Canopy position has a profound effect on soybean seed composition

Although soybean seeds appear homogeneous, their composition (protein, oil and mineral concentrations) can vary significantly with the canopy position where they were produced. In studies with 10 cultivars grown over a 3-yr period, we found that seeds produced at the top of the canopy have higher concentrations of protein but less oil and lower concentrations of minerals such as Mg, Fe, and Cu compared to seeds produced at the bottom of the canopy. Among cultivars, mean protein concentration (average of different positions) correlated positively with mean concentrations of S, Zn and Fe, but not other minerals. Therefore, on a whole plant basis, the uptake and allocation of S, Zn and Fe to seeds correlated with the production and allocation of reduced N to seed protein; however, the reduced N and correlated minerals (S, Zn and Fe) showed different patterns of allocation among node positions. For example, while mean concentrations of protein and Fe correlated positively, the two parameters correlated negatively in terms of variation with canopy position. Altering the microenvironment within the soybean canopy by removing neighboring plants at flowering increased protein concentration in particular at lower node positions and thus altered the node-position gradient in protein (and oil) without altering the distribution of Mg, Fe and Cu, suggesting different underlying control mechanisms. Metabolomic analysis of developing seeds at different positions in the canopy suggests that availability of free asparagine may be a positive determinant of storage protein accumulation in seeds and may explain the increased protein accumulation in seeds produced at the top of the canopy. Our results establish node-position variation in seed constituents and provide a new experimental system to identify genes controlling key aspects of seed composition. In addition, our results provide an unexpected and simple approach to link agronomic practices to improve human nutrition and health in developing countries because food products produced from seeds at the bottom of the canopy contained higher Fe concentrations than products from the top of the canopy. Therefore, using seeds produced in the lower canopy for production of iron-rich soy foods for human consumption could be important when plants are the major source of protein and human diets can be chronically deficient in Fe and other minerals.

118 analysis, and extremely small, wrinkled or off-color seeds were manually removed from all 119 samples before analysis. 120 121 Soy products 122 To produce flour, soybeans were blanched (boiled for ~25 minutes) and then baked 123 before grinding. To produce soymilk and okara (remaining solids), soybeans were blanched 124 (boiled for ~5 min) twice and then ground in water and cooled slightly. The soymilk (liquid 125 phase) and okara (solid phase) were separated using a cheesecloth and then dried separately and 126 reground before analysis.
127 Seed storage product analysis 128 Protein and oil were measured with an Infratech 1241 Grain Analyzer (FOSS Analytical 129 AB, Höganäs, Sweden), which is a true Near Infrared Transmission instrument that generates a 130 spectrum from 850 to 1050 nm via the monochrome light source and mobile grating system. A 131 50-ml seed sample was used that allowed for 10 subsample readings reported on a 13 % moisture 132 basis.
133 134 Ionomic analysis 135 Seed analysis was conducted as described in Ziegler et al. 2012. Briefly, single seeds 136 from each quadrant were weighed using a custom-built seed weighing robot and then digested in 137 concentrated nitric acid before loading onto an Elan ICP-MS. Internal standards were used to 138 control for differences in dilution and sample injection. Leaf and soy products were analyzed in 139 the same manner except that samples were added to digestion tubes by hand and weighed. 140 Custom scripts were used to correct for internal standards and correct for sample weight.  146 FW was extracted at room temperature with 1 mL of 50% methanol followed by addition of 800 147 µ1 of methanol:chloroform (1:2) as outlined in Supplemental File 7. Each extraction was 148 followed by centrifugation (5 min at 15,000 g), and the supernatants were collected. With the 211 Canopy position affects soybean seed protein, oil and mineral concentrations 212 We investigated positional effects with a core group of ten soybean lines (Supplemental Table   213 1) grown in Urbana, IL, over a 3-year period. Main stems were harvested at maturity and divided 214 into four canopy position quadrants (Fig. 1) and the seeds collected from each quadrant were 215 analyzed separately for major storage products (protein and oil) and various minerals. 216 Representative results obtained for one cultivar ('Chamberlain') are presented in Fig. 2A with 217 full plots provides as Supplemental File 1. As shown, protein concentration increased with 218 node position at which seeds developed going from bottom to top of the mainstem while oil and 219 iron (Fe) concentration decreased. For both protein and oil, which are the major seed 220 constituents, there was variation in the absolute concentrations among the 3 years of study, but 221 general trends were similar. Differences in absolute concentrations among years were most 222 apparent for protein concentration with highest levels obtained in 2010 and lowest in 2011, 223 presumably reflecting the impact of weather on seed development and composition. Another 224 confounding source of variation for canopy position analysis is genotype, and Fig. 2B highlights 225 the substantial variation in absolute concentrations of seed constituents due to both genotype and 226 year. As expected, absolute concentrations of Mg, S, K, P and Ca were highest (> 1000 ppm); 227 Mn, Fe, Rb, and Zn were intermediate (10 to 100 ppm), and Na, Co, Ni, Cu, Sr, Mo, and Cd 228 were present at trace levels (< 10 ppm).

229
In order to compare positional effects for various parameters across genotypes and years 230 without the confounding effects of differences in absolute values, we normalized each canopy 231 gradient to a mean value of one and the values for each quadrant were then expressed relative to 232 the normalized mean. However, because the weather in each year of the study differed 233 (Suppleemntal Table 2), the normalized results for each parameter are presented separately for 234 each year. Across the 10 soybean lines, oil concentration decreased progressively from bottom 235 to top of the canopy and was associated with a reciprocal increase in protein concentration (Fig.   236 3A). Protein and oil concentrations in soybean seeds are usually inversely related (Wilcox 1998) 237 and this was apparent with variation within the canopy as well. Single seed weight (designated as 238 sample weight in Fig. 3A) varied with canopy position with seed produced in the middle portion 239 tending to be slightly heavier than seeds produced at either the bottom or top of the canopy; 240 however, the storage product gradients were independent of seed weight variation. Storage 241 product gradients did not vary significantly across the three years of the study; however, absolute Manuscript to be reviewed 242 protein and oil concentrations varied among the three years of the study (Figure S1), This is 243 perhaps a result of weather that differed substantially in terms of temperature and precipitation 244 among the three growing seasons ( Supplemental Table 2). 245 We also found that canopy position significantly affected the seed ionome, which comprises 246 all of the minerals and trace elements found in mature seeds ( Fig. 3B and Supplemental File 2). 247

291
In addition to comparing parameters based on quadrant variation, it is also worthwhile to 292 compare plot averages, which will reflect genetic and environmental effects on absolute values 293 of the parameters. Figure 4B shows a matrix plot of correlations between plot means. Compared 294 to the corresponding plot that focused on quadrant variation (Fig. 4A), many more strong 295 correlations were apparent when comparing plot means. For example, protein concentration was 296 positively correlated with S and Zn (and more weakly with Fe). The correlation with S is 297 expected as the total seed S has been shown to track closely with high cysteine-and methionine-298 containing proteins in the soybean seed (Krishnan et al. 2012). The correlations between protein 299 content, Zn and Fe could be due to their primary role as cofactors of metalloproteins or to 300 variations senescence in leaves leading to nutrient remobilization (Uauy et al. 2006). 301 Accordingly, there was a significant negative correlation of Fe, S, and Zn with oil concentration. 302 Interestingly, there was also a strongly significant negative correlation of P with oil, whereas the 303 positive correlation of P with protein concentration was relatively weak. The majority of mineral  (Fig. 4A), but when analyzed in terms of plot means in Fig. 4B the association of 313 P with Mn, Fe, and Cu became apparent as well. It is worth noting that in terms of plot means, 314 there was no association between Ca and Sr suggesting that these chemical analogs do not 315 always behave similarly. There was one also a strong negative correlation between Mo and Sr,  (Fig. 5A). 328 Increased light energy to drive photosynthesis at most leaf positions and increased temperature at 329 lower positions could both favor increased protein accumulation at lower nodes thereby reducing 330 the difference between top and bottom seeds. However, while thinning significantly altered the 331 main stem gradients in major storage products there was relatively little effect on minerals. As 332 shown in Fig. 5B, the canopy positional effect on Mg, Fe and Cu was unaltered by the thinning 333 treatment whereas Ca and Sr were similar to one another and showed a significant effect of 334 thinning but only in one of the two test years (2010). The general conclusion is that thinning 335 affects the canopy positional effect on some but not all minerals. This suggests that at least for 336 Mg, Fe and Cu, the transport and homeostasis mechanisms are generally independent of 337 instantaneous environmental factors and the transport of sucrose and amino acids into the 338 developing seeds is not the sole factor driving their movement into seeds. In general, seeds lower in the canopy fill over a longer 348 period but at a lower rate compared to seeds at the top of the canopy (Raboy & Dickinson 1987) 349 so that at maturity, final seed size tends to be rather constant through the canopy rather than 350 increasing progressively from bottom to top of the canopy. We measured the SFPs with our core 351 group of ten lines and found substantial differences in SFPs at the bottom and top of the canopy 352 (Supplemental Table 3). Top SFP was generally correlated with bottom SFP, as would be 353 expected, but the difference in SPF (bottom -top position) was not correlated with the canopy 354 gradients of protein, oil, or Fe (Fig. 6). Therefore, factors other than the duration of the SFP are 355 responsible for the documented variation in composition with nodal position.   (Supplemental File 7). 419 In general, most metabolites did not show diurnal changes in concentration, but there were 420 differences in concentrations as a function of seed size and node position. The metabolite plots in 421 Fig. 9 illustrate some of the different patterns observed. The concentration of sucrose (Fig. 9A) 422 in developing seeds did not vary diurnally and remained relatively constant but the concentration 423 was slightly higher in the smallest seeds (day 1, top seed) compared to the larger seeds sampled 424 at the bottom position on day 1 or top position on day 7. The decrease in sucrose concentration 425 comparing top seed on day 1 and day 7 likely reflects in part the dilution effect caused by storage 426 product accumulation as the seeds increased in size by roughly 2-fold. In contrast, the 427 concentration of citrate in developing seeds was roughly equal among the three samples ( Fig.  520 et al. 2015). Conceivably, remobilization may also be triggered from leaves of all species under 521 certain conditions.

522
While multiple seed constituents exhibited canopy concentration gradients, it seems 523 unlikely that they are all caused by the same factors. Changing the microenvironment by 524 thinning plants to allow increased light penetration into the canopy altered the protein and oil 525 gradients but did not affect observed gradients for most of the minerals (Fig. 5). Furthermore,526 while the slope of many gradients changes across lines, treatment and year, the way that they 527 change is not well correlated between the different constituents, as illustrated in the plot 528 normalized correlation matrix (Fig. 4A), where relatively few strong correlations among the 529 various parameters were apparent. However, numerous correlations were apparent when mean 530 plot values were compared (Fig. 4B). Several minerals (e.g., P, Mn, Fe, Zn, S, and Co) had a 531 negative relationship with oil concentration and increased with protein concentration. Thus, 532 some coordination between seed storage product accumulation and mineral uptake into seeds is 533 evident. However, the results suggest that total uptake of a mineral and the allocation among 534 nodal positions are controlled by different mechanisms, and in general, canopy positional effects 535 on minerals and protein/oil appear to be controlled by distinct mechanisms. It should be noted 536 that altering the microenvironment by thinning plants did affect the observed gradients in seed 537 concentrations of Ca, Mn, and Sr, which were also the minerals altered in distribution in 2010 538 (the year of this study with above normal precipitation). These results highlight the differences 539 among minerals in terms of factors controlling their distribution among seed produced at 540 different node positions. Clear, continued studies in the future will be required to sort out the 541 different mechanisms involved. 547 in countries where plant-based diets are prominent. As discussed above, nodal position affected 548 the concentration of several minerals such as Mg, Fe, and Cu that were present at higher 549 concentrations in seeds produced at the bottom of the canopy. Iron is of particular interest and 550 was generally 20% higher in seeds produced lower in the canopy relative to the top and as 551 expected, differences in seed iron concentrations affected the concentration of iron in soy food 552 products made from those seeds (Fig. 7). Soy flour preserved more Fe than did milk; perhaps 553 mineral retention improvement through product preparation is possible. An immediate 554 application of our results with respect to human nutrition would be to use seeds from the top and 555 bottom halves of the canopy for different purposes, with seeds produced in the lower half 556 reserved for production of iron-rich soy foods for human consumption. Thus, knowledge of 557 these canopy position effects provides an unexpected approach to link agronomic practices to 558 improve human nutrition and health.  Another area that will be interesting to explore is the impact of elevated CO 2 on the 574 canopy positional effects described in the present study. It was recently reported (Loladze 2014;