Selection of plant material

Seeds of R. communis were used from the Mexican plateau (22° 36' 12" North latitude, 100° 24’ 47” West longitude), located in a backyard during a germplasm collection. The origin of the genotype could not be determined, but given its characteristics it was considered medium-sized (2.0 3.0 m tall), stem, leaf and purple cluster and fruits and grayish-purple seeds of medium size (^{SAGARPA-SNICS, 2014}).

The seeds of M. oleifera were obtained from a commercial plantation located in the south central region of Mexico (18° 30’ 09” North latitude 98° 35’ 14” West longitude). The origin of the genotype was not provided. The mother plants had average heights between four and five meters, their pods a length between 20 and 30 cm and seeds of average size and weight, in relation to that reported for this species (^{Foild et al., 2001}; ^{Ayerza, 2011}; ^{Oloyede et al., 2015}).

The harvest of both species took place three months before its sow. To correlate their weights and dimensions, 240 seeds randomly selected observing that they were free from defects or damage. The seeds were weighed with a digital precision scale (resolution of 0.001 g) and digital vernier measurements (resolution of 0.01 mm) on three dimensions: length (L), width (A) and thickness (E) (Figure 1). To estimate the percentage of the reserves and the embryo (almond) on the total weight of the seed, 80 seeds were extracted from the previous sample, to which the testa was removed and weighed again. The proportion of the almond was estimated as the ratio between the weight of seed without seed on the total seed weight.

Figure 1 Distribution of sizes and weights of seeds of Moringa oleifera and Ricinus communis. 

Substrate, experimental design, environmental and variable conditions

The sand was obtained from a coastal dune in the state of Veracruz, Mexico. To analyze and interpret the results, the procedures of NOM-021-SEMARNAT-2000 were used. The texture, by the method AS-09 (Bouyoucos), obtained a content of sand of 96%, silt 2.5% and clay 1.5%, reason why it was considered sandy. The pH, by the AS-02 method, was 7.8 (moderately alkaline). Organic matter and N, by methods AS-07 and AS-08, were not detected, so they were considered very low. The P, by the AS-10 method, was 2.8 mg kg^-1, so it was considered low. Given these characteristics, the substrate was considered very low in nutrient content.

The experiment began on September 1th at the facilities of the Postgraduate College, Veracruz campus, located in the municipality of Manlio Fabio Altamirano, Veracruz (19° 11' 55" North latitude, 96° 09’ 07” West longitude, 18 meters above sea level). A completely randomized arrangement was implemented in an open-air plot, under a 50% shade mesh, with each seed-seedling as an experimental unit and 80 seeds per species, extracted from the 220 previously weighed and measured. The seeds were placed in an upright position, with a caruncle downwards for the case of Ricinus and with the vertical wings in the case of Moringa and buried approximately 1.00 cm below the surface of the substrate; which was placed in plastic bags of 20 cm deep by 15 cm wide. The bags were irrigated daily in the mornings throughout the experiment.

The variables to be evaluated were: germination day from sowing; stem length in cm (with 1 mm precision tape), stem diameter in mm (with digital vernier 0.01 mm precision), number of leaves, cotyledon length and width or first true leaf when these they were no longer present (with metric tape), in cm. With the measurements of the leaves, their leaf area was estimated according to the equations described for Ricinus, ^{Naeem et al. (2011)}, and by calculating areas with scanned leaves and processed by ImageJ version 1.48 for Moringa.

The stems were measured every three days, to have 15 measurements at the end of the period, while the leaves were measured every seven days, with seven records during the experimental period. The average temperature of the experimental period, recorded by the meteorological station of the site, was 26.9 ±1.2 °C. At the end of the experiment the seedlings were unearthed, the lengths of their roots were measured and later the plants were separated in root, stem and leaves; weighed in fresh and placed in drying oven at 70 °C until constant temperature, to estimate dry weight. Both measurements were made using the Ohaus H-5276 analytical balance (0.001 g).

To estimate the potential content of oil in the seeds, a random sample of 20 g of seeds per species was taken, to which the husk was removed and then ground in a porcelain mortar. The oil extraction process was carried out by the Soxhlet method using 125 mL of n-hexane as solvent for 7 h.

Statistic analysis

On the measurements of the seeds their descriptive statistics were obtained (average, standard deviation, coefficient of variation (CV) and frequency distribution). To determine the relationship between seed weight and length (length, width and thickness), germination and seedling growth, the seed weight was correlated against their sizes, the number of days to germinate, the stem length and diameter, number of leaves, leaf area and robustness index (length of stem above base diameter) of the 15 samplings, as well as the fresh and dry weights of the extracted seedlings. The statistical significance of the correlations was analyzed using a two-tailed t-test with significance levels of 0.05 and 0.01. The possible relationship between seed weights and each character evaluated was estimated by testing five types of regression (linear: y= a + bx; exponential: y= a e ^x; potential: a x ^b ; logarithmic: a ln(x) + b; quadratic: y= ax² + bx + c), and the model with the highest coefficient of determination (r²) was selected, as the one that could explain the greater percentage of variation, which resulted in a total of 1200 tested models per species. To evaluate correlations, models and their regressions the Sigmaplot V10.0 program was used.

Distributions of seed sizes and their energy potential

For the Ricinus variety, both the sizes and weights of the seeds had little dispersion, with greater abundance of heavy seeds (negative asymmetry, as shown in Figure 1). The weight of the almond constituted 78% of the total weight of the seed, and its oil content was 55%. These results indicate a low variation in the seeds, when compared with studies in other R. communis varieties, where CVs of up to 52% were obtained for the weights (^{Naeem et al., 2011}; ^{Barrios et al., 2013}). A low CV, predominance of high weights, high oil content and high percentage of almond on husk, are characteristic of improved genotypes (^{Nielsen et al., 2011}; ^{Barrios Gómez et al., 2013}), so this origin could be considered an elite material for commercial use.

For Moringa, both sizes and weights were more dispersed, with a predominance of light seeds (positive asymmetry, as shown in Figure 1). The variation of the weights was found between the limits higher than those reported for African and Asian seeds, with values between 20 and 22% (^{Ogunsina, 2006}; ^{Ayerza, 2011}), except for the distribution, which in other origins was symmetrical or it had a slight negative asymmetry (^{Foild et al., 2001}; ^{Ogunsina, 2006}; ^{Ayerza, 2011}; ^{Oloyede et al., 2015}).

The positive asymmetry in seed weights is attributed to environmental pressures in mother plants or to a premature harvest in other tropical oilseeds (^{Valdes-Rodríguez et al., 2013}; ^{Severino and Auld, 2013b}), data that the producers of this origin did not provide, but it was found out that the mother plants came from imported seeds, reason why their adaptation could generate environmental pressures that increased the dispersion and the quantity of seeds of low weight (^{Ayala-Cordero et al., 2004}). The almond constituted 71% of the total weight of the seed, with 34% of oil, values similar to those of commercial genotypes in plantations in Africa, Asia and America, with percentages of oil between 30 and 42% (^{Olagbemide and Philip, 2014}).

Relationships between weights and dimensions of seeds

In both species, the correlations between weights and seed sizes were significant (Table 1). Although only for Ricinus were strong, with models that managed to explain more 75% of the variation; while for Moringa the correlations were weak and the best models only explained about 20% of the variation, with similar results in other Moringa provenances from India (^{Ogunsina, 2006}).

These values obtained are due to the shape of the cover of the seeds, which in Ricinus is adjusted to its almond, which generates very direct proportions with their weights; while the shape of the Moringa almond differs from its winged testa (Figure 2), which increases the variation in its relationships. For both species, the best regressions between weights and sizes were obtained with quadratic functions of convex curves (Table 1), which indicates that the lighter seeds had larger husks in relation to the seeds of medium weight. Although no similar analyzes were found on these two species, an analogous relationship was also found in Mexican J. curcas seeds (^{Valdes-Rodríguez et al., 2013}).

Figure 2 Cross sections of seeds of Moringa oleifera and Ricinus communis where the relationship between almonds and their heads is observed. 

Germination

The seeds of Ricinus reached a germination percentage of 96%, with average germination time of 4.7 days, while the percentage of Moringa was 95%, with average germination time of 5.9 days. The correlation between the weight of the seed and the days required to germinate was negative in both species, although it was only significant for Ricinus (Table 1), with a potential regression model (y= ax ^-b ), which was significant. For Moringa, the regression function was quadratic, but it was not significant (p= 0.47). However, another species of the genus (Moringa Peregrina) did show a negative and significant correlation between seed weight and days to germinate, as well as another subtropical arboreal species from India (^{Upadhaya et al., 2007}; ^{Gomaa and Xavier- Pico, 2011}).

In this regard, it has been recognized that in certain species the lighter seeds germinate faster in order to growth earlier and improve their life expectancies in relation to their heavier peers (^{Delgado et al., 2008}). However, the climatic conditions recorded during this experiment favored rapid and high germination rates, both in Moringa (^{Tesfay et al., 2016}) and in Ricinus (^{Ribeiro et al., 2015}), which is why it is considered that other tests that include greater environmental stress to increase these relationships; since environmental factors can promote diverse responses in seeds, depending on their biomass or size (^{Cordazzo, 2002}).

Seedling growth and its relation to the weight of the seed

The highest correlation between seedling weights and sizes was observed in the height of Moringa (r= 0.68, Figure 3), followed by the foliar area (r= 0.6), highly significant (p< 0.001) in both cases. The best regression was obtained with a potential model, which explained 50% of the variation in height. In Ricinus, the highest correlations were obtained with stem diameter (r= 0.41) and height (r= 0.4) and were highly significant (p< 0.001). The best regression was given by a quadratic model that predicted 21% of the diameter variation. Positive correlations, although minor (r= 0.55), between seed weight and stem length were also obtained in experiments with Moringa Peregrina in an arid substrate (^{Gomaa and Xavier Pico, 2011}) and in 15-day-old Ricinus seedlings with a model quadratic (r²= 0.13) (^{Naeem et al., 2011}), as well as in another oilseed (Jatropha curcas) in sandy substrate poor in nutrients (^{Valdés et al., 2014}).

Figure 3 Correlation coefficients (r) and determination (r²) of the models obtained between seed weights and growth variables of Moringa oleifera and Ricinus communis seedlings. Figure f) shows two examples of models obtained to define the relationship between seed weight and plant height in Moringa and Ricinus on the ninth day after germination. 

The relationship with the robustness index showed no significance for Ricinus and had a weak correlation, although significant, in Moringa (p< 0.05), which indicates that the seedlings of heavy seeds were relatively slender. In the roots the correlations with the weights of the seeds were positive and very significant (p< 0.01) for Moringa in their maximum lengths (r= 0.42) and the diameters of their pivoting (r= 0.44), followed by the number of roots laterals (r= 0.24), while for Ricinus the highly significant correlations were for the diameter of its pivot (r= 0.45), the average diameter of its lateral roots (r= 0.35) and the number of laterals (r= 0.39). The thicker roots and in greater quantity in seedlings of heavier seeds imply that part of the energy resources of the seed were invested in these organs (^{Kabeya and Sakai, 2003}), something very important in environments with limited resources, as in this sandy substrate.

Seedling growth and its relationship with seed sizes

For Ricinus, the highest correlation between growth and seed dimensions was obtained between seed length and stem length (r= 0.24). These results were consistent with what was reported in stems of 15-day-old Ricinus seedlings correlated with the size of their seeds, where r= 0.34 was reported for this relationship (^{Naeem et al., 2011}), which could be explained by the fact that The axis of the embryo is directly related to the length of the seed, thus that longer seeds could develop longer stems (Figure 4a). However, the dimensions of the seed did not seem to have significant effects on the diameter of the stem, number of leaves or leaf area. For Moringa, the width and the thickness of the seed were the dimensions with the highest correlations found in the length of the stem (r= 0.61 and 0.56, respectively), while the number of leaves and stem diameter had minor correlations with these measures (Figure 4b). In this regard, no reports were found on these relationships in Moringa, but the fact that it was the width and the thickness of the seed the measures with the highest correlations could be due to the fact that thicker seeds have greater reserves to nourish the seedling (Fotouo-M et al., 2015). In both species, the dimensions of the seed that best correlated with their weight were those that had the highest correlation with the dimensions of the emerging seedling.

Figure 4 Correlation coefficients (r) between seed sizes and growth variables of Moringa oleifera and Ricinus communis seedlings. 

The highest correlations in both weight and seed sizes obtained with Moringa could be due to the way in which it germinates, considered as hypogeo-cryptocotylar, where the cotyledons remain as reserves for the plant for up to 25 days after germination (^{Ramos et al., 2010}; Fotouo-M. et al., 2015), while in Ricinus germination is epigeous and albumen reserves are consumed when the cotyledons emerge, which are transformed into photosynthetic tissue (^{Severino and Auld, 2013a}).

According to this strategy, it is considered that the high-energy content of Ricinus seeds gives them a vigorous emergence, with large leaves, taller and thicker stems, which allow them to take better advantage of solar energy to produce more foliage and rise above other plants that could compete on the same substrate (^{Sanyal and Decocq, 2016}) (Figure 5).

Figure 5 Stem growth curves, leaf area and number of leaves in Ricinus and Moringa seedlings. 

On the other hand, Moringa emerges with thinner and smaller stems, as well as a much smaller leaf area, but the reserves of its seed allow it to maintain constant growth even in poor substrates, such as this sandy soil, where it was observed that after 27 days Ricinus stopped producing foliage and started its defoliation (although it was just at this stage when the heavier seedlings managed to maintain a greater number of leaves), while Moringa managed to maintain a 9% increase in the number of leaves until the day 42 (Figure 5d). This indicates that Moringa, coming from more arid environments (^{Parrotta, 2009}), has developed a strategy to consume the reserves of its seeds for a longer time and thus have greater opportunities to settle on less fertile substrates.

Biomass of seedlings against seed weights

For the dry weights of the plants, the stem had the highest positive and significant correlation recorded against the weight of the seed in both species (Table 3), followed by the root and leaves. According to the regression analysis, for both species the best adjustment was obtained with a potential model, where the weight of the seeds explained 21% of the variation of the biomass of the stems of Ricinus and 33% of those of Moringa; as well as 15% of the variation of root biomass in Ricinus and 28% in Moringa. Similar results were found with other seeds of Ricinus and their 15-day seedlings, where a quadratic model obtained an r² of 0.34 (^{Naeem et al., 2011}).

The obtained models indicate that the weight of the seeds positively influenced the final weights of the stems of their seedlings (Figure 6). Although the lower predictive capacity of the models for Ricinus was due to the high variation in the biomass of their plants, which at 49 days were completely autotrophic, while in Moringa some seeds were still found adhered to the base of the stems, which implies that their reserves could have been consuming for more than 25 days (^{Ramos et al., 2010}).

Figure 6 Models obtained to relate weight of seeds (ps) against dry weight of stems (pt) of Moringa and Ricinus seedlings. 

In relation to the organs, another investigation with recently emerged seedlings of Q. crispula indicated that the reserves of their seed are also transported to the roots to be maintained in case the stems are eaten by predators (^{Kabeya and Sakai, 2003}). Therefore, it can be considered that the biomass of these seeds had more impact on the stems, thus that the plants achieved higher heights and could compete better for light while more profuse roots ensure a better survival by exploring more widely a medium environment poor in resources, such as the substrate where they were.