The Seed–Seedling Transition in Commercial Soybean Cultivars with the Presence of Greenish Seeds in the Sample: A Perspective from Classical Genetic Parameters

Abstract

Chlorophyll retention reduces the quality of soybean grain and seeds, but little is known about the relationship between this anomaly and the seed–seedling transition or classical genetic breeding. Taking this into account, we provide a new perspective about how greenish seeds are associated with genetic material. For that, we used seed–seedling transition measurements of nine commercial cultivars cultivated by different seed farmers. Our findings demonstrated that: (i) The anomaly of greening in soybean seeds seems to be more expressed as a function of thermal variation than rainfall; (ii) There are commercial cultivars with distinct sensitivity to the greening phenotype, indicating potential for classical breeding for the species; (iii) The seed–seedling transition is significantly modified by the presence of greenish seeds, and the diagnosis of their presence can be anticipated by the analysis of embryo protrusion.

1. Introduction

Although reports on greenish seeds in soybean lots (also called ‘green soybean’ or ‘green seed’) are relatively recent [1,2,3,4], the problem has drawn attention due to the negative impact on the commercialization of the species [3,5,6,7]. This impact has increased over the last few years, stimulating studies concerned with understanding the phenomenon from a physiological and molecular point of view [2,3,5,8]. What is known from the physiological point of view so far is that the increase in the number of greenish seeds (above 9%) in the sample can cause a reduction in physiological quality, measured by the number of normal seedlings developed [2,6]. It is also known that the phenomenon is expressed mainly due to abiotic (specifically, water and thermal stresses) and biotic stresses (such as Asian soybean rust) affecting the mother plant at the end of seed development (i.e., after R5, Fehr and Caviness 1977 sense) [4]. Because it is a green pigmentation anomaly, the phenomenon has been associated with chlorophyll fluorescence peaks and, therefore, has been attributed to failures in plastid degradation metabolism [8]. However, the impact of the genotype–environment interaction on the occurrence of this anomaly is not yet known. The impact of this anomaly on the seed–seedling transition per se is also little understood, considering only germination or post-germination. This is particularly intriguing, since a link between the processes can speed up the diagnosis of the anomaly.

For optimum soybean seed production, it is necessary to have average temperatures < 23.5 °C at the time of histodifferentiation, maturation and harvest [9]. Thus, a high temperature could be the stress that most contributes to inflating the number of greenish seeds of a sample, especially when associated with water deficit at maturation [10]. This explanation was confirmed in practice with data from the 2011/2012 crop, in which soybean producers from the South of Brazil (Rio Grande do Sul, West of Santa Catarina and Southwest and West of Paraná) had great losses, with the overwhelming presence of the anomaly due to high temperatures and low precipitation in the filling phase [11]. In general, this situation was also observed in other large soybean-producing regions in the country. The unusually large presence of the greenish seed anomaly was attributed to the phenomenon “La Niña”. This and other climatic phenomena associated with global climate change have reinforced the relevance of studies on agroclimatic zoning [12], especially concerning soybeans [9,13]. These studies have been carried out with the purpose of delimiting the regions with the lowest climatic risk to seed production considering altitude, temperature and rainfall, emphasizing the need to understand the genotype–environment interaction regarding the seed–seedling transition.

The link between embryo protrusion and immediate post-germination (seedling development) can provide effective biometric markers not only for the production chain of the species but also for genetic improvement [14,15]. This instigated the following questions: (i) Is there differentiated sensitivity in soybean cultivars regarding greenish seeds? (ii) Will the genotype effect manifest itself in greenish seeds? (iii) What is the genetic contribution to the seed–seedling transition when there are greenish seeds in the sample? To answer these questions, we used classical germination and seedling development measurements in post-germination. The use of different classical germination measurements is still restricted in studies on cultivated species such as soybean, but many studies have used these measurements to determine basic aspects of seed biology [14,16]. The studies of genetic and environmental effects, although already mentioned for characteristics such as the percentage of emergence and germinability, have only recently been detailed, using native species as a biological model [14]. In any case, these studies should be intensified not only to make algorithms freely available but also to increase the selection speed of characteristics desired by the breeder.

Overlapping these sets of classical measurements can help us understand the contribution of the genotype–environment interaction, which results in the undesired effect of soybean greenish seeds. We expect to point out ways to clarify the reasons why this phenomenon reduces the production of normal seedlings of the species so much. By the way, from this, it will also be possible to interconnect basic and practical aspects of seed science. Thus, the hypotheses arise that (i) mathematical dissociations between thermal and rainfall amplitudes can signal paths for agroclimatic zoning for soybean seed production; (ii) the effects of greening on the quality of commercial seed lots are determined by significant genotype–environment interaction, where germination is more sensitive to this interaction than the production of normal seedlings. The objective of this study is to characterize the physiology in soybean seed lots with different percentages of greenish seeds using cultivars produced in varied regions and environmental conditions and, thus, to elucidate how and why the increase in this anomaly impacts the development of seedlings in commercial lots.

2. Materials and Methods

2.1. Seed Production and Sample Processing

The seeds used in the test were produced in commercial fields in different Brazilian producing regions during the 2014/2015 harvest, under the responsibility of Embrapa Soja/Londrina-PR. We used the lots of nine cultivars (NS 5959 IPRO, 97Y07, 8473 RSF, 71MF00 RR, SYN1059 RR, NA 5909 RG, W 799 RR, 6266 RSF IPRO and M6972 IPRO), chosen for being the most used by producers in the regions of Goiás, Mato Grosso, Paraná and Santa Catarina.

At harvest, seed samples of each cultivar were sent to the laboratory for the screening of the expression of the greening phenotype (anomaly) and for the measurement of the water content. The percentage of greenish seeds was visually determined using four replicates with 100 seeds each. In this case, the criterion for determining greenish seeds was the presence of any green pigmentation on the seed, a procedure adopted according to the routine of technicians in the area. The water content of the seeds was determined by the oven method at 105 ± 3 °C for 24 h, with two replications of 25 seeds, whose results were expressed as a percentage of water [17]. In order to study the maximum and minimum expression of greening in the commercial material, the lots that presented a greater or lesser expression of the anomaly were classified as highly or little greenish. Thus, among 37 seed lots of different cultivars obtained in the harvest, two lots of each cultivar were selected with the most extreme values in the expression of the anomaly.

2.2. Seed–Seedling Transition

To determine how the degree of greening in a sample alters the genetic pattern of the sowing material, the seed–seedling transition was evaluated by germination analysis (root protrusion), phenotyping and seedling growth in post-germination.

2.2.1. Seed Germination

The experiment was conducted in a completely randomized design, with four replications of 25 seeds each. The sowing was carried out on filter paper with water added, equivalent to 2.5 mL for each mg of its dry mass. The experiment was conducted in a germination chamber at 25 °C (with continuous light). The germination criterion adopted was the protrusion of the radicle. The final reading occurred after no germination was observed in an interval of 8 days. The remaining seeds, after this period, had coalescing tissues.

The evaluated characteristics were calculated according to [16,18], including the germinability (G), time for first germination (t₀), mean germination time ($\overline{t}$ or MGT), median germination time (t₅₀), time for last germination (t_l), coefficient of variation of germination time (CV_t), mean germination rate ($\overline{v}$ or MGR), uncertainty (U) and synchrony (Z).

2.2.2. Phenotyping of Seedlings

The analysis followed the protocol described according to [17]. For this, four replicates with 25 seeds each were sown on filter paper substrates, following the same protocol described above for seed germination. The evaluations of the test were performed on the fifth and eighth day after sowing, classifying the seedlings by visual evaluations, as described by [17]. The percentage of normal and abnormal seedlings and dead seeds was determined. There were no records of dormant seeds. In parallel to the test, seedlings that presented early development, counted on the fifth day after sowing, were considered vigorous [17].

2.2.3. Seedling Growth

Four replicates of 10 seeds per genotype were used, following the same protocol described above for seed germination. The seeds were arranged in a line drawn longitudinally in the upper third of the wet substrate paper, with the hilum facing the lower part of the paper, aiming to guide the seedling growth in the straightest way possible [19]. The substrates were rolled up and kept vertically in a germination chamber at 25 °C, in the absence of light. Seven days after the beginning of the test, the average length of the shoot and of the primary root was measured for the normal seedlings, and the dead seeds were counted. The results were expressed in centimeters. After the evaluation of the length, the shoot was separated from the roots, identified and kept in an air circulation oven at a temperature of 60 °C until weight stabilization. After that, the material was weighed on a precision scale to calculate the average dry mass.

2.3. Hierarchical Model or Nested Analysis

The seed–seedling transition was studied using a hierarchical model (also called nested analysis) for two levels, where the cultivar was the fixed factor and the expression of the greening anomaly was the random factor, according to:

y_ijk = μ + α_i + β_j(i) + e_k(ij),

for i = 1, 2, 3, …, a; j = 1, 2, 3, …, b_i and _k

= 1, 2, 3, …, r_ij,

where μ is the overall average, α_i is the effect of genotype i, β_j(i) is the greenish effect on seed sample j within genotype i, e_k(ij) is the effect associated with the residual, a is the number of samples with a distinct expression of the anomaly, b_i is the number of genotypes, r_ij is the number of repetitions of i and j is the number of samples with different expressions of the greenish anomaly [14]. The part of the total variance attributed to the variance in the anomaly within a genotype was calculated by mean squares.

2.4. Genetic Parameters Associated with Seed–Seedling Transition Characterization

The estimation of genetic parameters for the characterization of the seed–seedling transition was processed with the GENES software [20]. The estimate of genetic variance was given by ${\widehat{\mathsf{\sigma }}}_{\mathrm{g}}^2=({\mathsf{\sigma }}_{\mathrm{B}/\mathrm{A}}^2-{\mathsf{\sigma }}^2)/\mathrm{r}$, and that of phenotypic variance was given by ${\widehat{\mathsf{\sigma }}}_{\mathrm{f}}^2={\mathsf{\sigma }}_{\mathrm{B}/\mathrm{A}}^2/\mathrm{r}$, where ${\mathsf{\sigma }}_{\mathrm{B}/\mathrm{A}}^2$ is the variance of genotypes within the degree of greening as a function of the environment, ${\mathsf{\sigma }}_{\mathrm{B}/\mathrm{A}}^2$ is the residual variance and r is the number of replications. The sensu stricto heritability [21] was calculated by ${\mathrm{h}}^2={\widehat{\mathsf{\sigma }}}_{\mathrm{g}}^2/{\widehat{\mathsf{\sigma }}}_{\mathrm{f}}^2$, where ${\widehat{\mathsf{\sigma }}}_{\mathrm{g}}^2$ is the genetic variance and ${\widehat{\mathsf{\sigma }}}_{\mathrm{f}}^2$ is the phenotypic variance; the coefficient of genetic variation was given by ${\mathrm{C}\mathrm{V}}_{\mathrm{g}}=\raisebox{1ex}{$100\sqrt{{\widehat{\mathsf{\sigma }}}_{\mathrm{g}}^2}$}\!\left/ \!\raisebox{-1ex}{$\widehat{\mathrm{m}}$}\right.$, where ${\widehat{\mathsf{\sigma }}}_{\mathrm{g}}^2$ is the genetic variance and $\widehat{\mathrm{m}}$ is the average of genotypes in samples with different degrees of greening. The ratio between the coefficient of genetic and environmental variation was also calculated (CVg/CVe). Ratios greater than one indicate that genetic variation predominates in relation to environmental variation.

2.5. Statistical Analysis

The assumptions of the model were verified by the Shapiro–Wilk test to determine the normal distribution of the residues and by the Levene test for homoscedasticity, both at 0.01 significance. Whenever the data transformation improved the probability of the tests, even if the assumptions were not met, parametric statistics were chosen [22]. The comparison of the significance of the model was made by Snedecor’s F test, and the multiple comparisons of the means were made possible by the Scott–Knott test, both at 0.05 significance. Genetic parameterization was performed using the GENES software [20]. Surface maps were generated by considering the degree of seed greening in the samples under study and the climatic aspects of the producing regions (temperature and rainfall). Principal Component Analysis (PCA) and hierarchical clustering (groups determined by cuts in the dendrogram at 10% of the Minkowski distance) were performed to determine patterns of variances and seed–seedling transition structures. These patterns were confirmed by means of correlation for character residues of the seed–seedling transition, performed by the spearman correlation by pairs. From the significant coefficients of Spearman’s correlation, a heatmap of the correlation matrix was generated. These analyses were processed at 0.05 significance, with the aid of routines coded in the R project (https://cran.r-project.org/, accessed on 1 January 2023). The interpretation for correlation was carried out by using the perspective of Davis [23], in which only r ≥ 0.7 is a substantial correlation.

3. Results

3.1. Seed Production and Maternal Environment

During seed production, the increase in variation between the average of the maximum and minimum daily temperature of the environment was associated with a marked increase in the variation of the daily mean rainfall amplitude (Figure 1A). However, the highest proportion of greenish seeds (50% of seeds in a sample) was associated with variations in the daily average between 10 and 12 mm in rainfall amplitude and variations in the mean between 1.0 and 1.5 °C of the daily thermal amplitude in the maternal environment (Figure 1B). It should also be noted that with variations in the mean between 16 and 18 mm in the rainfall amplitude in environments with an average variation of up to 0.5 °C in the thermal amplitude, there were up to 40% of greenish seeds in the sample, whereas with variations between 8 and 10 mm at 1.5 °C or 14 mm at 0.5 °C, there were smaller proportions of greenish seeds (between 0 and 10%). This demonstrates that thermal amplitude is an important environmental trait for the development of greenish seeds in soybean cultivars.

3.2. Seed–Seedling Transition and Classical Genetic Parameters

The soybean seeds analyzed showed similarity in water content (WC = 9.31% ± 0.22%; mean ± SD) at the beginning of the assays, due to the wide intraspecific variation in the samples of some cultivars (see CI in Figure 2). This ensured that the observed physiological effects are attributed to the genotype and/or the greater presence of greenish seeds in the sample. The relative frequency of germination, in turn, is polymodal, with cultivar-specific peculiarities, especially when there are greenish seeds in the sample (Figure 3). Except for the overlap of germination events (0.12 ≤ Z ≤ 0.25), similar among genotypes even when the sample had more greenish seeds, the germination pattern was modified, but the magnitude was dependent on the cultivar (

Table 1. Values of statistics, probability (P) and genetic parameters of seed–seedling transition measurements in Glycine max (L.) Merr. samples with different proportions of greenish seeds: aspects of the germination process.

Statistics							Characters
			tf (h)		$\overline{\mathit{t}}$		tl (h)		T50 (h)		G (%)		CVt (%)		U (bit)		Z		Rate (Embryo Protrusion h−1)		$\overline{\mathit{v}}$ (h−1)
ANOVA Assumptions		W (P)	0.84 (0.01)		0.97 (0.16)		0.88 (0.01)		0.94 (0.01)		0.86 (0.01)		0.98 (0.45)		0.97 (0.16)		0.87 (0.01)		0.97 (0.16)		0.97 (0.08)
		1F (P)	2.50 (0.01)		3.37 (0.05)		5.50 (0.01)		3.83 (0.01)		6.30 (0.01)		2.20 (0.02)		5.10 (0.01)		6.32 (0.01)		3.42 (0.01)		4.34 (0.01)
Model	Source of Variation	DF	MS	2F	MS	2F	MS	2F	MS	2F	MS	2F	MS	2F	MS	2F	MS	2F	MS	2F	MS	2F
Nested (Hierarchical) ANOVA	Genotype	8	654.00	4.16	689.37	14.27	1404.67	5.85	438.12	4.61	800.96	6.90	27.82	1.05	0.14	1.49	0.004	1.21 ns	800.96	6.90	2.0 × 10−5	10.90
	Greenish seeds (Genotype)	9	794.67	5.05 **	511.62	10.59 **	944.00	3.93 **	294.95	2.82 **	1434.96	12.35 **	88.15	3.32 **	0.26	2.81 **	0.010	1.72 ns	1434.96	12.35 **	1.2 × 10−5	6.53 **
	Error	54	157.33		98.31		240.00		104.71		116.15		26.57		0.09		0.0036		116.15		1.8 × 10−5
		CV (%)	29.40		8.93		13.46		14.91		13.46		19.17		12.85		33.15		13.46		10.14
Genetic Parameters	h2 =		0.76		0.93		0.83		0.96		0.86		0.05		0.33		0.17		0.86		0.91
	CVg (%) =		21.32		13.27		12.10		23.50		13.34		1.69		3.67		6.15		13.34		13.02
	CVg/CVe =		0.72		1.48		0.90		2.04		0.99		0.09		0.28		0.19		0.99		1.28
	r (%) =		34.48		68.86		44.72		80.72		49.56		0.78,		7.53		3.33		49.56		62.27

W: statistic of the Shapiro–Wilk test for residual normality (p ≥ 0.01); ¹F: statistic of the Levene test for homogeneity of variances (p ≥ 0.01); DF: Degrees of Freedom; MS: Mean Square; ²F: statistic of the Snedecor test (Nested ANOVA; p < 0.05); t_f: time to first germination; $\overline{t}$: mean germination time or tm; tl; time to last germination; CV_t: coefficient of variation of the germination time; T50: median germination time; G: germinability U: uncertainty of the germination process; Z: synchronization index; Rate: Maguire’s rate or VE;$\overline{v}$: mean germination rate. The statistical analyses for characters were based on transformed data by √x or, in case of percentage data, by arcsine $\sqrt{x/100}$. CV: Coefficient of Variation (experimental ANOVA). h²: Broad-sense heritability; CVg: coefficient of genetic variation; CVg/CVe: ratio between the coefficient of genetic and environment variation; r: intraclass correlation. ** p ≤ 0.01; ^ns: no significance.

). The uniformity (CV_t) and predictability (U) of embryo protrusion presented low heritability (h² ≤ 0.33). These measurements were also the ones with the lowest genotypic variability (CVg ≤ 6.15%), reinforcing the role of the environment in the expression of the greenish phenotypic variation. This was confirmed by the low CVg/CVe ratio (≤ 0.28) and the low degree of association between the genotype and the presence of greenish seeds in the sample (r ≤ 7.53%; Table 1). The time for the first germination of soybean seeds, on the other hand, proved to be a potential measurement for selection gain, given the high heritability (h² = 0.76) associated with high genotypic variation (CVg = 34.48%) (Table 1). Greater sensitivity to the production environment (CVg/CVe = 0.72) and a correlation of 34.48% between genotypic effects and the greater presence of greenish seeds reinforce this idea (Table 1). The other characteristics of the germination process, in general, present a prevailing genetic process in relation to the presence of greenish seeds in the sample (h² ≥ 0.83; CVg ≥ 12.10%; 0.90% ≤ CVg/CVe ≤ 2.04; r ≥ 44.72%), highlighting the high heritability associated with T50, which proved to be a prevalently genotypic characteristic (h² = 0.96; CVg = 23.50%; CVg/CVe = 2.04; r = 80.72%).

The highest proportion of greenish seeds in a sample significantly affected all protrusion measurements of only one of the nine cultivars analyzed (Figure 4). The first germination of the sample and MGT were higher for two of the nine cultivars analyzed and lower for another cultivar (Figure 4). In addition, the highest proportion of greenish seeds did not always culminate in higher values of T₅₀ (Figure 4). In general, embryo protrusion started 28 h after sowing (HAS). However, the first protrusion occurred at 84 HAS when the proportion of greenish seeds in the sample was higher (t_f; Figure 4). The last germination occurred between 88 and 152 HAS, it being lower in samples with fewer greenish seeds (t_l; Figure 4). The peak of germination events occurred between 60 HAS in samples with the lowest number of greenish seeds and 109 HAS in those with a higher number (MGT; Figure 4). Some cultivars took up to 54 HAS to have 50% germination when there was a lower number of greenish seeds, while others needed up to 91 HAS (T₅₀; Figure 4).

Two cultivars had reduced germinability by approximately 10%, one by 50% and the other by 74% when there was a higher number of greenish seeds in the sample (Figure 4). Only two of these cultivars had no change in germination uniformity associated with the proportion of greenish seeds (CV_t; Figure 1). Higher proportions of greenish seeds in the sample, similarly, did not always reduce the predictability of germination; on the contrary, for three cultivars, the uncertainty decreased (U; Figure 4). This logic was also observed for germination velocity or MGR (Figure 1). Once again, cultivar 1, with extreme values for samples with a lower (MGR = 0.1433 h) and a higher (MGR = 0.0093 h) proportion of greenish seeds, stood out for its sensitivity to the physiological anomaly (Figure 4). This sensitivity was also ratified by means of the daily frequency of germination (Rate), where samples with a higher number of greenish seeds presented 0.3323 embryo protrusions per hour⁻¹ (Figure 4).

Moreover, the genotypic characteristics of soybean seedling phenotyping were also affected by the higher presence of greenish seeds in the samples, except for abnormal seedlings, in which the genotype effect is predominant (Table 1,

Table 2. Values of statistics, probability (P) and genetic parameters of seed–seedling transition measurements in Glycine max (L.) Merr. samples with different proportions of greenish seeds: aspects of the seedling phenotype.

Statistics			Characters
ANOVA Assumptions			NS (%)		AS (%)		SD (%)		VS (%)
	W (P)		0.88 (0.01)		0.92 (0.02)		0.80 (0.01)		0.91 (0.02)
	1F (P)		6.65 (0.01)		3.36 (0.01)		9.48 (0.00)		5.37 (0.01)
Model	Source of Variation	DF	MS	2F	MS	2F	MS	2F	MS	2F
Nested (Hierarchical) ANOVA	Genotype	8	2446.04	6.16	805.73	8.53 **	695.16	3.08	2452.50	6.93
	Greenish seeds (Genotype)	9	11,233.04	3.11 **	118.68	1.26 ns	804.82	3.57 **	865.02	2.44 **
	Error	54	397.06		94.42		225.71		354.11
		CV (%)	30.03		52.22		99.97		31.98
Genetic Parameters	h2 =		0.93		0.98		0.98		0.96
	CVg (%) =		13.67		35.05		37.82		14.22
	CVg/CVe =		1.26		2.73		2.31		1.82
	r (%) =		61.48		88.18		84.26		76.86

W: statistic of the Shapiro–Wilk test for residual normality (p ≥ 0.01); ¹F: statistic of the Levene test for homogeneity of variances (p ≥ 0.01); DF: Degrees of Freedom; MS: Mean Square; ²F: statistic of the Snedecor test (Nested ANOVA; p < 0.05); NS: normal seedlings; AS: abnormal seedlings; SD: dead seeds; VS: vigorous seedling. The ratio between the shoot and root dry mass. The statistical analyses for characters were based on transformed data by √x or, in the case of percentage data, by arcsine $\sqrt{x/100}$. CV: Coefficient of Variation (experimental ANOVA). h²: broad-sense heritability; CVg: coefficient of genetic variation; CVg/CVe: ratio between the coefficient of genetic and environment variation; r: intraclass correlation. ** p ≤ 0.01; ^ns: no significance.

and

Table 3. Values of statistics, probability (P) and genetic parameters of seed–seedling transition measurements in Glycine max (L.) Merr. samples with different proportions of greenish seeds: aspects of seedling growth.

Statistics		Characters
ANOVA Assumptions		RS (cm)			ShS (cm)		SS (cm)		RDrM (mg)		ShDrM (mg)		SDrM (mg)		ShS/RS		ShDrM/RDrM
	W (P)	0.80 (0.01)			0.86 (0.02)		0.75 (0.01)		0.82 (0.01)		0.98 (0.62)		0.97 (0.11)		0.90 (0.01)		0.94 (0.01)
	1F (P)	7.00 (0.02)			6.08 (0.01)		5.74 (0.01)		1.37 (0.26)		3.76 (0.02)		3.91 (0.01)		3.99 (0.01)		2.90 (0.04)
Model	Source of Variation	DF	MS	2F	MS	2F	MS	2F	MS	2F	MS	2F	MS	2F	MS	2F	MS	2F
Nested (Hierarchical) ANOVA	Genotype	8	153.58	1.18 ns	53.17	4.28 **	132.58	2.11 ns	2034.32	1.49	28,339.6	3.77 **	29,653.87	2.43 *	0.29	3.56 *	7.81	5.35 **
	Greenish seeds (Genotype)	9	44.42	2.06 ns	11.61	0.93 ns	94.43	1.50 ns	551.3	2.64 **	8763.30	1.17 ns	23,026.20	1.89 ns	0.04	1.61 ns	2.35	1.61 ns
	Error	54	12.43		12.43		62.94		1361.45		7471.66		12,187.95		0.08		1.46
		CV (%)	36.97		35.20		35.16		56.64		38.01		37.73		39.25		31.27
Genetic Parameters	h2 =	0.95			0.98		0.90		0.90		0.90		0.91		0.92		0.95
	CVg (%) =	61.69			82.80		24.92		23.74		26.85		26.54		44.58		29.53
	CVg/CVe =	1.58			2.79		1.06		1.04		1.06		1.15		1.21		1.51
	r (%) =	71.34			88.63		52.90		51.95		52.87		56.92		59.25		69.53

W: statistic of the Shapiro–Wilk test for residual normality (p ≥ 0.01); ¹F: statistic of the Levene test for homogeneity of variances (p ≥ 0.01); DF: Degrees of Freedom; MS: Mean Square; ²F: statistic of the Snedecor test (Nested ANOVA; p < 0.05). RS: root size; ShS: shoot size; SS: seedling size; RDrM: root dry mass; ShDrM: shoot dry mass; ShS/RS: the ratio between the shoot and root size; ShDrM/RDrM: the ratio between the shoot and root dry mass. The statistical analyses for characters were based on transformed data by √x or, in the case of percentage data, by arcsine $\sqrt{x/100}$. CV: Coefficient of Variation (experimental ANOVA). h²: broad-sense heritability; CVg: coefficient of genetic variation; CVg/CVe: ratio between the coefficient of genetic and environment variation; r: intraclass correlation. ** p ≤ 0.01; *: p ≤ 0.05; ^ns: no significance.

). This characteristic presents the highest heritability (h² = 0.98) among those studied for seedling phenotyping, which is explained by a high genetic variation (CVg/CVe = 2.73) and an intraclass correlation of 80.18% between genotypes and the greater presence of greenish seeds, although it presents potential for improvement by selection (CVg = 35.05%). This is the same pattern expressed by dead seeds at the end of the morphofunctional test of seedlings (Table 2). The percentage of abnormal seedlings presented great variability in the experimental analysis (CV = 99.97% and 52.22%, respectively) due to the excess of experimental plots with aberrant values, inherent to the measurement. The other measurements of seedling phenotyping also presented high heritability (h² ≥ 0.93), with great genotypic influence (1.26 ≤ CVg/CVe ≤ 1.82; r ≥ 61.48%).

The percentage of normal seedlings was negatively affected by the higher number of greenish seeds in the sample of four of the nine cultivars, and two cultivars (one with intermediate ability and the other with low ability to form normal seedlings) were insensitive (Figure 5). The samples with the best results developed 96% of normal seedlings, whereas those with the worst results had only 23% (Figure 5). However, some of the cultivars, such as cultivar eight, although with a greater presence of greenish seeds, unexpectedly obtained the best performance for normal seedlings (96%). By the way, this cultivar and cultivar three also showed the highest production of vigorous seedlings (90%) when the sample presented a higher proportion of greenish seeds (Figure 5). Only cultivar two had a similar performance, but only when the presence of greenish seeds was smaller. Three of the cultivars had a sudden reduction in vigorous seedlings when the sample had a higher number of greenish seeds. Three of the cultivars were insensitive to the presence of greenish seeds. These aspects of seedling phenotyping were deliberately associated with seed mortality rates throughout the data collection period, since the samples that presented higher values of dead seeds at the end of the test were those that had the worst normal and/or vigorous seedling results (Figure 5). However, the number of abnormal seedlings was determined by the genotype, being higher in genotype five (46%), which presented the lowest production of normal and/or vigorous seedlings, and lower in genotypes three, six and eight (between 7.34% and 10%), with satisfactory performances for the other phenotyping attributes (Figure 5).

The root dry mass was the only measurement of seedling growth in which the genotype effect was affected by the greater presence of greenish seeds, while the shoot size and dry mass, as well as the seedling dry mass and quotients between shoots and roots and between the shoot and root dry mass, were only affected by the genotype (Table 3). On the other hand, similar to germination synchrony, the seedling size (12.11 cm ≤ SS ≤ 30.80 cm) and root size (5.81 cm ≤ RS ≤ 18.66 cm) were species-specific aspects, not being affected by the genotype and/or the presence of greenish seeds (Table 3). It should be emphasized that, with high heritability (h² ≥ 0.90) and genetic propensity (1.04 ≤ CVg/CVe ≤ 2.79), as well as the great degree of association between the genotypic effect and the greater presence of greenish seeds in the sample (r = 51.95%), seedling growth measurements still presented high genotypic variation, especially in terms of shoot size (CVg = 82.80%).

Cultivar nine was the one that most functionally invested in the growth of the shoot, with 14.51 cm and 367.3 mg of dry mass (Figure 6). On the other hand, although originating from samples with a higher number of greenish seeds, the seedlings of this cultivar invested little in roots (RDrM between 62.87 mg and 75.97 mg) (Figure 6). In general, the cultivars with the highest seedling mass (SDrM between 337.15mg and 389.3 mg) were those in which the shoot/root ratio was higher (ShS/RS between 0.8374 and 1.0545). In addition, as expected, the cultivars that invested less in shoot mass (see ShDrM) were the ones that invested the most in roots (ShDrM/RDrM between 5301 and 5344) (Figure 6). As for growth, it is also noteworthy that the four cultivars that presented the best performance were also insensitive to a higher number of greenish seeds, regarding the accumulation of dry mass in the seedlings (Figure 6).

3.3. Relationship between the Seed–Seedling Transition Measurements

The greater presence of greenish seeds caused, therefore, the seed–seedling transition behavior of the cultivars to be altered, to the point that samples of the same cultivar ended up in two distinct groups of similarities (Figure 7). In general, the measurements of the germination process and post-germination events contributed differently to this grouping (Figure 7). These measurements add up to 59.6% of the variation attributed to seed–seedling transition, with germination measurements being explained by component one (38.4% of the variation) and the attributes of phenotyping and seedling growth being explained by component two (21.2%) (Figure 8).

Expressive linear correlations (r ≥ 0.7) ratify these associations, being observed mainly between seedling phenotyping and growth, and T50, while the other germination measurements expressed a correlation only with each other (Figure 9). In this sense, positive linear correlations between the normal seedlings and the vigorous seedlings, shoot dry mass and seedling size and dry mass, as well as between the vigorous seedlings and the shoot dry mass, demonstrate that all growth and phenotyping measurements are not necessarily related (Figure 9). On the other hand, all seedling growth measurements correlate linearly with each other (Figure 9), indicating that soybean seedling growth is finely regulated. Another interesting positive linear correlation is between the mean time and time for first germination, demonstrating that samples that germinated first reached the peak of protrusion early (Figure 9). This explains the positive linear relationship between the uniformity and predictability of germination (r between CVt and U in Figure 9). The positive linear relationships between Rate and Germinability and T50 and dead seeds (Figure 9) indicate, respectively, that the daily frequency of germination data are more related to the ability of the seed to germinate than the velocity at which the process occurs and that samples with more lethargic seeds have a higher possibility of death at the end of the test. Negative linear correlations between abnormal and normal or vigorous seedlings, as well as between dead seeds and normal seedlings, vigorous seedlings and growth measurements (Figure 9), show that samples that present a more functional seed–seedling transition are also those with a greater propensity for initial plant growth. This is reinforced by the negative linear correlation between the median germination time and the attributes of phenotyping and seedling growth (Figure 9). In addition, negative linear relationships between the average germination rate and mean germination time, as well as those between the Rate and MGT (Figure 9), indicate that samples with faster seeds, which have a higher daily frequency of protrusion, are those that reach the germination peak earlier. Besides this, a negative linear correlation between synchrony and uncertainty and the time for last germination denotes that samples that terminate germination late are less synchronous and present a less predictable germination pattern.

4. Discussion

The results reported here lead to three important and unprecedented findings: 1. The anomaly of greening in soybean seeds seems to be more expressed as a function of thermal variation than rainfall; 2. There are commercial cultivars with distinct sensitivity to the greening phenotype, indicating potential for classical breeding for the species; 3. The seed–seedling transition is significantly modified by the presence of greenish seeds, and the diagnosis of a higher proportion of green seeds can be anticipated by the analysis of embryo protrusion.

To date, the maternal trigger for the expression of the greening phenotype in soybean seeds does not consider the relationship between thermal and water variation from a quantitative and independent perspective [4,6,9,10,13]. In general, temperature, because of rain fluctuation, from the R5 stage on, has been considered a very important factor driving the increase in greenish seeds [4]. Thus, the first contribution of this work is not only in quantifying thermal stress as a preponderant factor for the appearance of greenish seeds but also in designing a relationship between thermal and rainfall variation.

Dissociating temperature and precipitation effects in the commercial field is not trivial. Our findings demonstrate that it is possible to have a considerable rainfall amplitude range (12–16 mm) for soybean seed production when considering greenish seeds, as it does not differ from thermal amplitude. It is difficult to say whether this is an optimal range for chlorophyll degradation due to our experimental design, but this indicates that the role of water and thermal stress can actually be additive; i.e., they act by independent routes in order to promote the anomaly, but on specific agroclimatic regimes, one stress intensifies the other. Although there is this perception in the field [4,7,8], there are no studies effectively dissociating the effects, but there are inferences about the role of hormones that moderate growth, such as ethylene and ABA [8,24,25,26]. Therefore, this work demonstrates the need to segregate the effects in the field in order to understand the processes of chlorophyll degradation in seeds whose mother plants have suffered distinct triggers for the expression of the anomaly. One point that supports our statement is that temperature variations of up to 1 °C in the maternal environment, in general, promote the expression of greenish seeds, but at a high rainfall amplitude (18 mm), there is a sharp increase in the anomaly. In this case, the high rainfall amplitudes point to a less uniform distribution of rainfall, and, therefore, it is possible that there is a water deficit experienced by the mother plant. In this regime, there were greater amplitudes of the minimum and maximum temperature. However, it was the variation between the maximum and minimum temperature that most intensified the expression of greenish seeds, demonstrating a catalyst relationship of thermal variation and water stress. On the other hand, environments with thermal variation of 1.5 °C presented the expression of greenish seeds conditioned to rainfall variation. There was a higher expression in an intermediate range of variation (10–12 mm) and a lower expression in a smaller range (6–8 mm). This indicates that when there is uniformity in the occurrence of rainfall during the life of the mother plant, even though the maximum and minimum temperature vary less, the high thermal amplitude is predominant for the anomaly, indicating a preponderant effect of temperature on rainfall. These examples converge on the idea of agroclimatic zoning for oilseed production [12], including the consideration of climate change scenarios such as global warming.

Although there are studies that have attempted to determine agroclimatic zoning for soybeans, there is still little information for guiding the choice of variables that should be analyzed when the product in question is seeds, not grains [9,13]. This information needs to be better developed for the soybean crop, since other crops, such as coffee, already have projects promoting production regions for the next 100 years [27]. The fact is that the thermal and rainfall amplitude are two of the main points explored by theorists on climate change [28], with the consensus that the distribution of rainfall will be increasingly uneven and the thermal amplitude will be greater. Thus, our findings not only raise the greening of soybean seeds as a biomarker of climate aggravation in production regions, such as Brazil, but also as a phenotype that should be considered in order to promote regional and international agroclimatic zoning for cultivation. This statement gains greater robustness when one thinks that the anomaly also impacts the quality of grains, as well as the longevity of the seeds. Nevertheless, here, we demonstrate that the greatest impact of greenish seeds during the seed–seedling transition is on the post-germination development, which includes seedling vigor and normal seedling production. It is necessary to consider that other researchers have already found a similar relationship regarding these characteristics, but this is the first work to determine the relationship between germination (embryo protrusion) and immediate post-germination. In any case, it is intriguing that the greatest impact of seed greening is on the initial development and not on protrusion. This indicates that alarming scenarios of climate change can not only impact agricultural production, since soybeans are the world’s largest source of animal protein feed and the second-largest source of vegetable oil [29]. Mentioning the impact of climate change in the face of our experimental design can be considered speculative, but it serves as a warning for a drastic scenario and promotes further studies mainly for biomarkers, whether biometric or molecular.

Some authors attributed the distinct sensitivity of commercial cultivars to greenish seeds [2,8]. However, the attribution was promoted by means of subjective criteria, and the characteristics that led to the classification were not quantified or specified. Here, however, a first indication for the distinct sensitivity of commercial cultivars was the finding that the germination pattern described by the relative polymodal frequency was affected differently by the presence of greenish seeds in the samples. A questionable point of this observation is that the relative frequency of germination does not allow for statistically dissociating the effects between the different proportions of greenish seeds in the samples of the different cultivars. Therefore, the hierarchization or nesting of the proportion of greenish seeds within the cultivar enabled us to ratify the statement that there are commercial cultivars tolerant to the phenomenon. This information is important, since the studies that characterize the effects of greenish seeds on the physiology of seeds and/or seedlings disregard the proportion of greenish seeds factor as a random effect, making inferences from factorial models with fixed factors. This can not only inflate type I and II errors from inferences but also minimize the physiological conclusions, resulting in difficulties in selecting materials for classical breeding.

Considering classical breeding [21], a relevant point is that the heritability of post-germination measurements is higher than those associated with the protrusion process. This effectively demonstrates a significant role of the genotype in phenotype variation during the seed–seedling transition, when considering the proportion of greenish seeds produced. This has been ratified by CVg, CVg/CVe and r. With high phenotypic variability, there is potential for selection gain. As a reinforcement of this idea, in addition to h² being higher than 0.7, the CVg of most of the analyzed measurements is higher than the values found for native cerrado species, such as Lithrea mollioides [14]. Therefore, it is possible that the anomaly of greenish seeds can be overcome through breeding programs. In this sense, as already observed, molecular biology should be a catalyst, and, therefore, it is expected that future studies use selection lines to promote a better understanding of genotype–environment interactions. These studies will promote accurate and rapid analysis tools, such as the SNP-array or GBS panels, used by other segments of the species’ genetic breeding. Regardless, it is possible to point out some potential morphofunctional diagnostic biomarkers and selection indicators for the anomaly. The absolute measurements of the physiological potential of the sample (germinability and normal seedlings), the time for last protrusion and the mean germination time, as well as the dry mass, proved to be good indications for the presence of the anomaly and may be the objective of classical breeding programs and the diagnosis of the problem. On the other hand, the germination synchrony, root size and seedling size were species-specific measurements and, therefore, suffered little influence either of the genotype or the presence of greenish seeds. We understand, of course, that our work used a small number of commercial cultivars, but the information contained herein cannot be mischaracterized and, therefore, we encourage studies in the near future to validate our information.

Another interesting perspective was that of T50 and first protrusion. Although they are the common measurements for the study of the vigor of seed samples and demonstrate great selection potential, especially because they are phenotypes whose variation is of genetic origin, they were measured with little magnitude and, therefore, are difficult for diagnosis. However, we emphasize the strong relationship between T50 and post-germination measurements. By the way, the analysis of seedling vigor pointed out some important aspects, the first being about the relationship between mass and size. Although the size was different for sensitive cultivars, it was the dry mass that was most negatively affected. This, in a way, can be expected. The seedling size is a measurement often masked by the etiolation of seedlings, natural in the development process in the germination chamber, while mass represents the conversion rate of reserve material into dry mass and, consequently, allows for a better view of initial growth and development. In the case of the presence of greenish seeds, the reduction in seedling mass may be a consequence of less reserve material in seeds with the anomaly. This is because it induces high respiratory rates [30], therefore reducing the amount of energy available to the seedling, such as what occurs with Pisum [31]. The respiratory rate/photosynthetic rate ratio during the seed–seedling transition, however, needs to be further studied, but our findings on mass point to an imbalance in favor of the respiratory rate. This would even justify the shorter longevity of samples with the anomaly that has been reported by technicians in the area but not yet academically clarified.

Another finding that intrigued us, and is actually worrying, is that the presence of the anomaly modifies the seed–seedling transition to the point of mischaracterizing the genotype effect of the commercial cultivar. This is alarming because the desired characteristics for a commercial cultivar include the initial formation of a stand based on the indicative area. By significantly affecting the transition, greenish seeds can give the sample an undesirable characteristic for a specific region and, therefore, affect the entire production chain.

The fact that the higher proportion of greenish seeds affects the variation in embryo protrusion and early post-germination measurements indicates that the critical phase of the anomaly in the transition may be protrusion. The presence of greenish seeds delays the germination peak and time for the last germination of the sample. Consequently, the measuring of post-germination, by not considering the date of protrusion as a weighting effect of growth, can give a false perception that the greatest weight occurs at this stage. Other points that reinforce the importance of protrusion analysis are the increases in dead seeds and abnormal seedlings with the increase in greenish seeds in the sample. Thus, studies of germination can contribute to accelerating the diagnosis and breeding to resist the effects of seed greening.

5. Conclusions

In view of the above, it is possible to conclude that the seed–seedling transition is sensitive to the greenish seed anomaly, and germination is the most sensitive process. With this in mind, it is possible that classical breeding can use embryo protrusion measurements as morphological markers. Sensitive and tolerant cultivars exist in the market and can be parental to overcome the problem. Finally, field experiments that contemplate the segregation between thermal and water amplitude in the expression of greenish seeds should be encouraged.

6. Patents

There are no patents resulting from the work reported in this manuscript.