An experiment was conducted at the Henry A. Wallace Agricultural Research Center facility, USDA-ARS, Beltsville, Maryland, USA using six controlled environment growth chambers in 2014 (EGC Corp., Chagrin Falls, OH, USA). The experiment was repeated over time in the same year. Soybean (cv. Spencer) was planted in 18 pots (one plant per pot, volume 7.6 L) filled with quartz silica sand (#2Q-ROK^®, US Silica Company, MD, USA) in each chamber. After emergence, the treatments were initiated in a combination of three levels of potassium (K) treatments, a sufficient level as 5.0 mM (control) and two deficient levels as 0.50 mM (moderate) and 0.02 mM (severe), in a modified Hoagland’s nutrient solution each under ambient (aCO₂, 400 µmol∙mol⁻¹) and elevated (eCO₂, 800 µmol∙mol⁻¹) atmospheric CO₂ concentrations. Two destructive plant harvests were conducted at 28 and 42 days after planting to avoid plant competition. Thereafter, six plants per chamber were maintained until the soybean maturity. The day/night chamber’s air temperature was maintained at 28/22˚C (12h/12h) throughout the experiment. The light as photosynthetically active radiation (PAR) was maintained at plant canopy height to the mean value of 900 ± 15 µmol∙m⁻²∙s⁻¹ using controllable ballasts (Osram Sylvania, MA, USA). Injection of either CO₂ or CO₂-free air was determined using a TC-2 controller that monitored CO₂ every 3 s, measured from an absolute infrared gas analyzer (WMA-4PP-systems, Haverhill, MA, USA). The relative humidity was not controlled and varied between 50% and 70% among chambers. More details of the experimental conditions have been reported previously by Singh and Reddy [11].

At maturity, plants were harvested and separated into plant organs of leaves, roots, stems, and pods. All plant parts, except the pods, were dried to a constant weight at 70˚C. Pods were dried at 35˚C in a forced-ventilation air for 10 days then the seeds were separated from pods. Dry weight of plant components and seeds were measured separately. The plant components and seed were ground using a Wiley Mill (Wiley^® Mill, Thomas Scientific, NJ, USA) to pass through a 1 mm screen. The plant tissue K concentration and concentrations of seed mineral elements of phosphorus (P), K, calcium (Ca), magnesium (Mg), sulfur (S), iron (Fe), manganese (Mn), zinc (Zn), copper (Cu), and boron (B) were determined using Spectro ARCOS EOP-Inductively Coupled Plasma (ICP) Spectrophotometer (Spectro Analytical Instruments, Mahwah, NJ, USA) at the Agriculture Diagnostic Laboratory, University of Arkansas, Fayetteville, AR, USA, according to the standard procedure [31]. The concentrations of seed carbon (C) and nitrogen (N) were determined by combustion of the ground materials using a CHN-2000 (Carbon Hydrogen Nitrogen-2000: LECO Corporation, St. Joseph, MI, USA) at the Adaptive Cropping System Laboratory, USDA-ARS, Beltsville, MD, USA. The weighted plant (whole-plant) K concentration was determined as the sum of the products of the dry mass of plant and seed, and their K concentration divided by the total dry mass [32].

The seed metabolic profiles were determined in 30 mg of ground seed materials from each plant. The pulverized seed material was suspended in 1.4 mL of 70% ice-cold methanol containing a final concentration of 2.5 mmol∙L⁻¹ α-aminobutyric acid and 2 mg∙mL⁻¹ ribitol as internal standards. The suspended samples were vortexed vigorously, allowed to stand at room temperature for 30 min and then centrifuged as described previously (Sicher, 2008). The pellets were extracted again with 1.4 ml of the above solvent, incubated in shaker bath for 15 min at 45˚C, kept at room temperature for 30 min and centrifuged as above. The supernatants were combined in a 15 mL Falcon tube and stored at −20˚C to be used for the determination of metabolic profile as described previously by our group [33].

In brief, the soluble carbohydrates and derivative (sucrose, glucose, fructose, maltose, ribose, myo-inositol-inositol, pinitol, and mannitol), organic acids (citrate, aconitate, succinate, fumarate, malate, pyruvate, Oxalic, shikimate, glycerate, adipic, and malonic), and amino acids (valine, phenylalanine, leucine, isoleucine, threonine, glycine, glutamine, proline, alanine, serine, and putrescine) were measured by gas chromatography (HP/Agilent 6890A, Agilent Technologies, Wilmington, DE) coupled to mass spectrometry. A 10 µL aliquot of each extract was transferred in a 1 mL reactivial and dried overnight under vacuum. The dried samples were dissolved in 100 µL of pyridine containing 20 mg∙ml⁻¹ methoxyamine and the vials were incubated in a shaker bath for 90 min at 30˚C. Subsequently, 100 µl of N-methyl-N-(trimethylsilyl) fluoroacetamide (MSTFA) was added to each vial, which was subsequently incubated in shaker bath for 30 min at 37˚C.

Concentrations of seed protein and oil were determined non-destructively using near infrared (NIR) spectroscopy in a 25 g sample of randomly chosen mature soybean seeds (moisture 13%) from each plant. The NIR analyses were performed by the National Center for Agricultural Utilization Research (NCAUR), USDA-ARS, Peoria, IL, USA. The concentrations of saturated fatty acids palmitic (C16:0) and stearic (C18:0) and unsaturated fatty acids oleic (C18:1), linoleic (C18:2), and linolenic (C18:3) of the total seed oil from three randomly selected seeds from each plant were determined by gas chromatography (GC) at the same facility as described by Singh et al. [34]. In brief, seeds were hammer-cracked manually (not pulverized) and the oil was extracted by soaking overnight in closed vials containing 1.5 ml of a mixture of CHCl₃:Hexane:MeOH (8:5:2 v:v:v ). Then, 0.1 ml of sodium methoxide (~0.35 M) solution and 0.5-1.0 ml of the extraction fluid were added and vials placed into the GC auto-sampler. Samples were analyzed on a GC (Agilent 6890 GC with an Agilent J&W GC column, Quantum Analytics, Inc., Foster City, CA) equipped with a flame ionization detector (320˚C, H₂ @ 35 mL/min, air @ 400 mL/min, makeup gas of N₂ @ 25 mL/min).

The SAS software (SAS Enterprise Guide, 4.2, SAS Institute Inc., NC, USA) was used for the statistical analysis. To test treatment effects and their interaction, PROC MIXED procedure with Kenward-Rogers (kr) adjustment of degrees of freedom was used for analysis of variance (ANOVA) using the mean value for a given seed component from each repetition of the experiment. Treatments (K and CO₂) and their interaction were considered as fixed effects and repetition of the experiment as a random effect. The ANOVA significance levels (P-values) are presented as *, **, *** and ^ns representing P ≤ 0.05, P ≤ 0.01, P ≤ 0.001, and non-significant i.e., P > 0.05, respectively.

Seed minerals elements and metabolic profile were also regressed against the plant tissue K concentration using the values from individual plant from each experiment across K treatment. The PROC MIXED procedure of SAS was used to select the best regression fit (quadratic versus linear) using ‘Akaike information criterion’ (AIC, lower is better), and to test the uniformity of coefficients of regression analysis between two CO₂ levels. In case the quadratic term or the slope (i.e., linear regression) significantly differed at alpha ≤0.05 between CO₂ levels, a regression relationship was established separately for each CO₂ level. Alternatively, the relationship was established across CO₂. The PROC REG and PROC GLM procedures were used for regression analysis to determine the level of the significance (P-value) and the coefficient of determination (r²). The significance is presented as *, **, *** and ^ns indicating P ≤ 0.05, P ≤ 0.01, P ≤ 0.001, non-significant (P > 0.05), respectively.

K deficiency significantly affected nine out of 12 seed mineral elements (Figure 1). Across CO₂ levels, several of these mineral elements declined under K deficiency compared to control, except N and Cu. The decline was relatively greater under the severe K deficiency. Seed C concentration decreased from an average value of 495.5 mg∙g⁻¹ to 479.5 mg∙g⁻¹, especially under severe K deficiency, across CO₂ levels. Among primary macronutrients, seed N increased 13% - 15% while P and K declined 2.5% - 9.2% and 24% - 38%, respectively, under K deficiency when averaged across CO₂ levels (Figure 1). The ratios among concentrations of primary macronutrient and C were also significantly affected by K treatment, except the C:P ratio. On average, K deficiency decreased C:N ratio ≈ 14% while increased C:K, N:P, and N:K ratios about 21-68% (Figure 1). Among secondary macronutrients, a significant K effect was mainly observed for Mg that was decreased by 27% - 45% under K deficiency. The concentration of micronutrients Mn and B declined on average 11% - 19% under K deficiency. Averaged across CO₂ levels, Zn concentration declined 15.5% primarily under moderate K deficiency. However, Cu increased ≈ 47%, especially under severe K deficiency. A significant K × CO₂ interaction was observed for B concentration and N:P ratios (Figure 1). The eCO₂ significantly decreased concentrations of P, S, Zn, and B with the minimum and maximum decline of 8% and 21% for S and B, respectively,

when averaged across K levels. Conversely, eCO₂ increased C:P and N:P ratios by 11% - 27%, particularly under the control and moderate K deficiency.

The concentration amino acids were significantly affected by K treatment and consistently increased under K deficiency across CO₂ levels, except glutamine (Figure 2). However, these increases were of greater magnitudes (3 - 18 folds relative to control) under severe K deficiency. The K × CO₂ interaction was significant for valine, phenylalanine, isoleucine, glycine, proline, and serine. Averaged cross CO₂, essential amino acids valine, phenylalanine, leucine, isoleucine, and, threonine concentrations were enhanced 24% - 72% under moderate K deficiency and 277% - 1600% under severe K deficiency compared to control, with threonine showing the greatest enhancement. The non-essential amino acids glycine, proline, alanine, and serine had 6% - 75% greater values under moderate K deficiency than control average across CO₂ levels. However, under severe K deficiency, these amino acids were approximately 300% - 1700% greater than control, with serine and alanine having the highest and lowest increment, respectively. The CO₂ effect was also significant for phenylalanine, alanine, and serine (Figure 2). Relative to aCO₂, the eCO₂ decreased phenylalanine and serine concentrations by 20% - 40%, especially under severe K deficiency. Conversely, eCO₂ stimulated alanine concentration by 36% under the similar situation.

The concentrations of protein, oil, and fatty acids were assessed mainly between the control and moderate K deficiency across CO₂ levels due to very low production of seeds under the lowest K treatments, which were mostly used for the seed mineral and metabolic profiling (Figure 3). The seed production across these

treatments have been reported previously (Singh and Reddy 2017). The effect of K was significant for these seed components. Averaged between CO₂ levels, concentrations of protein and oil increased and decreased by 8.3% and 4.5% under moderate K deficiency, respectively, compared to control (Figure 3(A)). The K deficiency also increased protein:oil ratio by 14% across CO₂ levels. Concentrations of both saturated fatty acids, palmitic and stearic acids, decreased 7% - 8% under moderate K deficiency, when averaged across CO₂ levels. However, that of mono-unsaturated fatty acid, oleic acid, was increased under moderate K deficiency by 34% and 58% under aCO₂ and eCO₂, respectively. Conversely, under moderate K deficiency, essential fatty acids (i.e., linoleic and linolenic acids), declined by 11% - 17% and 22% - 26% under aCO₂ and eCO₂, respectively. The eCO₂ significantly increased oleic acid by 25% while decreased linoleic acid by 14%, particularly under the moderate K deficiency (Figure 3(A)). There was an inverse linear relationship between seed protein and oil concentration when regressed across the CO₂ levels (Figure 3(B)).

Among the soluble carbohydrates in soybean seeds, sucrose was the most abundant followed by pinitol (Figure 4). Relative to control, K deficiency decreased sucrose concentration by 16% - 40% averaged across CO₂ levels. However, concentrations of soluble sugars, glucose, fructose, and ribose, were more than doubled, particularly under severe K deficiency. Maltose was not significantly affected by the treatments. Averaged between CO₂ levels, sugar alcohols, myoinositol and mannitol, also increased 51% - 200% under moderate K deficiency and 350% - 1000% under severe K deficiency compared to control. The sugar

alcohol, pinitol, decreased about 40% then increased by 58% under moderate and severe K deficiency, respectively. The effect of CO₂ was significant for glucose, fructose, myo-inositol, and mannitol (Figure 4). Seed glucose concentration consistently decreased 20% - 60% under eCO₂ versus aCO₂ across K levels. However, fructose concentration increased 29% - 52% under eCO₂. Relative to the aCO₂ level, eCO₂ also increased concentration of myo-inositol by 23% - 124% and mannitol by 28% - 45% across K levels, but these increases were the greatest under severe K deficiency (Figure 4).

Majority of the measured organic acids were significantly affected by K treatments while the effect of CO₂ was mainly significant for aconitate and pyruvate (Figure 5). Averaged between CO₂ levels, concentration of tricarboxylic acid (TCA) cycle intermediates, citrate, fumarate, and malate, consistently declined about 42% - 72% under K deficiency whereas TCA intermediates, aconitate and pyruvate, increased 114% - 200%, especially under severe K deficiency. The TCA intermediates, succinate and oxalic acid, were not significantly affected by the

treatments. The moderate K deficiency primarily enhanced the glycerate and adipate concentrations about 92% - 105% (Figure 5) while severe K deficiency enhanced the concentrations of shikimate, glycerate, adipate, and malonate approximately 45% - 728%. The highest and lowest increases were observed for glycerate and malonate, respectively, when averaged across CO₂ levels. The eCO₂ decreased aconitate concentration 24% - 45% but increased pyruvate concentration approximately 90% - 250% across the K treatments (Figure 5).

Figure 6 & Figure 7 illustrate the relationship patterns between plant K status and concentrations of various seed constituents and ratios. The regression relationships between tissue K concentration and seed constituents exhibited a similar trend (decreasing or increasing) as observed due to K deficiency as induced by the K-deficient treatments. Out of 56 relationships (49 seed constituents + six ratios), 51 regression relationships of plant tissue-K concentration with the seed constituents (or the ratio) were significant (P ≤ 0.05). Out of 51 regression relationships, 20 were best described by a linear regression equation while 31 by a polynomial second order (quadratic) regression equation. However, the remaining four (Ca, S, maltose, and oxalate concentrations) did not show a significant relationship with the plant K concentration (Figure 6 & Figure 7). The minerals N, P, K, and their ratios, protein, oil, and fatty acids had a linear relationship with the plant K concentration while a quadratic relationship was observed for carbohydrates (except ribose), organic acids (except fumarate) and all amino acids. About 29 seed constituents including ratios tended to increase with a decrease in plant tissue K concentration. For instance, concentrations of minerals such as N, Cu, and ratios (C:P, C:K, N:P, N:K), majority of organic acids, soluble carbohydrates (except sucrose), amino acids, protein, and oleic acid increased when plant K concentration decreased to the minimum. Conversely, seed constituents including C, P, K, Mg, Oil, palmitate and essential fatty acids, citrate, and malate decreased with plant K concentration. The relationships also differed significantly between CO₂ levels for the ratios of C:N, C:P, N:P, and N:K, and for 12 seed constituents including C, N, P, B, glucose, ribose, myo-inositol, fumarate, malate, and protein (Figure 6 & Figure 7).

Influence of K deficiency on soybean was obvious from the substantial decline of K concentration in seeds and plant tissues across CO₂ levels. This study revealed that soybean seeds are likely to contain deficit levels of several minerals when grown in K-deficient soils, thus affecting seed nutritional quality. Interestingly, K deficiency affected concentrations of several mineral elements although they were not limiting in the rooting media. Alterations in vegetative tissues and seed mineral concentrations have been reported previously under nutrient deficiency in several crops [35] [36]. A concomitant decline of K and P concentrations in the seeds from K-deficient soybean was parallel to the deficit of seed K under P deficiency, suggesting a close association between these two macronutrients under

limiting conditions [27]. Similar to our study, Nelson et al. [36] reported decreased Mg concentration in soybean seeds under K deficiency. Walker et al. [35] also found significantly lower concentrations of B and Zn in K-deficient seeds of at least one out of five soybean cultivars tested under field conditions. Although B declined across K fertilization, Zn primarily decreased under the moderate K deficiency in this study. Zn is an essential nutrient in the human diet for proper function of immune systems [18].

The changes in seed mineral elements also significantly affected the ratios among C concentration and macronutrients (N, P, and K). In K-deficient seeds, a greater seed C:K ratio mainly resulted from a larger decline of K than C concentration. However, greater seed N:P and N:K ratios were primarily caused by an increased N while decreased K and P concentrations under K deficiency. In the seeds of K-deficient soybean, the decreased C:N ratio was consistent with the previous observations [27] [36]. Moreover, a lower C:N or higher N:K ratios might have resulted from a decreased level of C-rich compounds (e.g., carbohydrates, oil) and tissue K concentration, respectively, without a commensurate decline or an accumulation of N-rich compounds (e.g., amino acids, protein) under K deficiency [27]. When plant growth is limited by macronutrients other than N in the rooting media (e.g., K), the excess N in tissues are typically stored as N-rich compounds leading to an altered C:N ratio in plant tissues [27] [37] [38]. A consistent increase of amino acids (N-rich compounds) concentrations in seeds under K deficiency across CO₂ levels was also found in this study. The glutamine was the most abundant amino acid as often reported in other studies [27] [39]. The accumulation of proline and putrescine (a polyamine) signified their role in the osmotic adjustments and desiccation avoidance under stress situations [40].

Protein and oil are considered as the major soybean seed reserves and directly contribute to the economic value [30]. However, an inverse relationship between seed protein and oil concentration observed under K deficiency situations indicated a trade-off between these components [21] [27] [37]. Nevertheless, increased protein concentration clearly indicated that the K deficiency did not limit protein synthesis in soybean seed [37]. Alternatively, K deficiency not only decreased oil concentration but also reduced the nutritive value of its fatty acids’ components by decreasing accumulation of essential fatty acids (linoleic and linolenic). Gaydou and Arrivets [41] reported comparable changes in seed linoleic and linolenic acids in soybean grown across various nutrient fertilization, which was consistent with the observation made in this study.

Sucrose is the primary form of translocated photo-assimilate and is crucial for biosynthesis of protein and oil [24] [26]. Sucrose is also the most abundant soluble sugar in soybean seeds as observed across treatments in this study and elsewhere [27] [33]. Hymowitz et al. [42] studied 60 soybean lines and found that sucrose had a positive and negative correlation with oil and protein, respectively, which was consistent with their patterns observed against plant K concentration in this study (Figure 7). The enzymes phosphoenolpyruvate carboxylase and pyruvate kinase play important role in the carbon partitioning form sucrose to protein and oil of developing seeds, which is most likely influenced by the ratio of available amino acids and soluble carbohydrates [26] [43]. The lower sucrose concentration under K deficiency was accompanied by accumulation of simple sugars glucose, fructose, and ribose, and sugar alcohols, which agreed with the observations made under P deficiency [27]. Simple sugars play important role in seed germination and desiccation tolerance during seed maturation [44]. Accumulation of sugar alcohols such as mannitol, myo-inositol, and pinitol act as compatible solutes and help to maintain the cellular integrity of membranes, and are strongly correlated with plant tolerance to abiotic stresses [45].

Organic acids represent the stored pool of photosynthetic products and synthesized as intermediates in metabolic pathways including tricarboxylic acid cycle (TCA, i.e., reparation) and during the biosynthesis of secondary metabolites, fatty acids, and amino acids [46] [47]. Relative to the plant leaf tissues, reports on the concentrations of organic acids in crop grains/seeds are scarce, especially in nutrient-limited conditions [27] [33] [48]. The concentration of TCA intermediates citrate, malate, and fumarate decreased while pyruvate (the main substrate for TCA cycle) was enhanced under K deficiency. Generally, citrate followed by malate is the most abundant organic acid in plant tissues as also observed in this study across treatments [33] [47]. The lower seed citrate and malate concentrations under K deficiency contradicted the observation made under P deficiency, indicating a varying influence of these nutrients on the metabolic processes [27]. However, accumulation of TCA intermediate aconitate under K deficiency across CO₂ levels was consistent with P deficiency which might result due to the depressed activity of aconitase enzyme under stress situations [27] [49]. Gupta et al. [49] also suggested that the inhibition of aconitase activity might result in enhanced biosynthesis of amino acids, as observed in this study. The accumulation of organic acids shikimate and glycerate under severe K deficiency was comparable to the observation made in other study [27]. Under K deficiency, the accumulation of adipate might be attributed to the breakdown of amino acid lysine [50]. The malonate also increased, especially under the severe K deficiency. Malonate plays role in the symbiotic fixation and might serve as the precursor of lipid biosynthesis [47].

The CO₂ significantly affected 16 seed constituents, and eight of them (P, S, Zn, B, glucose, aconitate, linoleic, and linolenic acids) had lower values while six (fructose, myo-inositol, mannitol, pyruvate, alanine, and oleic acid) had higher values under eCO₂ versus aCO₂ consistently across K fertilization. However, differential sensitivity to eCO₂ in the concentration of several seed components was found because of the varying degree of responses among three K fertilization resulting in a K × CO₂ interaction. For instance, the concentration of essential amino acids (e.g., valine and phenylalanine) tended to be greater at eCO₂ versus aCO₂ under the control or moderate K fertilization but declined under severe K deficiency. Likewise, the eCO₂-mediated accumulation of fructose and myoinositol was relatively much greater under severe K deficiency than other K levels. The eCO₂ also increased the C:P and N:P ratio in soybean seeds, particularly when the K deficiency was not severe, and agreed with the previous observations [27] [51].

Lower concentrations of mineral nutrients and metabolites in plant tissues under eCO₂ have often been reported and found to vary between species, cultivars, and growing environments [3] [15] [16] [17] [27] [51]. Over the decades, dilution effects due to increased photosynthates, plant size, and leaf thickness, lower transpiration, and adjustment in the nutrient uptake and utilization have emerged as the nature of the decreased concentrations of minerals and other constituents in plant tissues under eCO₂ [16] [17] [19] [51]. Furthermore, an altered N metabolism under eCO₂ also influences biosynthesis of organic compounds, thus affecting the overall metabolic response of plants to eCO₂ [52]. Therefore, the inconsistent impacts of eCO₂ on various seed constituents suggested the existence of a combination of mechanisms contributing to the changes observed in this study. The insignificant effect of eCO₂ on the seed oil and protein concentrations have also been reported in soybean and other crops [15] [16]. However, eCO₂ changed the composition of fatty acids by enhancing oleic acid but decreasing linoleic and linolenic acids (essential fatty acid), especially under K deficiency. Högy et al. [53] also found decreasing trends in these essential fatty acids under eCO₂ in the Brassica sp., which indicates negative impacts of CO₂ enrichment on the seed nutritive values.