Chromatographic separations were performed using an Accela ultra-high performance liquid chromatography (UHPLC) system (ThermoFisher Scientific, Bremen, Germany) fitted with a Syncronis C18 100 × 2.1 mm analytical column with 1.7 μm particle size (ThermoFisher Scientific). The mobile phase consisted of (A) ultrapure water with 0.1% formic acid and (B) acetonitrile (MS grade) + 0.1% formic acid. The elution gradient program was as follows: 0.0 – 1.0 min, 5% B; 1.0 – 16.0 min, 5 – 95% B; 16.0 – 16.2 min, 95 – 5% B; 16.2 – 20.0 min, 5% B. The injection volume for all samples was 25 μL, and the flow rate was 0.3 mL/min.

The UHPLC system was coupled to a LTQ OrbiTrap mass spectrometer (ThermoFisher Scientific) equipped with a heated-electrospray ionization probe (HESI-II) operated in negative ionization mode. HESI-source parameters were the same as previously described (Banjanac et al., 2017). Mass spectra were acquired by full range acquisition covering 100–1,500 m/z. For the fragmentation study, a data-dependent scan was performed by deploying collision-induced dissociation (CID). The normalized collision energy of the CID cell was set to 35 eV.

Metabolites were identified and some of them quantified according to the corresponding spectral characteristics: mass spectra, accurate mass, characteristic fragmentation, and characteristic retention time. ThermoFisher Scientific Xcalibur software (version 2.1) was employed for instrument control, data acquisition, and data analysis. Fragmentation mechanism and characteristic MS² and MS³ fragments were confirmed using ThermoFisher Scientific Mass Frontier software (version 6.0).

Separation of organic acids was performed on a Waters HPLC system (Waters Corp., Milford, MA, USA) consisting of 1525 binary pumps, thermostat, and 717+ autosampler connected to the Waters 2996 diode array detector (DAD) adjusted at 210 nm. The HPLC system was fitted with a Supelco C-610H (300 x 7.8 mm) anion exchange column (Sigma-Aldrich Co., Barcelona, Spain) connected to the appropriate guard column. Isocratic elution was carried out with 0.1% H₃PO₄ as the mobile phase at a flow rate of 0.5 ml/min and a holding column temperature of 40°C. Quantification of organic acids was based on developing calibration curves of pure compounds. Values are expressed as μg per g of dry weight (μg g⁻¹ DW). Standards of tartaric, citric, isocitric, fumaric, malic, formic, and oxalic acids were purchased from Sigma Aldrich Co. (Steinheim, Germany).

Separation of sugars was performed on a Waters Breeze chromatographic system (Waters Corp.) connected to a Waters 2465 electrochemical detector (pulsed amperometric detection, PAD) with a 3 mm gold working electrode and hydrogen reference electrode. This HPLC system was fitted with a CarboPac PA1 (Dionex Corp., Sunnyvale, CA) 250 × 4 mm column equipped with the corresponding CarboPac PA1 guard column, held at 30°C, and used a flow rate of 1.0 ml/min. Sugars were eluted using 0.2 M sodium hydroxide (50% w/w, low carbonate; J.T. Baker Co., Deventer, Holland). Signals were detected in pulse mode using the following waveform: E₁ = +0.15 V for 400 ms; E₂ = +0.75 V for 200 ms; E₃ = −0.80 V for 300 ms and within 150 ms of integration time. Filter timescale was 0.2 s and range set at 500 nA for the full mV scale. For both sugars and organic acids, data acquisition and spectral evaluation for peak confirmation were carried out using Waters Empower 2 Software (Waters Corp.). Quantification was performed using calibration curves of pure compounds. Standards of sucrose, fructose, glucose, sorbitol, and mannitol were purchased from Sigma Aldrich (Steinheim, Germany).

Metabolic fingerprinting of methanol extracts of maize grain using UHPLC coupled with mass spectrometry (MS) detection, which combines high-resolution mass spectrometry (HRMS) and multistage mass spectrometry (MS² and MS³ fragmentation), identified a total of 90 compounds (Supplementary Table 1). The identity of 16 compounds (compounds 1, 5, 9, 12, 14, 21, 25, 26, 32, 66–68, 76, 87, 89, 90) were confirmed by direct comparison with appropriate standards, while the identity of the remaining 74 compounds were obtained after examination of MS data (exact mass and specific fragmentation), in consultation with the available literature on LC/MS and nuclear magnetic resonance (NMR) analysis of maize metabolites (Li et al., 2018; Zhou et al., 2019; Lux et al., 2020; Eleazu et al., 2021). Basic retention and MS data, as well as references confirming the presence and, in some cases, mass fragmentation of specific compounds in maize are listed in Supplementary Table 1. The structures of some compounds (compounds 10, 11, 15, 22, 27, 28, 33, 34, 38, 45, 70, 71) were proposed based only on MS data, because a search of available databases did not provide confirming information. These newly identified compounds are clearly labeled in Supplementary Table 1, as is the availability of very similar derivatives in the maize literature, which helped with their identification. Several compounds were identified in maize for the first time but previously detected in other species of the Poales order. These are compounds 29 (Lee et al., 2010), 42–44 (Shirota et al., 1996), and 72 and 73 (Kang et al., 2018). The chromatograms of all maize samples showed similar metabolic profiles. As an example, the base peak chromatogram of maize grain extract (Organic variety, Org6 cropping system) is shown in Supplementary Figure 1. The peak numbers in Supplementary Figure 1 correspond to the compound numbers in Supplementary Table 1. For easier understanding of Supplementary Table 1, the compounds were divided into subgroups based on structure: benzoic acid derivatives (12 compounds), cinnamic acid derivatives (33 compounds), hydroxycinnamic acid amides (20 compounds), flavonoids (3 compounds), lignans (5 compounds), plant hormones (11 compounds), and “other” compounds that didn't fit into this grouping (6 compounds).

Benzoic acid is the simplest aromatic carboxylic acid, with a carboxylic group bonded to the benzene ring. Benzoic acid derivatives were represented in the maize samples by simple acids (compounds 1, 5, 9, 12) and various glycosides (Supplementary Table 1). Mass spectra of all glycoside derivatives showed a similar fragmentation pathway characterized by loss of hexoside (162 Da), dihexoside (324 Da) or pentosyl hexoside (294 Da).

Cinnamic acid derivatives are polyphenolic compounds originating from the mavolanate-shikimate pathway that include simple phenolic compounds such as cinnamic acid, p-coumaric acid, ferulic acid, and caffeic acid. Monomers, dimers, and rarely trimers of hydroxycinnamic acids were detected in maize samples as sucrose acetyl esters (15 compounds) or bound to glycerol (5 compounds) (Supplementary Table 1). Consistent with the literature, there were numerous ferulic acid derivatives (dehydrodimers, glycosides, acyl glycosides, and soluble and insoluble amide derivatives) with a significant number of compounds containing two ferulic acids or one ferulic and one caffeic acid, sucrose, and several acetyl groups (Bento-Silva et al., 2018, 2020; Zhou et al., 2019). For example, compound 28, found at 763 m/z, showed a MS² base peak at 721 m/z (loss of one acetyl group – 42 Da) and MS³ base peak at 545 m/z (further loss of feruloyl unit – 176 Da). All other MS² and MS³ secondary peaks are depicted in Supplementary Figure 2. Based on all these findings, this compound was identified as feruloyl-caffeoyl-sucrose di-acetyl ester. The exact position of the two acetyl groups could not be determined. Supplementary Figure 2 presents the tentative structure and proposed fragmentation pathway of this compound.

Hydroxycinnamic acid amides are a diverse class of secondary metabolites that arise from condensation reactions between CoA esters of hydroxycinnamic acids with aliphatic/di/poly- or aromatic mono-amine groups. The combination of 96% methanol and ultrasound during extraction resulted in identification of 20 hydroxycinnamic acid amides in maize grain samples (Supplementary Table 1). Hydroxycinnamic acid amides, commonly present in maize, were detected as derivatives of putrescine (compounds 46, 55, 57, and 57–59), spermidine (compounds 47–54 and 61–63), and tryptamine (compounds 56, 64, 65) (Li et al., 2018; Bartolić et al., 2020; Bento-Silva et al., 2020), and one compound (60) was detected as a tyramine derivative (Marti et al., 2013). Only one glycosidic derivative (compound 47; 598 m/z) was detected, and the molecular formula, C₃₁H₄₁N₃O₉, was calculated by examining its exact mass. MS² and MS³ spectra confirmed that this compound can be identified as a dicoumaroyl spermidine hexoside (Supplementary Figure 3), which was previously detected in maize seeds (Li et al., 2018).

Many different hydroxycinnamic acid amides consisting of 2 hydroxycinnamic acids linked through their carboxylic acid moiety by a chain of putrescine or spermidine, were also detected in the maize grain samples (Supplementary Table 1). Some of the feruloyl putrescines previously reported for maize are coumaroyl-feruloyl putrescine, Di feruloyl putrescine (Sen et al., 1994; Acosta-Estrada et al., 2015), dicoumaroyl putrescine, feruloyl putrescine, and p-coumaroyl putrescine (Sen et al., 1994). Most of these compounds were also present in maize grain samples analyzed in the present study. According to previous studies, the majority of the hydroxycinnamic acid amides are in the outer tissues of maize grain (Burt et al., 2019) and can be involved in defense against postharvest insect pests (García-Lara et al., 2010; García-Lara and Bergvinson, 2014) and serve as co-pigments in colored maize (Collison et al., 2015). This group of compounds was not quantified due to the lack of available standards.

Flavonoids are synthesized through the phenylpropanoid pathway, transforming phenylalanine into 4-coumaryoyl-CoA, which is transformed via chalcone synthase, the first enzyme in the flavonoid biosynthesis pathway, into chalcone scaffolds from which all flavonoids are derived. All three flavonoids detected, naringenin, apigenin, and hesperetin (compounds 66–68), were previously detected in maize, and their identity confirmed using appropriate standards (Supplementary Table 1).

Lignans share common biosynthetic pathways and consist of two propyl-benzene units coupled by a ß, ß'-bond. Of the 5 lignans identified in the methanol extracts of maize grain (Supplementary Table 1), none were previously detected in maize. Compound 69 (syringaresinol 4-O-hexoside) was not common in the Poales order, but its structure was confirmed in the literature (Li et al., 2018). Oryzativol A and B (compounds 72, 73), specific for rice (Kang et al., 2018), as well as two derivatives of oryzativol without the cinnamic acid residue (compounds 70 and 71) were detected in methanol extracts of the maize samples.

A total of 11 phytohormone compounds were identified in methanol extracts of maize grain (Supplementary Table 1). Zeanosides A and C (compounds 74, 75), named after maize, are indole-3-acetic acid (IAA)-related compounds (Tateishi et al., 1987). Three gibberellins (compounds 78, 79, 81) were detected: gibberellin A53, A12, and A53-aldehyde (Kobayashi et al., 1996). The stress hormone abscisic acid (compound 76) was also detected in maize grain.

Six compounds did not fit into the above grouping. Of these “other” compounds, the most interesting was DIMBOA-Glc (compound 85; Supplementary Table 1). DIMBOA-Glc is a glucoside derivative of DIMBOA, a powerful antibiotic naturally occurring in maize (Sauter et al., 2002).

Data presented here are consistent with previous studies that revealed that grain of various maize varieties contain several phenolic compounds, including phenolic acids (e.g., vanillic, p-hydroxybenzoic, caffeic, p-coumaric, ferulic, syringic, synaptic, protocatechuic) and flavonoids (Kandil et al., 2012; Guo and Beta, 2013; Das and Singh, 2015, 2016; Thakur et al., 2017) with ferulic acid and p-coumaric acid present in high concentrations (Mathew and Abraham, 2004; Vitaglione et al., 2008; Santiago and Malvar, 2010). Ferulic acid, p-coumaric acid, and vanillin were the predominant phenolic compounds in analyzed maize grain samples, in both conventional and organic varieties. Amounts of ferulic acid ranged between 0.39 and 3.68 mg kg⁻¹ DW, while p-coumaric acid was present in amounts ranging between 0.99 and 7.62 mg kg⁻¹ DW. Vanillin (0.41–3.64 mg kg⁻¹ DW), vanillic acid (0.45–2.30 mg kg⁻¹ DW), and p-hydroxybenzoic acid (0.24–1.37 mg kg⁻¹ DW) were also abundant. All other phenolic compounds were present in significantly lower amounts. Among the flavonoids, the flavanones naringenin and hesperetin were detected in amounts up to 0.73 and 0.41 mg kg⁻¹ DW, respectively. The flavone apigenin was detected at ≤ 0.30 mg kg⁻¹ DW. Supplementary Table 2 contains concentrations of phenolics, sugars and organic acids by replicate, variety, cropping system, and year in maize grain for which quantitative data could be obtained.

Soluble sugars identified in maize grain samples (Supplementary Table 2) were the disaccharide sucrose, the monosaccharides glucose and fructose, and the sugar alcohols mannitol and sorbitol. Two more compounds, one monosaccharide and one disaccharide, were visible in HPLC-PAD chromatograms (Supplementary Figure 4); however, due to the lack of appropriate standards they were not unambiguously identified. The amounts of these unidentified saccharides in samples were calculated relatively, based on calibration curves for glucose and sucrose. The amount of sucrose in samples varied between 5.50 and 52.50 g kg⁻¹ DW, depending on the maize variety, cropping system, and year. The second most abundant soluble sugar in maize grain was fructose with concentrations ranging between 0.04 and 2.00 g kg⁻¹ DW, while glucose was detected between 0.04 and 1.41 g kg⁻¹ DW. The sugar alcohols mannitol and sorbitol were detected at 0.31 and 0.55 mg kg⁻¹ DW, respectively.

Organic acids identified and quantified in maize grain samples were oxalic, citric, tartaric, malic, iso-citric, formic, and fumaric acids (Supplementary Table 3). A representative HPLC-DAD chromatogram is shown in Supplementary Figure 5. The most abundant organic acid was tartaric acid, detected at 2.19 g kg⁻¹ DW, followed by formic acid, detected at ≤ 1.68 g kg⁻¹ DW, oxalic acid detected at ≤ 1.08 g kg⁻¹ DW, and iso-citric acid detected at ≤ 1.01 g kg⁻¹ DW. Malic, citric, and fumaric acids were present in significantly lower concentrations.

Quantitative data for 30 compounds detected in maize grain for which standards were available (15 phenolics, 1 phytohormone, 7 sugars, and 7 organic acids) were statistically analyzed. Variance decomposition showed that fixed effects (cropping system, maize variety, soil properties, and their interactions) explained little of the total variance in measured metabolites except for maize variety (Table 2). Importantly, cropping system was associated with very little of the variation in mean levels of any metabolites. Cropping system explained <7% of the variation among all 30 metabolites and 2% or less of the variation with 25 of the metabolites. Available soil fertility variables were not predictive (results not presented). Maize variety explained most of the variability associated for fixed effects with almost all metabolites. With certain metabolites (tartaric, isocitric, fumaric, formic, malic, unknown disaccharide, ferulic, caffeic) maize variety explained a sizable effect (>20%) of variation. In contrast to fixed effects, random effects, or naturally occurring impacts (year, interactions of year with fixed effects), explained a large percentage of the explainable variance. Of the 30 metabolites, 16 (including all sugars) had a sizable year effect (> 20% of the total variance). However, for most metabolites, the largest proportion of the total variance was unexplained (i.e., residual error), indicating that either there are other important and unmeasured factors influencing metabolite levels or that there was considerable biological variability in the system such as soil moisture and soil fertility factors that can vary substantially within fields.

In tests of significance, cropping system affected 9 of 30 metabolites and maize variety affected 11 of 30 metabolites (Table 2). Only one metabolite, unidentified disaccharide, was significantly affected by the cropping system × maize variety interaction. Tests of significance were also performed on soil fertility variables as cropping systems can affect soils. Soil organic matter impacted no metabolite while soil pH, P, and K impacted 4 to 6 metabolites (mannitol [K, pH], sucrose [pH], p-hydroxybenzoic [P], caffeic [P], vanillic [K, P], vanillin [P], ferulic [K, P], coniferyl aldehyde [P], oxalic [pH], citric [pH], malic [K], iso-citric [pH]; data not shown). Examining these results by metabolite, we found that two metabolites were affected by three fixed effect factors, eight metabolites were affected by two fixed effect factors, 11 by one fixed effect factor, and nine were unaffected by any of the fixed effect factors (Table 2). Thus, levels of 2/3 of the metabolites in maize grain tested responded to at least one potential cropping system management factor. However, while these changes in metabolite means were statistically significant, they accounted for very little of the total variance. Thus, most traditional crop production management practices represented among the five FSP cropping systems, such as tillage, crop rotation, fertility management, and pesticide use are poor candidates for manipulating metabolite levels in maize grain.

Separate quantitative data subsets for organic and conventional maize varieties were tested for the significance of cropping system impacts on the mean content of metabolites (Figure 1) as maize genetics (variety selection) can impact some metabolites substantially. Results indicate that cropping system is a significant factor determining the content of glucose, fructose and the unknown monosaccharide in the organic variety, and the unknown monosaccharide in the conventional variety. Among phenolics, the content of naringenin and luteolin in the conventional variety were significantly influenced by the cropping system. Other phenolic compounds, including the most abundant phenolic acids (ferulic acid, p-coumaric acid, and vanillin) and formic acid, were not influenced by the cropping system in either organic or conventional varieties.

The variance decomposition and results of F-tests for significance on the stepwise selected variables were largely consistent (i.e., fixed effects that explained a relatively large percent of the variance attributable to fixed effects were often significant). However, there were also instances where the two sets of results appeared contradictory. For example, there were four cases where the estimated percent contribution to the total variance was zero, yet that variable was deemed significant, and in several cases seemingly large percentages were not significant. Part of this, especially for low percentages, may have been due to the way the variables were constructed. For example, pH in the F-tests was a one degree of freedom linear effect; in the variance decomposition it was a four degree of freedom factor. However, for other effects, it appears that the parameters were simply poorly estimated within the model, i.e., with large standard errors. The variance decomposition does not include the accuracy of each estimated percent, and some were likely poorly estimated.

Metabolomic correlation analysis has been used to study impacts of genetic or environmental changes on metabolomes of plants and other organisms (Steuer, 2006; Szymanski et al., 2009; Fukushima et al., 2011; Fatima et al., 2016), and serves as a proxy indicating how metabolic pathways change under different conditions. Metabolomic correlation analysis was applied here to determine impacts of naturally occurring, cropping system, and crop variety impacts on the maize physiological state to complement the means analysis presented above. Metabolite interrelationships with significant changes in correlation (at α = 0.01) are given in Supplementary Table 2, except for six metabolites showing no changes in correlation and few significant changes among the various organic systems. The largest number of changes in metabolite correlations occurred among years, likely a response to strong environmental differences plants experienced with different years, such as rainfall. There were also a fair number of changes in correlations in the maize metabolome when conventional and organic varieties were compared. Changes in correlations among metabolites when organic systems were compared were small and few. When comparing Org2 to Org6 or Org3 to Org6, there were just two or three decreases in correlation. Between Org2 and Org3, there were two increases involving sugars and yield. More changes in correlation between metabolites occurred when comparing CT to the collective organic cropping systems (Org2, Org3, and Org6 collectively) or to the NT system. Importantly, correlation analysis showed that correlations between metabolites were not fixed, likely due to differences in regulation of biosynthetic pathways responsible for production of the evaluated compounds (e.g., see above the variety of biosynthetic pathways involved in production of phenolics detected with this present study) with different environmental conditions and crop varieties.

Metabolic fingerprinting of methanol extracts of maize grain using UHPLC coupled with MS detection identified a total of 90 compounds, the majority of which were phenolics. Phenolics from grain and other plant dietary sources are important phytochemicals that possess human health beneficial effects including antioxidant and anti-inflammatory activities that reduce certain chronic diseases (Liu, 2013; Zhang B. et al., 2020). Ferulic acid, p-coumaric acid, and vanillin were the predominant phenolic compounds detected, consistent with maize grain being considered an important source of ferulic acid and fiber in the human diet (Mathew and Abraham, 2004; Vitaglione et al., 2008). Other nutritional components of the maize grain metabolome, 7 carbohydrates and 7 organic acids, were also detected and analyzed. Macronutrients, minerals, vitamins, and essential amino acids and fatty acids important to human nutrition were not the subject of this investigation.

Means analysis on a subset of the metabolites we evaluated (15 phenolics, 1 phytohormone, 7 sugars, and 7 organic acids) indicated that the 5 cropping systems at FSP (two conventional and 3 organic systems) could significantly impact quantities of certain of these metabolites, and by extension other maize-grain nutrients important to human nutrition. However, cropping system effects explained very little of the total explainable variability, indicating that cropping system management tactics cannot be employed to consistently manipulate maize nutrient levels. By contrast, natural environmental influences (growing year) had a much greater impact on quantities of nutrients analyzed in maize grain. Correlation analysis of this same subset of compounds was consistent with the means analysis. Importantly, this is the first thorough analysis of the impact of cropping system on the maize metabolome where experiments were performed in well characterized soils where cropping systems had been in place long enough to alter soil edaphic and microbial properties (Green et al., 2005; Cavigelli et al., 2008; Spargo et al., 2011; Maul et al., 2019; Kepler et al., 2020; Schmidt et al., 2022). A second study comparing impacts of one organic and one conventional cropping system on a complementary set of maize grain metabolites (fatty acids, sugars, sugar alcohols, organic acids, amino acids and amines) support our findings as cropping systems impacts only led to minor changes in metabolite profiles and were subordinate to naturally occurring environmental impacts (Röhlig and Engel, 2010). However, the experiment was conducted only 1 year for this comparison. Our results are also consistent with the litany of studies finding no clear evidence for differences in nutritional quality between foods grown under conventional and organic production systems (Giampieri et al., 2022).

Large impacts of growing year on metabolite means and metabolite correlations were likely a response to strong environmental (non-management related) differences maize plants experienced during different years. This is consistent with findings from several metabolomic studies with maize looking at growing year impacts (Röhlig et al., 2009; Anttonen et al., 2010; Röhlig and Engel, 2010; Frank et al., 2012). Candidate factors in the present study that possibly varied yearly contributing to growing season impacts include rainfall, temperature, and light intensity. Substantial impacts have been found with geographically distant fields that differed in mean temperature, duration of day/night, average rainfall, and planting and harvesting date (Röhlig et al., 2009; Anttonen et al., 2010; Röhlig and Engel, 2010; Skogerson et al., 2010; Asiago et al., 2012; Frank et al., 2012; Baniasadi et al., 2014; Chen et al., 2016); which could influence the aforementioned factors.