stages (R7, R8 and R9). The significance of the groups was tested at a level of 1% (Anderson, 2017).

The Random Forest, a machine learning algorithm (Brieuc et al., 2018), was used to model the relationship between seed development stages and 22 seed quality variables. When applied to predicting seed development stages, the Mean Decrease Gini (MDG) method quantifies the extent to which each predictor variable contributes to the accuracy of the model. Variables with higher MDG values are considered more important in predicting seed development stages. Correlation analysis was calculated using the Spearman method. R software was used for the analyses (R Core Team, 2024).

The notable morphological changes of the fruits in the process of maturation ( Figures 2L–P ) occur in parallel with the physical changes of the seeds ( Figure 3A ). As regards water content, a high degree of seed moisture was observed at the beginning of development for stages R5 (58%) and R6 (47%) ( Supplementary Table 5 ). From stage R7 the water content of 40% rapidly decreased to 32% in stage R8 and again to 30% in stage R9. The water dynamics of the seeds were inverse to the accumulation of dry weight ( Figure 3A ), which increased significantly until the R7 stage (~7g/10 seeds). This point defined the maturity of the seed mass or the maximum filling of reserves. In general, seeds from late stages (R7, R8 and R9) presented different water contents and similar and stable dry weight ( Figure 3A , Supplementary Table 5 ).

Germination of fresh seeds occurred in the initial stages R5 and R6 with a high degree of humidity. However, seeds at the R5 stage were highly sensitive to water loss and did not germinate after drying. Interestingly, even with low germination (6.4%), the first signs of desiccation tolerance were only observed in seeds at the R6 stage. Desiccation tolerance was fully acquired at the R7 stage with germination at around 90%. In the following stages, R8 and R9, the results remained stable ( Figure 3B ). In general, late-stage seeds tolerate desiccation and do not lose viability after the drying step.

At the beginning of the development process, seeds from the initial stages (R5 and R6) did not germinate ( Figure 3C ). This did not result in t50 values, which was notable evidence of low seed vigor. It was interesting to note that by the time desiccation tolerance was fully acquired at the R7 stage ( Figure 3B ), t50 began to be accounted for ( Figure 3C ). Overall, it took about 50 h for half of the seeds to fully germinate at the R8 and R9 stages. In relation to total germination, seed performance increased significantly between development stages ( Supplementary Table 5 ). The R9 stage seeds had maximum germination in a short evaluation period (72 h) ( Figure 3C ).

The few seeds that germinated at the R6 stage gave rise to seedlings with poorly developed shoots and roots over the 7 days of evaluation ( Figure 3D ). Seeds from the late stages (R7, R8 and R9) produced seedlings with larger roots (~4 cm and 5 cm) and higher dry mass accumulation in the shoot (~16 mg and 19 mg) and root (~22 mg and 31 mg) ( Supplementary Table 5 ). Seeds at the R8 stage were able to produce seedlings with larger shoots, similar to those found in the R9 stage (around 1.2 cm). In general, seeds from late stages were more efficient in transferring reserves to seedlings, which was evidenced by well-developed shoots and roots ( Figures 3D, E ).

Under uncontrolled environmental conditions, seeds from the early stages (R5 and R6) occasionally gave rise to seedlings with reduced emergence capacity (3% and 12%, respectively). Notably, there was a higher percentage of seedlings that emerged at the R7 stage (~60%) and the results remained statistically similar at the R8 and R9 stages (~74% and ~80%, respectively). It was interesting to note that only seeds from stages R7, R8 and R9 showed higher emergence speed ( Figure 3F ). This constitutes a characteristic that expresses the superior potential of late-stage seeds to rapidly establish seedlings under broad environmental conditions ( Figure 3F ).

Seeds up to the R6 stage aged at a high temperature (42°C for 72 h) germinated very little (0.8%). Germination increased considerably in the following stages, reaching 72.8% in the R7 stage and 98.8% in the R8 stage, which had a maximum performance similar to the R9 stage ( Figure 3G ). After 1 year of storage (10°C/55% RH), seeds at early stages (R5 and R6) gave rise to approximately 8% to 26% plants after 40 days in the field ( Supplementary Table 5 ). The results were notably higher in the R7, R8 and R9 stages, where the percentage of established plants was around 70 to 89% ( Figure 3G ). This means that seeds from late stages were better able to guarantee the initial formation of a new peanut production cycle even after prolonged storage (one year) ( Figure 3G ).

Interestingly, only R9 stage seeds had the highest germination (70%) after 71 days stored under conditions of heat stress and high humidity (35°C/75% RH). Physiologically, maximum seed quality was acquired at the R9 stage ( Figure 3H ). In general, this ability to germinate even under long exposure under these conditions increased as the seeds entered late maturation from the R7 stage onwards ( Supplementary Table 5 ). Seeds from late stages (R7, R8 and R9) took longer to lose 50% viability (p50) during storage. This means they had superior longevity ( Figure 3H ). In the context of peanut maturation, seeds at stages R7, R8 and R9 have superior storage potential and post-harvest vigor ( Figure 3H ).

Under controlled conditions (temperature and humidity) prior to storage, the seeds generated seedlings, during 7 days of evaluation, only from the R6 stage (< 1 cm) onwards. The total length of the seedlings increased notably at the R7 stage (~5 cm). Seeds from the R8 stage gave rise to seedlings with a maximum size (~6 cm) similar to the R9 stage ( Figure 4 ). Interestingly, after 1 year of storage (10°C/55 RH), the seedling size result followed the same behavior as the percentage of normal seedlings (without defects or damage) in uncontrolled field conditions. For example, although some seeds at the R5 stage eventually germinated, they were not considered normal (absence of a developed root). The formation of normal seedlings was 15% at the R6 stage and 70% at the R7 stage. Notably, the maximum percentage of normal seedlings (80%) was obtained for seeds at the R8 stage, similar to R9 ( Figure 5 ). This means that the seeds acquired maximum vigor for the establishment of seedlings under broad environmental conditions at the R8 stage, and that this vigor did not change after storage.

The presence of pathogens decreased as the seeds entered the late maturation phase from the R7 stage onwards ( Figure 6 ). Seeds at stages R5 and R6 showed reduced germination (0 and 21.4%, respectively) and a high percentage of bacteria (65.7% and 37.1%, respectively). Regarding fungi: i) in the R5 stage, around 18% and 25% of Aspergillus ssp and Penicillium ssp were observed, respectively; and ii) in the R6 stage, this respective percentage was around 17% and 20% ( Figure 7 ). Interestingly, in the late stages there was no presence of bacteria and seed germination was around 85% to 90%, although they were in an environment contaminated with pathogens ( Supplementary Table 5 ). Furthermore, the seeds showed low susceptibility to Aspergillus ssp (~1% to 3%) and Penicillium ssp (<2%) ( Supplementary Table 5 ). In general, seeds from stages R7, R8 and R9 showed superior resilience to the proliferation of pathogenic microorganisms ( Figure 6 ).

The results of the entire research were divided into two distinct groups (PERMANOVA, p value < 0.001) by the similarity index (Bray-Curtis) and with 88.6% of variance in PCA₁ and 3.2% in PCA₂ ( Figure 7 ). In the first group, the highest values of germination, desiccation tolerance, vigor and seed longevity were strongly related to the late stages (R7, R8 and R9). On the other hand, the second group consists of the initial seed stages (R5 and R6). The main characteristics of these stages were: i) data associated with high contamination by Aspergillus, Penicillium and bacteria (Bacillus sp.); ii) high water content ( Figure 7 ) in the case of seeds before drying ( Figure 3A ); and iii) none of the related physiological quality parameters. In general, through the PCA, it was evident that the R7, R8 and R9 stages should be prioritized in monitoring the peanut harvest aiming for superior seed quality.

The ranking of the observed data, using the Random Forest tool, made it possible to detect in order of importance the best variables to discriminate the development stages of peanut seeds. The most important variables were: 1°) Normal Seedlings; 2°) Water content; and 3°) Longevity ( Figure 8A ). For the first highlighted variable: i) the percentage of normal seedlings was positively correlated with dry weight (r= 0.9) and with the physiological quality attributes of the seed (0.67 ≥ r ≤ 0.91); and ii) the correlation was negative for contamination by fungi and bacteria (pathogens). Subsequently, except in the case of pathogens (0.65 ≥ r ≤ 0.87), the higher the water content of the seed, the lower its mass, germination capacity, desiccation tolerance, vigor, and longevity (-0.76 ≤ r ≥ -0.98). Finally, interestingly, superior seed longevity was negatively correlated with water content (r < -0.98) and pathogen contamination (-0.62 ≤ r ≥ -0.89). In general, longevity was directly proportional to the physiological quality of the seeds, which enabled normal seedling production months after initial storage (0.78 ≥ r ≤ 0.9) ( Figure 8B ).

The tropical peanut maturity scale summarized all research results ( Figure 9 ). Seed quality was expressed based on the most important variable for classifying the stages of seed development obtained in the Random Forest analysis (machine learning): Normal seedlings ( Figures 5 , 8A ). The information allows monitoring the potential of newly harvested peanut seeds to form seedlings in the field after the storage phase. Thus, each stage of development classified at the time of harvest is related to a future percentage of seed quality. This way, it is possible to understand what the seed’s likely performance will be in the following crop season (new production cycle). For example: at stage R5 the seeds do not have the potential to generate plants in the field, therefore, they have 0% physiological quality; at stage R6 the seeds present around 20% quality; in stage R7, around 70% quality; in stages R8 and R9 around 90% quality ( Figure 9 ).

Water dynamics govern physical and biochemical changes in legume seeds (Righetti et al., 2015; Lima et al., 2017; Okada et al., 2021). Changes in water content occur in parallel with cellular events during seed development (Aldana et al., 1972; Bewley et al., 2013). For example, cell division and expansion in the embryogenesis phase depend on high humidity for the synthesis of nucleic acids and the formation of seed tissues (Setter and Flannigan, 2001; Armenta-Medina et al., 2021). The moisture required for this process after fertilization (stages R1 to R4) reflects the higher seed water content in the following development stages ( Figure 2 ). In the beginning of the filling phase (stage R5), the flow of assimilates from the plant to seeds occurs under high water demand (Copeland and McDonald, 2001; Okada et al., 2021). This means that water contents of around 60 to 40% are critical until the R7 stage ( Figure 3A ) for the accumulation of reserves such as sugars, oils, and proteins (Pattee et al., 1974). These reserves are crucial for cellular protection in face of sudden variations of embryonic humidity that occur until the end of late maturation (Sattler et al., 2004; Chatelain et al., 2012; Lima et al., 2017). Stability in reserve accumulation is achieved at water contents of around 40 to 30% in the late stages ( Figures 2 , 3 ). Water dynamics can help identify at which reproductive stage most seeds in field are ( Figure 8 ). Therefore, with this information it is possible to harvest more mature seeds with improved chemical composition.

The ability to survive desiccation (losing more than 90% of water) is a characteristic of angiosperm seeds and other organisms (Gaff and Oliver, 2013; Oliver et al., 2020). The natural desiccation of peanut seeds in the soil and subsequent artificial drying to 10% humidity ( Figures 3A, B ) cause structural changes in the cellular membrane system (Hoekstra et al., 2001). Part of the reserves accumulated in the filling phase play an essential protective role when under these circumstances (Smolikova et al., 2021). Among the main seed reserves we have: i) LEA proteins (Chatelain et al., 2012); ii) HSP proteins (Wehmeyer and Vierling, 2000); and iii) sugars from the raffinose family (Hell et al., 2019). A system involving these compounds preserves the protein structure and imparts molecular stability as water is removed from the developing seed ( Figures 3A, B ) (Buitink and Leprince, 2004). This system allows the late-stage seeds to remain viable after desiccation and to germinate when rehydrated ( Figure 3B ) (Verdier et al., 2013; Lima et al., 2017). This means that ending the peanut cycle with a high proportion of immature stages (i.e., R5 and R6) can cause seed death during the drying step ( Figure 3B ). The solution to optimizing the acquisition of desiccation tolerance is to establish a judicious harvest based on development stages (Lima et al., 2017; Moreno et al., 2022). This increases the chances of obtaining peanut seeds with superior viability.

In late maturation, when the reserves are practically stabilized, the seeds complete the acquisition of vigor and longevity under low levels of humidity (Basso et al., 2018; Okada et al., 2021). The “long life” mechanisms of leguminous seeds that preserve their biodiversity are then established ( Figure 3H ) (Verdier et al., 2013; Righetti et al., 2015). In particular, the formation of the vitreous cytoplasm, the action of antioxidants and the repair mechanisms in cells, ensure the metabolic stability of dry seeds during storage (Buitink and Leprince, 2004, 2008; Zinsmeister et al., 2020). These molecular and chemical defenses (i.e., tocopherols) delay cellular aging after desiccation (Finch-Savage and Bassel, 2016; Ramtekey et al., 2022). Interestingly, the protective capacity described for acquiring longevity is exclusive to late stages ( Figure 3H ) (Lima et al., 2017; Zhou et al., 2019). In them, vigor is preserved as much as possible so that the seeds germinate and produce seedlings months after the harvest ( Figures 4 , 5 ) (Sano et al., 2016; Reed et al., 2022). However, harvesting seeds with these benefits requires assiduous monitoring of late maturation (Leprince et al., 2017; Basso et al., 2018). It is in this context that the peanut harvest indicators linked to the development stages presented have great practical utility ( Figure 2 ) (Williams and Drexler, 1981; Boote, 1982). Through the joint utilization of these well-established harvest indicators, late maturation can be accurately monitored in tropical fields enabling the obtention of peanut seeds with maximum physiological quality.

Seed maturity determines its bioprotection ( Figure 6 ) (Commey et al., 2021; Mendu et al., 2022). Seeds accumulate phenolic compounds, flavonoids, carotenoids, and other antioxidant agents as they develop (Wan et al., 2016; Zinsmeister et al., 2020). In addition, they acquire a thick cell wall to protect against biotic stresses both physically and chemically (Polo et al., 2020; Ninkuu et al., 2023). All these defense agents create a natural barrier against pathogen infection, especially Aspergillus ssp, a fungus of global concern in peanuts ( Figure 6B ) (Kumar et al., 2017; Miura et al., 2019; Norlia et al., 2019; Commey et al., 2021). Taking this into account, it is understandable that seeds from the more immature stages still have deficient protection, as they have not completed their development ( Figure 6 ) (Smýkal et al., 2014; Okada et al., 2021). In the late stages, most of the immune response configurations are molecularly programmed (Lima et al., 2017; Nayak et al., 2017). Thus, the biological meaning of seed maturity boils down to having resources to survive in the dry state, protect itself for long periods, germinate and provide new plant cycles in time and space (Sano et al., 2016; Song et al., 2022). In tropical agriculture, bioprotection benefits can be obtained with the harvesting guidelines proposed here ( Supplementary Tables 3 , 4 ).

Harvesting seeds that are resilient to environmental adversities is a highly relevant strategy given the numerous stresses that crops are subjected to in the field (Sun et al., 2019; Reed et al., 2022). Peanut seeds go through post-harvest stages that are very prone to oxidative stress (Barbosa et al., 2014; Groot, 2022). This stress can be mitigated with seed maturity ( Figures 5 , 6 ). In tropical peanut-producing nations, examples such stresses are: i) drying in the field exposed to the weather; ii) low availability of cold storage structures; and iii) sowing under extreme variations in temperature and soil moisture (Nakagawa and Rosolem, 2011; Sun et al., 2019; USDA, 2023). Only seeds from late stages (high seed quality) are able to tolerate biotic and abiotic stresses with maximum efficiency ( Figure 7 ) (Sano et al., 2016; Ellis, 2019; Zinsmeister et al., 2020). This is because a significant part of the mechanisms that mitigate the deterioration of seeds in the dry state are dependent on the moment they were harvested (Leprince et al., 2017; Ramtekey et al., 2022). Therefore, the maturation scale, when assisting in harvest programming, is an opportunity for technological advancement in the peanut production sector in different countries ( Figure 9 ) (USDA, 2023).