An Arabidopsis GCMS chemical ionization technique to quantify adaptive responses in central metabolism

Abstract

We present a methodology to survey central metabolism in ¹³CO₂-labeled Arabidopsis (Arabidopsis thaliana) rosettes by ammonia positive chemical ionization–gas chromatography–mass spectrometry. This technique preserves the molecular ion cluster of methyloxime/trimethylsilyl-derivatized analytes up to 1 kDa, providing unambiguous nominal mass assignment of >200 central metabolites and ¹³C incorporation rates into a subset of 111 from the tricarboxylic acid (TCA) cycle, photorespiratory pathway, amino acid metabolism, shikimate pathway, and lipid and sugar metabolism. In short-term labeling assays, we observed plateau labeling of ∼35% for intermediates of the photorespiratory cycle except for glyoxylate, which reached only ∼4% labeling and was also present at molar concentrations several fold lower than other photorespiratory intermediates. This suggests photorespiratory flux may involve alternate intermediate pools besides the generally accepted route through glyoxylate. Untargeted scans showed that in illuminated leaves, noncyclic TCA cycle flux and citrate export to the cytosol revert to a cyclic flux mode following methyl jasmonate (MJ) treatment. MJ also caused a block in the photorespiratory transamination of glyoxylate to glycine. Salicylic acid treatment induced the opposite effects in both cases, indicating the antagonistic relationship of these defense signaling hormones is preserved at the metabolome level. We provide complete chemical ionization spectra for 203 Arabidopsis metabolites from central metabolism, which uniformly feature the unfragmented pseudomolecular ion as the base peak. This unbiased, soft ionization technique is a powerful screening tool to identify adaptive metabolic trends in photosynthetic tissue and represents an important advance in methodology to measure plant metabolic flux.

A soft ionization gas chromatography–mass spectrometry technique yields insights into jasmonate-induced metabolic changes in primary metabolism of ¹³C-labeled Arabidopsis rosettes.

Introduction

Gas and liquid chromatography–mass spectrometry (GCMS and LCMS) have become indispensable tools for studying plant metabolism (Cajka and Fiehn, 2016; Perez de Souza et al., 2017). The addition of stable isotopes to the experimental system further broadens the analytical capabilities of metabolomics approaches to characterizing metabolic flux (Freund and Hegeman, 2017). ¹³CO₂ whole plant labeling and subsequent LC and GCMS analysis of isotopically enriched metabolite pools (i.e. isotopolog metabolomics or fluxomics) is a powerful technique for monitoring carbon flux in plant metabolic networks (Tcherkez et al., 2009; Hasunuma et al., 2010; Szecowka et al., 2013) and quantitatively characterizing metabolic responses to stress. This technique depends on the measurement of isotopologs, molecules that differ in their number of heavy isotopes without explicit reference to their position. In contrast, isotopomers constitute distinct positional isomers of the same isotopically modified element (Seeman and Paine, 1992). The former can be readily quantified by MS based on a shift of the corresponding peak in the mass spectrum. The rate of appearance of ¹³C tracer in downstream metabolite pools provides the raw data for calculating fluxes in primary metabolism as well as the rates, directions, and magnitudes of multiple metabolic pathways simultaneously (Ma et al., 2014). Conventional electron impact GCMS is generally the preferred technique in untargeted metabolomics due to its high chromatographic resolution, sensitivity, reproducibility, and broad metabolome coverage (Roessner et al., 2000; Hummel et al., 2007; Kanani et al., 2008). However, its utility for isotopic labeling studies is limited by the fragmentation characteristic of hard ionization, where degradation of the molecular ion cluster results in loss of information needed to calculate label incorporation.

One of the advantages of LCMS for isotopolog metabolomics is the nondestructive nature of electrospray ionization, which almost uniformly preserves the molecular ion cluster during the ionization step and facilitates quantification of isotopolog peaks needed for exact labeling calculations of primary plant metabolites obtained through whole plant labeling experiments (Wright et al., 2014; Tsugawa et al., 2019; Xu et al., 2021). Indeed, a combination of GCMS and LCMS was frequently employed in early labeling studies to account for the hard ionization problem (Kind and Fiehn, 2010). Comparable soft ionization in GCMS can be obtained with ammonia positive chemical ionization (NH₃-PCI) (Warren, 2013). Among the conventional reagent gases employed in CI, ammonia has the highest proton affinity (845 kJ·mol⁻¹) (Keough and DeStefano, 1981), a property which results in low energy transfer to receptor molecules and minimal fragmentation during the ionization step. A mixture of [M + H]⁺ and [M + NH₄]⁺ product ions are typically formed whose masses differ by 17 Da, providing unambiguous nominal molecular mass information for each analyte peak. NH₃ has previously been employed as a reagent gas in GCMS to exploit this soft ionization property. It has therefore found uses in the identification of nitrogen containing compounds in coal extracts (Buchanan, 1982) and in the analysis of lipids (Crawford and Plattner, 1983), carbohydrates (Horton et al., 1974), essential oils (Lange and Schultze, 1992), and alkaloids (McCoy et al., 1983). More recently, the soft ionization properties of NH₃-PCI have also led to its application in GCMS-based ¹³C-metabolic flux analysis in yeast and mammalian systems (Yang et al., 2006; Chu et al., 2015; Okahashi et al., 2019). This technique has thus far received little attention in plant metabolism research, despite its potential to improve metabolite annotation and facilitate label quantification in isotopolog metabolomics analysis, especially for recalcitrant metabolites.

Many metabolic processes in plants remain inaccessible to all but the most highly specialized analytical techniques. For instance, photorespiration remains a challenging target for direct quantification and flux analysis despite its centrality in nitrogen and carbon assimilation (Bloom, 2015). Typically, photorespiratory rate is estimated from gas exchange parameters (Sharkey, 1988; Busch, 2013), and few comprehensive reports exist describing the direct detection of its intermediates (Arrivault and Obata, 2017). These difficulties may be due to the high charge density and low molecular weight of the C₂ and C₃ organic acids of the photorespiratory cycle, which result in poor retention on both LC and GC stationary supports.

Measuring flux through the shikimate pathway, which yields chorismate for aromatic amino acid (AAA) formation, is likewise a challenging endeavor which limits our current ability to investigate its complex regulation. An LCMS method to estimate stable isotope labeling of shikimate was implemented to confirm arogenate as the dominant postchorismate route to Phe and Tyr (Maeda et al., 2010) and identify phenylalanine ammonia lyase regulation as the main mechanism for controlling flux through the shikimate pathway (Lynch et al., 2017). Improvements in shikimate isotopolog detection would facilitate additional inquiries into understanding regulation of this pathway. Another example can be found in attempts to identify substrate supply to the 2C-methyl-D-erythritol-4-phosphate (MEP) pathway, which supplies the precursors to many photosynthetic co-factors in green tissue (Phillips et al., 2008). A robust, untargeted GCMS methodology that facilitates isotopolog analysis without the limitations of hard ionization would therefore impact many areas of plant research.

Here, we present an untargeted NH₃-PCI-GCMS analytical methodology for quantifying ¹³C isotopologs of GC-amenable central metabolites in plants labeled under physiological conditions. We provide complete chromatographic, mass spectral, and kinetic labeling data for intermediates from photorespiration, the citric/tricarboxylic acid (TCA) cycle, the shikimate pathway, and carbohydrate and amino acid metabolism. The potential of this method to provide direct input into mathematical models of metabolic flux is discussed.

Results

Soft ionization GCMS provides nominal mass and isotopolog ratios of derivatized leaf metabolites

To evaluate the usefulness of this analytical technique for profiling plant metabolism, we first prepared methanolic extracts of Arabidopsis (Arabidopsis thaliana) rosette tissue and analyzed the derivatized polar extracts by GCMS using electron ionization (EI), methane positive chemical ionization (CH₄-PCI), and NH₃-PCI. Comparison of their total ion chromatograms (TICs) showed that the three methods produced similar chromatographic profiles (Figure 1) though with clear differences in response factor between ionization types for certain metabolites. Using retention index (RI) to align chromatographic peaks obtained in different ionization modes, the EI and PCI spectra of each analyte could be compared. This evaluation indicated that although overall sensitivity of derivatized polar metabolites in both PCI modes was reduced compared with EI (∼20-fold for NH₃, >100-fold for CH₄), fragmentation was dramatically reduced, and improved retention of the molecular ion cluster was evident, consistent with previous literature reports (Yang et al., 2006; Chu et al., 2015). Initial PCI scans of low mass regions (m/z 55–650) revealed a trimethylsilyl fragment ion in CH₄-PCI (m/z 73) and NH₃-PCI (m/z 90) that was <20% of the pseudomolecular ion in most cases (Supplemental Figure S1), and subsequent full scans were restricted to m/z 150–950 to improve the sampling rate and focus detector time on intact pseudomolecular ions. In EI mode, without further deconvolution, an average of 328 ± 20 features could be detected in TICs of control leaf extracts acquired in untargeted scans (m/z 50–550). Analysis of the same samples by NH₃-PCI (m/z 150–950) and CH₄-PCI (m/z 150–950) under identical chromatographic conditions revealed 209 ± 12 and 115 ± 59 features, respectively. A total of 25 features detected by NH₃-PCI could not be detected by EI. With respect to MeOX/TMS-derivatized Arabidopsis leaf metabolites, NH₃-PCI was superior to CH₄-PCI in terms of both sensitivity and preservation of the intact analyte molecule. We therefore optimized this technique further as a tool for untargeted plant GCMS metabolomics analysis. Of the >200 features detected by NH₃-PCI in control tissue extracts (Supplemental Table S1), we provide the full NH₃-PCI spectra (m/z 150–950) for the 100 most robust metabolic features across ∼100 leaf extracts (Supplemental Data set 1). Within this subset, 78 were unambiguously identified through a combination of mass spectral library searches of their corresponding EI spectra, comparison of EI to PCI spectra, RI matching, and comparison to authentic standards where possible (Supplemental Table S1). All 78 annotated analytes detected by NH₃-PCI corresponded to the expected small molecules of primary metabolism involved in respiration, photorespiration, amino acid biosynthesis, carbohydrate metabolism, and phenolic, lipid, and sterol biosynthesis, in agreement with previously described EI-GCMS analysis of Arabidopsis leaf extracts (Fiehn, 2006; Gu et al., 2011). Metabolite annotation was also facilitated by pseudomolecular ion detection by NH₃-PCI, which generally provided unambiguous nominal molecular mass information that complemented EI spectral information obtained for the same peaks after changing ion sources. Chemical ionization-based GCMS analysis of phosphorylated sugars using methane (Chu et al., 2015) and ammonia (Yang et al., 2006; Okahashi et al., 2019) reagent gases has been reported. While we detected signals for these metabolites using purified standards, our protocol did not permit reliable quantification in plant tissues needed to accurately calculate label incorporation.

Figure 1.

Metabolite profiles of an Arabidopsis leaf extract analyzed by GCMS following MeOX/TMS derivatization in three different ionization modes. The total ion abundance of each mode is displayed as the maximum peak height obtained from the same sample injected under identical conditions to indicate the relative sensitivity of each mode. The gap on the y-axes in NH3-PCI and CH4-PCI represents discontinuous axes.

In NH₃-PCI mode, the molecular ion formed the base peak nearly 100% of the time (Supplemental Data set 1). Fragment ions were rarely present above 10% of the monoisotopic peak intensity. Most analytes yielded two separate pseudomolecular ion clusters consisting of the protonated form ([M + H⁺]) and the ammonium adduct ([M + ${\text{NH}}_4^{+}$]). In some cases, the diammonium adduct was also present. The [M + H]⁺/[M + NH₄]⁺ ratio depends on the proton affinity of the recipient molecule relative to the ammonia reagent gas. A ranking of 50 derivatized analytes by these ratios partly recapitulated the expected order of the original metabolites with reducing sugars, now bearing a basic nitrogen atom from the methyloxime group, demonstrating the highest proton affinity, followed by amino acids, and organic acids showing the lowest (Figure 2). The nonreducing sugar sucrose, which does not acquire a methyloxime group during derivatization, demonstrated very low proton affinity. For calculating label incorporation, the molecular ion cluster with the highest abundance was generally used.

Figure 2.

Ratios of the [M + H]⁺/[M + NH₄]⁺ pseudomolecular ions formed during NH₃ chemical ionization of 50 derivatized Arabidopsis leaf metabolites, ranked in order from highest to lowest proton affinity (left to right). The ratio increases with increasing proton affinity of the recipient molecule. The proton affinity of derivatized analytes largely reflects the relative proton affinity of their metabolite precursors with carbohydrates showing the highest affinity, followed by amino acids, and organic acids showing the lowest. * indicates IS. β-OGP, β-octylglucopyranoside. Error bars show the standard error of 7–106 independent biological replicates. Fructose 1 and 2 refers to methyloxime cis/trans isomers.

Accuracy of isotope ratio data obtained from NH₃-PCI-GCMS

Reliable calculations of label incorporation into plant metabolites depend on accurate quantification of isotopologs representing its differentially labeled forms. Prior to applying this method to the analysis of labeled plant tissue, we evaluated the accuracy of isotope ratio measurements obtained with NH₃-PCI-GCMS. To do this, we compared the observed and expected relative isotope peak intensities of the synthetic chlorinated compound chlorpyrifos (Figure 3A) in scan and selected ion monitoring (SIM) mode (Supplemental Table S2). Due to high natural abundance of ³⁷Cl isotopes, chlorpyrifos (C₉H₁₁Cl₃NO₃PS) produces a distinctive secondary isotope pattern (Figure 3B) which we exploited to evaluate the precision and accuracy of this method. All three expected pseudomolecular ions ([M + H]⁺, [M + NH₄]⁺, and [M + N₂H₇]⁺) predicted by the established ionization mechanism of ammonia (Figure 3C) were readily detected in scans. Analysis of the [M + H]⁺ molecular ion cluster of a chlorpyrifos standard (n = 8) showed that the observed relative abundances of its [M + 1] to [M + 6] secondary isotopes (m/z 366.9–372.9) were within 0.3% of expected values on average (Supplemental Table S2). In scan mode, we observed a higher deviation from theoretical values for highly abundant ions (m/z 366.9 and 370.9 corresponding to the M + 0 and M + 4 ions) relative to the base peak (m/z 368.9) than for lower abundance ions (M + 5 and M + 6), which typically produced values within 0.03% of expected. A SIM method monitoring the same M + 0 through M + 6 ions of the [M + H]⁺ pseudomolecular ion cluster produced the opposite effect: the high abundance ions matched expected values most closely with the highest errors seen for low abundance ions. Despite these differences between acquisition modes, their magnitude was small, and no relative isotope measurement deviated from expected values by >0.84%.

Figure 3.

Assessment of isotope ratio accuracy acquired with NH₃-PCI-GCMS. A, The analytical trueness of the technique was evaluated by comparison of secondary isotope peak intensities of a model chlorinated compound (chlorpyrifos) to predicted values. B, Close-up of characteristic protonated molecular ion cluster ([M + H]⁺) of chlorpyrifos used to evaluate quadrupole calibration. C, Formation of reactive gas species during NH₃-PCI.

Isotopolog profiling by NH₃-PCI-GCMS to monitor ¹³C enrichment of central metabolite pools

By comparing spectra from untargeted scans in labeled and unlabeled Arabidopsis leaf extracts, we identified 111 metabolic features in time-course labeled plants where isotopic enrichment of at least 1% above control tissue levels could be detected (Supplemental Table S1). Using this information, we designed a SIM method focusing on their molecular ions and isotopologs. We then used this revised method to analyze leaf extracts from a cohort of time-course labeled plants (n = 66) exposed to ¹³CO₂ between 3 min and 2 h. Pool size data were also acquired using external calibration curves. By modeling the time-course labeling data to an exponential rise to maximum (F = C + A(1−e^-kt), where F is fractional labeling, C is the natural isotopic abundance, A is plateau labeling level, t is time in minute, and k is the kinetic rate constant), we estimated their kinetic constants, half-lives, and plateau labeling (Table 1). The robust standard deviation of the residuals (RSDRs) indicated a highly reliable fit in all cases (Motulsky and Brown, 2006).

Table 1.

Kinetic parameters of selected polar metabolites detected by NH₃-PCI-GCMS

					Control				MJ
		Nominal	Molecular		Labeling				Labeling
Metabolite	Analyte	mass	Ion	Adduct	plateau	k	t 1/2 (min)	RSDR	plateau	k	t 1/2 (min)	RSDR
Photorespiration
Glycolate	2TMS	220	238	[M + NH4]+	0.36	0.101	6.85	0.075	0.31	0.144	4.81	0.072
Glyoxylate	MeOX/TMS	175	193	[M + NH4]+	0.05	0.181	3.83	0.026	0.05	0.314	2.20	0.024
Glycine	3TMS	291	292	[M + H]+	0.51	0.056	12.30	0.080	0.38	0.116	5.98	0.063
Serine	2TMS	249	250	[M + H]+	0.51	0.039	17.56	0.076	0.50	0.037	18.52	0.100
Glycerate	3TMS	322	340	[M + NH4]+	0.42	0.049	14.12	0.063	0.44	0.049	14.09	0.079
Organic acids
Pyruvate	MeOX/TMS	189	207	[M + NH4]+	0.25	0.044	15.58	0.047	0.20	0.052	13.30	0.063
Shikimate	4TMS	462	480	[M+NH4]+	0.19	0.012	60.24	0.009	0.32	0.007	102.31	0.023
TCA cycle
Citrate	4TMS	480	481	[M + H]+	0.04	0.116	5.97	0.035	0.45	0.010	67.54	0.055
Fumarate	2TMS	260	278	[M + NH4]+	0.07	0.096	7.20	0.053	0.04	0.194	3.57	0.036
Malate	3TMS	350	368	[M + NH4]+	0.07	0.013	53.03	0.011	0.06	0.010	66.53	0.015
Succinate	2TMS	262	280	[M + NH4]+	0.01	0.021	32.69	0.006	0.00	0.385	1.80	0.003
α-Ketoglutarate	MeOX/2TMS	319	337	[M + NH4]+	0.02	0.383	1.81	0.015	0.01	0.386	1.80	0.008
N cycle
Aspartate	2TMS	277	278	[M + H]+	0.08	0.011	61.32	0.008	1.00	0.001	865.06	0.005
Free sugars
Sucrose	8TMS	918	936	[M + NH4]+	0.81	0.007	103.99	0.020	0.68	0.008	84.24	0.029
Fructose	MeOX/5TMS	569	570	[M + H]+	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.	n.d.
Trehalose	8TMS	918	936	[M + NH4]+	0.05	1.019	0.68	0.050	0.08	3.796	0.18	0.047
Glucose	MeOX/5TMS	569	570	[M + H]+	0.09	0.005	126.96	0.011	0.15	0.009	77.72	0.045

TMS, trimethylsilyl derivative; MeOX, methyl oxime derivative; k, exponential rate constant; t_1/2, half-life of metabolite; MJ, methyl jasmonate treatment; TCA, tricarboxylic acid cycle; RSDR, robust standard deviation of the residuals for fractional labeling curve; n.d., not done.

We first analyzed the intermediates of photorespiration (Figure 4) with the exception of 2-phosphoglycolate and hydroxypyruvate, neither of which could be reliably detected with this method. Photorespiratory intermediates quickly accumulated ¹³C in the initial minutes of labeling, and most intermediates reached plateau incorporation levels of 40 ± 6% atom labeling by 30 min (Figure 4, A–E). Exceptionally, glyoxylate reached a plateau of only 4% (Figure 4B). Comparison of their steady-state pool sizes indicated significant differences in their molar concentrations (Figure 4H). For instance, glycerate and glycolate were present in 2.9- and 1.5-fold molar excess over glyoxylate (P < 10⁻⁸; two-tailed student’s t test), while glycine was present at a 2.3-fold lower concentration than glyoxylate (P < 10⁻⁷) but reached the highest plateau labeling of the photorespiratory intermediates (51%; Table 1). Serine was present at the highest concentration of all photorespiratory intermediates (53-fold molar excess over glyoxylate), likely reflecting additional roles outside of photorespiration. Adjusting for pool size, rate of appearance of ¹³C label was judged to be ∼0.1 pmol·mg⁻¹·min⁻¹ for glyoxylate and 1.7, 5.1, 432, and 3.9 pmol·mg⁻¹·min⁻¹ for glycolate, glycerate, serine, and glycine, respectively (Supplemental Figure S2). This observation suggested that under steady-state conditions, flux through glyoxylate appears much lower than that of the other photorespiratory intermediates. The calculated half-lives of photorespiratory intermediates ranged from 5.8 to 6.0 min for glyoxylate and glycolate to 14 min for glycerate (Table 1), which are slightly slower than previously reported values (Szecowka et al., 2013).

Figure 4.

¹³C label incorporation into photorespiratory intermediates and quantification of pool size. A–E, Fractional labeling (0–1.0, indicating % atom labeling) of intermediates in Arabidopsis leaf over a 2-h time-course shows a plateau in labeling between 0.35 and 0.45 for most intermediates with the exception of glyoxylate, which remains mostly unlabeled. Each point represents a single rosette stage Arabidopsis plant incubated for the indicated time in a ¹³CO₂ atmosphere. F, Overlay of curves fitted to an exponential rise to maximum. G, TIC showing photorespiratory intermediates and pyruvate in a leaf extract. H, Pool sizes in leaf tissue based on comparison to external standard curves, corrected for recovery of an IS cocktail (see “Materials and methods”). For estimated plateau levels and rate constants, see Table 1. DW, dry weight. Box plot elements are as follows: Data points outside the bounds Q1 − 1.5× interquartile range or Q3 + 1.5× interquartile range were removed as outliers, and for the remainders, the center line represents the median, upper and lower box limits indicate quartiles 1 and 3, respectively, and whiskers show the minima and maxima of the dataset. Sample sizes following outlier removal: glycerate (61), glycine (61), glycolate (64), glyoxylate (63), and serine (58).

Using the same samples, we next interrogated the labeling and concentrations of TCA intermediates and other organic acids of central metabolism (Figure 5). Citrate and α-ketoglutarate were present at low concentrations (90 and 332 pmol·mg⁻¹ DW, respectively), and no intermediate from the TCA surpassed 10% fractional labeling under our time-course conditions. Succinate and malate were present at 17- and 38-fold higher concentrations, respectively, compared with citrate while fumarate was present at 22.4 nmol·mg⁻¹ DW, 249 times the molar concentration of citrate. Labeling of α-ketoglutarate, succinate, and fumarate only reached maxima of 1%–2%, respectively, suggesting there was little to no active turnover of these intermediates in leaves under illuminated conditions. However, malate, which had a pool size more than double succinate, and citrate showed labeling maxima of 9% and 12%, respectively. Aspartate, which is interconverted with oxaloacetate in the mitochondrion (Schultz and Coruzzi, 1995) reached a plateau of 5% labeled. The differential degree of plateau labeling we observed here is consistent with noncyclic flux through the TCA cycle (Sweetlove et al., 2010). The labeling kinetics of these intermediates indicated that succinate and malate appear to have half-lives of 27 and 64 min while citrate was much shorter at ∼9 min, consistent with a presumed role of citrate in export to the cytosol for NADPH production in illuminated leaves (Igamberdiev, 2020) or conversion to acetyl-CoA by ATP:citrate lyase (Fatland et al., 2002). Overall, intermediates of the TCA cycle were present at much higher concentrations than intermediates of photorespiration but reached much lower fractional labeling levels. This is generally consistent with a lower flux through the TCA cycle during illuminating conditions (Tcherkez et al., 2009).

Figure 5.

¹³C label incorporation into nonphotorespiratory organic acids of central metabolism and quantification of pool size. Fractional labeling (0–1.0) indicates % atom labeling. In time-course labeled Arabidopsis leaves, the turnover of shikimate and pyruvate (A and B) is higher than that of TCA intermediates (C–G) or aspartate (H) under illuminating conditions. I and J, Pool sizes in leaf tissue based on comparison to external standard curves, corrected for recovery of an IS cocktail (see “Materials and methods”). For estimated plateau levels and rate constants, see Table 1. Box plot elements are as follows: Data points outside the bounds Q1 − 1.5× interquartile range or Q3 + 1.5× interquartile range were removed as outliers, and for the remainders, the center line represents the median, upper and lower box limits indicate quartiles 1 and 3, respectively, and whiskers show the minima and maxima of the dataset. Sample sizes following outlier removal: citrate (62), α-ketoglutarate (61), malate (64), succinate (63), fumarate (64), pyruvate (65), and shikimate (54). Alpha-KG, alpha-ketoglutarate.

Plastidic pyruvate and shikimate are involved in terpenoid precursor formation (Rohmer et al., 1996) and AAA biosynthesis (Tohge et al., 2013), respectively, among several other roles. We were able to detect well resolved peaks for each corresponding to the ${\text{NH}}_4^{+}$ adduct of the MeOX/TMS derivative for pyruvate (m/z 207) and the ${\text{NH}}_4^{+}$ adduct of the 4TMS derivative of shikimate (m/z 480) (Figure 6). The predicted labeling plateaus were similar for both (22%–23%), but a visual inspection indicated shikimate had not reached isotopic steady state after 2 h of labeling. Its labeling was slower than other organic acids, but the plant-to-plant variability in shikimate flux was minimal in comparison with plant-to-plant variability in fumarate and citrate labeling. Its projected half-life was 82 min (pool size 304 pmol·mg⁻¹ DW), while pyruvate was at steady state in ∼45 min with a pool size three-fold as large (half-life 10 min, 911 pmol·mg⁻¹ DW). Pyruvate labeling kinetics outpaced that of citrate, suggesting that in the light, pyruvate may originate from multiple sources (Figure 5).

Figure 6.

Detection of ¹³C-labeled shikimate in Arabidopsis leaf extracts by NH3-PCI-GCMS. A, Extracted ion chromatogram (m/z 480.2) showing detection of shikimate in plant extracts. B, NH₃-PCI spectrum of trimethylsilylated shikimate obtained from an authentic standard (top) or unlabeled control plant (mirrored) in scan mode. Arrows indicate the protonated (H⁺) and ammonium adduct (${\text{NH}}_4^{+}$) pseudomolecular ion clusters. C, Close-up view of shikimate pseudomolecular ion cluster (ammonium adduct) from the mass spectrum of an unlabeled control plant. D, the same mass spectrum from a plant subjected to 2 h ¹³CO₂ labeling. (C) and (D) were acquired in SIM mode (m/z 480.2–487.2).

Free sugars were considerably more abundant than most organic acids, but only sucrose incorporated appreciable label under illuminating conditions (Figure 7). Sucrose displayed rapid labeling kinetics, but the large pool size made an accurate assessment of its labeling plateau difficult (estimated at 63% with a half-life of 76 min and pool size of 43 nmol·mg⁻¹ DW. The monosaccharides fructose and glucose were present at comparable concentrations as sucrose (34 and 52 nmol·mg⁻¹ DW, respectively) though with extremely long half-lives (4 and 12 h, respectively; Table 1). This indicated that during the day, as expected, these sugars do not undergo rapid turnover in the light. Free sugars could be arbitrarily grouped into high abundance (sucrose, glucose, and fructose) and low abundance (mannose, galactose, and trehalose, detected at 560, 761, and 78 pmol·mg⁻¹ DW; Figure 7F). In the case of the latter group, low ¹³C incorporation rates precluded prediction of turnover rates, although free trehalose reached plateau labeling levels of 4%. Labeling into myo-inositol was too low to fit to an exponential curve, although a linear regression of fractional labeling suggested a rate of label incorporation of 0.7% per hour during a 2-h time-course, suggesting an extremely low rate of turn-over for this metabolite under illuminating conditions.

Figure 7.

¹³C Label incorporation into free sugars and quantification of pool size. Fractional labeling (0–1.0) indicates % atom labeling. In time-course labeled Arabidopsis leaves, the sucrose pool (A) rapidly assimilates isotopic label compared to other free sugars (B–D). E, An overlay of their fitted curves. F, Pool sizes in leaf tissue based on comparison to external standard curves, corrected for recovery of an IS cocktail (see “Materials and methods”). High abundance free sugars (sucrose, fructose, and glucose) are present at 100-fold molar excess over low abundance free sugars (trehalose, galactose, and mannose), which are plotted on the right vertical axis as indicated by the dashed line. For estimated plateau levels and rate constants, see Table 1. Box plot elements are as follows: Data points outside the bounds Q1 − 1.5× interquartile range or Q3 + 1.5× interquartile range were removed as outliers, and for the remainders, the center line represents the median, upper and lower box limits indicate quartiles 1 and 3, respectively, and whiskers show the minima and maxima of the dataset. Sample sizes following outlier removal: sucrose (58), fructose (57), glucose (64), trehalose (54), galactose (62), and mannose (64).

NH₃-PCI-GCMS is reportedly effective for the analysis of sugar phosphates in yeast and cultured human cells (Chu et al., 2015; Okahashi et al., 2019). These central metabolic intermediates are also of great importance to understanding plant metabolism, and we attempted to replicate their detection in our samples by NH₃-PCI-GCMS. The analysis of phosphorylated metabolic intermediates has more commonly been reported using tandem LC-MS/MS (Arrivault et al., 2009; González-Cabanelas et al., 2016; Bergman et al., 2021a, 2021b), where the near universal production of a common phosphate product ion (m/z 97 or 79) during collision induced dissociation provides a convenient means to quantify isotopologs from labeled sources (Wright et al., 2014). The NH₃-PCI-GCMS signals we observed for phosphorylated sugars in plant extracts were insufficient for reliable quantification, and we have therefore not included them in peak tables of the present method. Finally, major intermediates from lipid metabolism were detected in untargeted scans, mostly as intact molecular ions suitable for labeling calculations. These included palmitate, oleate, stearate, and major leaf sterols (Supplemental Data set 1 and Supplemental Table S1), but they were not investigated further.

Exogenous defense hormones induce changes to the pool size and flux of central metabolites

We tested the ability of this method to quantitatively characterize shifts in central metabolism by subjecting wild-type Arabidopsis plants to treatment with exogenous salicylic acid (SA), methyl jasmonate (MJ), or water (control) the day before labeling. Application of these defense hormones is known to induce changes to the concentrations of many intermediates of central metabolism (Bari and Jones, 2009; Wei et al., 2021), and the quantitative detection of these changes would provide additional proof of principle for this methodology.

When we compared the pool sizes of photorespiratory intermediates after MJ treatment, we observed significant increases in concentrations of early intermediates such as glycolate and glyoxylate and steep decreases in the later intermediates serine, glycine, and glycerate (Figure 8). Serine and glycine both declined ∼97% in MJ-treated plants while SA had the opposite effect, resulting in four- and six-fold increase in glycine and serine concentrations, respectively, over controls (Supplemental Table S3). While jasmonate signaling is known to downregulate glycine decarboxylase at the transcript (Jung et al., 2007) and protein levels (Cho et al., 2007), these results imply that it triggers a block of the transamination of glyoxylate to glycine. When we examined the rate of ¹³C incorporation into these pools, all but glycerate displayed a lower maximal labeling plateau following MJ treatment (Supplemental Figure S3 and Table 1). The decrease in pool size and labeling maxima of the latter intermediates suggest that flux through the photorespiratory pathway is negatively impacted by jasmonate treatment. This observation is consistent with the downregulating effect of this phytohormone on RUBISCO activity (Shan et al., 2011), which should lead to reduced carboxylation and oxygenation activity.

Figure 8.

Effect of exogenous hormones on the pool size of central metabolic intermediates as determined by NH₃-PCI-GCMS. Plants were extracted and analyzed 24 h after treatment with either100 µM MJ, 100 µM SA, or water (control). Quantification was accomplished by comparison to external standard curves with a correction to a cocktail of ISs (see “Materials and methods”). Box plots show median, upper and lower interquartile range while the whiskers indicate the minima and maxima of the dataset. N.S., no significant difference, N.D., not detected, *P < 0.05, based on a Student’s two-tailed t test. N = 66 (control), 16 (MJ), and 5 (SA).

The effect of SA and MJ on the TCA cycle, pyruvate, and shikimate revealed inverse effects of these defense hormones on the organic acids of central metabolism. Flux through the TCA cycle is highly reduced in the light (Tcherkez et al., 2005) and we observed low levels of ¹³C incorporation during time-courses in all treatments, complicating an accurate assessment of the response to phytohormones. In general, MJ caused decreases in respiration. Pyruvate displayed a decrease in both pool size and labeling plateau (Figure 8 and Supplemental Figure S3), and the decrease in concentration was even larger in SA-treated plants (∼60% reduction). Pyruvate was the only metabolite examined that responded to both SA and MJ in a similar fashion. Fumarate and malate concentration increased by 201% and 160%, respectively, following MJ treatment. Citrate and α-ketoglutarate, in contrast, were reduced in concentration following MJ treatment by 81% and 78% compared with controls, but the labeling plateau of citrate rose from 12% to 29%. This suggests the noncyclic flux through the TCA cycle seen under standard illuminating conditions (Figure 5) is altered by jasmonate signaling. SA-treated plants showed almost uniformly contrasting effects with decreases in fumarate, malate, and succinate and increases in citrate and α-ketoglutarate (Figure 8). The pool size of shikimate did not change in response to SA but declined 56% following MJ treatment (Figure 8). Nonetheless, the kinetics of shikimate labeling were unchanged (Supplemental Figure S3). Together, these observations suggest that MJ signaling inhibits TCA, while SA exerts the opposite effect. The shikimate pathway, in contrast, was affected to a lesser extent by these phytohormone treatments.

We next compared the concentrations of the free disaccharides sucrose and trehalose in MJ and control samples. Both sucrose and free trehalose declined by 29% in MJ-treated plants (Figure 8; P = 0.001 [sucrose] and 0.008 [trehalose]). Trehalose-6-phosphate is a signaling sugar which coordinates sucrose supply to sink organ growth, while free trehalose is involved in abiotic stress responses (Paul et al., 2008; Fichtner and Lunn, 2021). Trehalose is known to suppress expression of jasmonate biosynthetic genes (Bae et al., 2005), but the reverse interaction is not well understood. The MJ-induced decrease in both sucrose and trehalose may reflect a general diversion of sugar reserves toward defensive metabolism. The free hexose pools (glucose, fructose, mannose, and galactose) showed only minor changes in response to hormone treatment including a small decline in fructose in MJ-treated plants and minor declines in galactose, glucose, and mannose upon SA treatment (Figure 8).

Discussion

The soft ionization GCMS methodology presented here is a powerful approach to quantifying stable isotope incorporation into the metabolite pools of plant central metabolism. It has been employed in the isotopolog analysis of sugar phosphates in yeast (Chu et al., 2015), cultured human cells (Okahashi et al., 2019), and perfused rat livers (Yang et al., 2006), but it has thus far not been explored for photosynthetic organisms. This technique expands our ability to perform untargeted, global metabolite profiling of plants under physiological conditions. Although NH₃-PCI sensitivity drops 20–25-fold compared with EI, the absolute abundance of the molecular ion of each feature and its signal-to-noise characteristics typically improves several hundred fold. The extremely reduced fragmentation of pseudomolecular ions led to a major reduction in the total number of ion signals across spectra and this had the additional benefit of reducing interference between co-eluting substances. This approach, therefore, offers clear advantages over conventional EI-GCMS when intact mass information or isotopolog ratios are paramount. While we have here demonstrated the utility of this CI technique on a conventional single stage quadrupole instrument, isotopolog analysis using CI can be broadly applied to more versatile MS detectors types such as triple quadrupole and high-resolution mass spectrometers and even other reagent gases such as isobutane (Capellades et al., 2021). Indeed, Capellades and co-workers recently described the merits of isobutane as a reagent gas, which, unlike NH₃, does not suffer from a loss of sensitivity compared with EI.

In addition to molecular mass information, NH₃-PCI also yields information on proton affinity of analytes through the ratio of their [M + H]⁺/[M + NH₄]⁺ ions (Figure 2), which unexpectedly displayed a sigmoidal rather than a linear relationship across a broad range of analytes. We observed small but significant differences in isotope precision in SIM versus scan modes for all the principal ions in the pseudomolecular ion cluster except for m/z 367.9 (M + 1) (Supplemental Table S2) which might be attributable to incomplete baseline separation of peaks during quadrupole data acquisition. The highest experimental error was noted for mass peaks within 1 Da of highly abundant ions, consistent with increased interference between incompletely resolved peaks. The standard technique for measuring isotope ratios is an elemental analyzer–isotope ratio mass spectrometer (EA-IRMS), a magnetic sector instrument whose accuracy is at least 10-fold better than the quadrupole-based method we describe here (Muccio and Jackson, 2009). However, the methodology presented here benefits from relative ease of use compared with EA-IRMS and takes advantage of established GC-based metabolomics protocols using a standard single quadrupole GCMS system fitted with a CI source. In particular, we report advances in the ability to monitor labeling of photorespiratory and shikimate pathway intermediates.

Photorespiration recycles 2-phosphoglycolate produced by the Rubisco oxygenation reaction into 3-phosphoglycerate, which can reenter the Calvin–Benson cycle (CBC). It accounts for ∼25% of the activity of Rubisco in C₃ plants (Sharkey, 1988) and therefore constitutes a major route of carbon flux. Methodologies for the quantification of photorespiratory intermediates have been reported (Arrivault and Obata, 2017), but the inclusion of isotopolog analysis to quantify rates remains a challenging prospect. The pool sizes, label incorporation rates, and half-lives of glycolate, glyoxylate, glycine, serine, and glycerate could be readily determined with this method (Figure 4). However, hydroxypyruvate recovery, which requires low temperature extraction conditions (Pick et al., 2013), was not compatible with this broad metabolome methodology. Half-lives of most photorespiratory intermediates were on the order of 6–14 min, which is several minutes longer than previous estimates (Szecowka et al., 2013). The labeling plateau and concentration of glyoxylate were conspicuously low compared with other photorespiratory intermediates. Although glyoxylate is thought to be mainly produced by glycolate oxidase (GO) and transaminated to glycine in the peroxisome (Somerville and Ogren, 1980; Igarashi et al., 2003), in Arabidopsis it may also be formed by glycolate dehydrogenase (GDH) in mitochondria (Bari et al., 2004; Niessen et al., 2007) or chloroplasts (Goyal and Tolbert, 1996; Goyal, 2002; Blume et al., 2013). Rice (Oryza sativa) plants with suppressed GO activity unexpectedly had higher rather than lower glyoxylate concentrations (Lu et al., 2014), underscoring its complex metabolism. Conversion of glyoxylate to malate through the latter half of the glyoxylate cycle could conceivably reduce glyoxylate concentration, but isocitrate lyase and malate synthase are not active in adult or senescent leave of Arabidopsis (Charlton et al., 2005). The lower than expected glyoxylate pool size indicates the labeling plateau cannot be explained by isotopic dilution from unlabeled pools in isolated compartments. Glyoxylate concentrations may reflect enzymatic activities that minimize the activity of this aldehyde outside the photorespiratory sequence due its reactivity toward lysine resides in proteins (Schmitt et al., 2005) and inhibition of Rubisco activation in the chloroplast (Chastain and Ogren, 1989). Indeed, cytosolic and plastidial glyoxylate reductases have been proposed to play a role in reducing glyoxylate toxicity (Allan et al., 2009). Its concentration may be further constrained by pyruvate dehydrogenase mediated decarboxylation in the plastid and mitochondria (Blume et al., 2013). Although glyoxylate does not appear to undergo turnover at the same rate as other intermediates in the photorespiratory sequence based on its labeling kinetics and pool size, this may simply reflect activities aimed at reducing its toxicity (Dellero et al., 2016) or decarboxylation to formate (Igamberdiev and Kleczkowski, 2018). In response to MJ treatment, the concentration of glyoxylate and glycolate rose significantly while serine and glycine become nearly undetectable (Figure 8), suggesting a block at the serine:glyoxylate and/or glutamate:glyoxylate aminotransferase step (SGAT and GGAT). GO is a major source of H₂O₂, which is involved in activation of SA defense signaling (Sørhagen et al., 2013). The jasmonate-induced suppression of photorespiration we observed, which also implies a reduction in H₂O₂ production, would be consistent with the generally antagonistic relationship between jasmonate and salicylate signaling (Kangasjärvi et al., 2012). The separate observation that SA application has the opposite effect of MJ on photorespiratory intermediates, increasing glycine and serine concentrations above control levels, also supports the notion of their opposing effects on photorespiratory flux.

Shikimate is the central intermediate in the eponymous pathway that produces the precursor for AAAs, phenylpropanoids, and various alkaloid and phenolic secondary metabolites. It has been estimated that one-third of fixed carbon passes through the shikimate pathway to phenylalanine (Razal et al., 1996). It is one of the most biotechnologically exploited pathways for generating industrial and pharmaceutical products in microbes (Averesch and Krömer, 2018). Our current understanding of how flux in this pathway is regulated in plants is limited compared with microbes. Unlike the analogous route in bacteria, the first committed step in Arabidopsis, catalyzed by 3-DEOXY-D-ARABINO-HEPTULOSONATE 7-PHOSPHATE SYNTHASE (DHS), is not under negative feedback regulation by Phe and Tyr, except in the seedling where an AAA feedback sensitive isoform (DHS2) is expressed (Yokoyama et al., 2021). The main photosynthetic isoform expressed in adults, DHS1, is insensitive to AAAs but inhibited by chorismate and caffeate. Flux through the shikimate pathway is partially controlled by substrate supply, which depends on transketolase activity for d-erythrose-4-phosphate (E4P) and import of phosphoenolpyruvate (PEP) by the PEP-phosphate transporter (PPT) (Maeda and Dudareva, 2012). Shikimate isotopolog detection by LCMS has been described (Maeda et al., 2010; Lynch et al., 2017), but exact measurements are complicated by co-eluting isobaric background compounds in the liquid phase. In the gas phase, the ammonia adduct of the 4TMS-derivatized shikimate (m/z 480.2) could be detected free of background noise, and we observed the expected ratio of M + 0 through M + 7 isotopes in control plants (Figure 6). The pool size was comparatively small (∼0.5 nmol·mg⁻¹ D.W.) compared with other organic acids of central metabolism such as pyruvate. The low level of shikimate labeling after 2 h can be explained by two factors. First, the E4P and PEP entering the shikimate pathway peak at ∼60% labeling in ¹³CO₂ feeding experiments (Sharkey et al., 2020). Second, phenylalanine labeling suggests the presence of a second, inactive, extraplastidic pool of shikimate (Abadie et al., 2021) which dilutes label in the actively metabolized pool in chloroplasts. Labeling of pyruvate and shikimate do not match despite an ostensibly common origin through PEP. This, together with its incongruency with citrate labeling, further supports the notion of a fast and slow labeling pyruvate pool in photosynthetically active chloroplasts. Besides a low level of production through cytosolic glycolysis, pyruvate might also arise by transamination with alanine in the peroxisomal steps of photorespiration (Liepman et al., 2019; Parthasarathy et al., 2019) or β-elimination from the Rubisco carboxylation reaction (Andrews and Kane, 1991).

Upon exogenous treatment with MJ, the shikimate pool size declined, but its labeling kinetics did not change. This may reflect a general decrease in Calvin cycle-derived substrate availability following MJ treatment, which suppresses photosynthesis rate (Attaran et al., 2014). Quantification of shikimate isotopologs by NH₃-PCI-GCMS did not suggest the flux of this pathway is under direct control of defense hormone signaling. However, a complete metabolic analysis of all intermediates in this pathway is required to understand a possible role of phytohormone signaling in regulating flux to chorismate. Although shikimate was the only intermediate from this pathway which was detected in untargeted screens of polar leaf extracts, the quantification of the remaining intermediates of this pathway using a targeted approach should also be feasible with NH₃-PCI-GCMS.

Plant mono and disaccharides serve as both energy source and signaling compounds (Gibson, 2005) and the ratio of invertase/sucrose synthase-mediated cleavage of sucrose is an important indicator of developmental progression (Koch, 2004). Estimating flux into sucrose has previously relied on modeling using labeling of its phosphorylated upstream precursors sucrose-6-phosphate, fructose-6-phosphate, and UDP-glucose (Szecowka et al., 2013; Heise et al., 2014; Ma et al., 2014), but direct detection of the complete complement of sucrose isotopologs by GCMS or LCMS, while possible, remains technically challenging due to fragmentation (Dethloff et al., 2017). Analysis of sucrose labeling in whole plant systems has instead typically relied on nuclear magnetic resonance spectroscopy (Ettenhuber et al., 2005; Römisch-Margl et al., 2007). We were able to detect intact sugars of central metabolism and their isotopologs using the NH₃-PCI-GCMS technique described here. Among the free sugars detected by this method, only sucrose demonstrated appreciable label incorporation rates. Although we readily detected label incorporation into fructose, glucose, and trehalose above background levels with this method, their label incorporation levels in leaves were low, consistent with a low level of invertase activity in this source tissue (Koch, 2004). The direct quantification of sucrose label incorporation rates described here can complement existing models of carbohydrate metabolism.

Trehalose deficiency has been linked to myriad phenotypes and metabolic irregularities (Paul et al., 2008) and its synthesis is thought to be important for drought and oxidative stress tolerance in tomato (Solanum lycopersicum) (Yu et al., 2019). The role of trehalose in plant stress responses is poorly understood, and efforts to expand our knowledge of trehalose signaling is hampered by the difficulty in detecting this low abundance disaccharide in plant tissue. Quantification of trehalose by NH₃-PCI-GCMS indicated that its concentration and label incorporation were extremely low and its pool size further decreased in response to MJ treatment (Figure 8). Little is known regarding the effect of jasmonate-signaling on trehalose function. However, exogenous trehalose has been shown to suppress lipoxygenase expression in Arabidopsis (Bae et al., 2005; Iordachescu and Imai, 2008), a key enzyme in jasmonate biosynthesis, suggesting that in isolation, these signals play opposing roles, consistent with our observation that MJ induces a decline in trehalose concentration. However, in soybean (Glycine max), application of trehalose following inoculation with the endophytic bacterium Sphingomonas sp. LK11 produced an increase in jasmonic acid (JA) (Asaf et al., 2017), underscoring the importance of the metabolic context in phytohormone signaling. The ability to quantify trehalose and its isotopologs as described here contribute to our ability to finely dissect the role of the sugar in plant responses to mixed biotic/abiotic stresses.

The TCA cycle is an iconic energy pathway that exemplifies metabolite channeling and metabolon function in plant cells (Zhang and Fernie, 2018). With no optimization, we were able to detect five of the eight intermediates, including their complete isotopolog complements. In the past 15 years, metabolic modeling and isotopic labeling studies have revealed divergent flux modes under which the TCA cycle can operate, depending on the tissue type and metabolic state of the plant (Tcherkez et al., 2009). These include noncyclic flux modes in autotrophic tissue where, in the light, malate as well as pyruvate are the main forms of substrate entering the TCA cycle (Sweetlove et al., 2010), while citrate exported from the mitochondria to the cytosol for NADPH production is the main product (Igamberdiev, 2020). Labeling data obtained by NH₃-PCI-GCMS support a stable, hemicyclic flux mode in illuminated Arabidopsis rosettes that includes high flux into citrate and malate but almost no turnover of TCA intermediates from α-ketoglutarate to fumarate (Steuer et al., 2007). Our observations in illuminated, autotrophic tissue are consistent with the TCA providing citrate as its principal output, rather than reduced NADH for ATP production, which is supplied by the chloroplasts under these conditions. Interestingly, treatment with defense hormones provoked a major reprogramming of this noncyclic flux mode (Figure 8). In analogy with the effects of these hormones on photorespiration, MJ and SA produced antagonistic effects on the TCA. Succinate, malate, and fumarate all rose in concentration in response to MJ treatment but fell following SA treatment. Citrate and α-ketoglutarate did the opposite, declining in concentration in response to MJ but increasing following SA application. These results suggest that jasmonate signaling induces a reprogramming of the TCA to a cyclic flux mode that more closely resembles that of heterotrophic tissues, whereas salicylate further stimulates noncyclic flow. A recent review emphasized the dual effects of various biotic and abiotic stressors that alternately promote or suppress noncyclic TCA flux in plants (Shelp et al., 2017). The observed shift toward cyclic TCA flux triggered by jasmonate signaling may be a consequence of the downregulation of photosynthesis that accompanies this phytohormone (Attaran et al., 2014). Under these conditions, photosynthetic ATP production is hindered and the cell shifts from citrate export to ATP production through the TCA, in effect downregulating the citrate valve (Igamberdiev, 2020) to meet its energetic needs as resources are redirected to defense.

Accurate label calculations in metabolites with 17 or more carbons are potentially compromised by overlapping signals of the [M + H]⁺ and [M + NH₄]⁺ pseudomolecular ion peaks. This would affect, at most, late eluting C₁₈ (or longer) fatty acids as well as phytosterols, precluding the simplified labeling calculations applied to other metabolites presented here. In practice, very few compounds face this restriction for the following reasons. First, product leaving the CBC is maximally 65% labeled in short-term labeling assays (Evans et al., 2022) due to dilution of CBC intermediates through the action of the xylulose-5-phosphate transporter (Sharkey et al., 2020), reducing the abundance of fully labeled species of all metabolites, especially M + 17 or higher isotopologs which may cause overlap of pseudomolecular ion clusters. Second, fatty acids and sterol biosynthesis occur more slowly than carbohydrate and amino acids (Szecowka et al., 2013), suggesting that only longer labeling assays (>2 h) may be subject to these effects. Finally, interference between pseudomolecular ion clusters is only problematic when they are of comparable intensities, and most analytes favor one or the other (Figure 2). For instance, the [M + H]⁺ monoisotopic ion of linoleic acid (TMS) is only 2.8% the intensity of its [M + NH₄]⁺ counterpart, and in the case of stearic acid (TMS), the same value is ∼5% (see Supplemental Data set 2). The chance of its M + H + 17 isotopolog interfering with the M + NH₄ + 0 monoisotopic ion under these circumstances is minimal. On the other hand, in the case of β-sitosterol (TMS), the [M + H⁺] ion is 34% of its ammonium adduct counterpart and therefore problematic. Reviewing the ∼200 NH₃-PCI spectra we include here, β-sitosterol appears to be the only metabolite subject to this issue in that it meets all the necessary criteria to fall under this restriction (17 or more carbons, comparable [M + H]⁺ and [M + NH₄]⁺ peak intensities, and measurable rate of label incorporation). Therefore, while potentially important, overlap of pseudomolecular ion clusters appears rare.

In summary, the use of ammonia as a GCMS chemical ionization reagent gas presents a major technical advance to quantifying fluxes of multiple central metabolic cycles simultaneously in plants. Here, we have provided a broad overview of targeted and untargeted approaches to surveying carbon flux in ¹³CO₂ labeled Arabidopsis rosettes which has generated several unexpected observations regarding photorespiration, the shikimate pathway, the TCA pathway, and the MEP pathway. Efforts to test these hypotheses in further detail are currently underway.

Materials and methods

Plant materials and cultivation conditions

Wild-type Arabidopsis (A.thaliana) ecotype Columbia-0 plants were used in this study. All plants were grown in environmentally controlled growth chambers with the following conditions: 9-h light photoperiod, 150 photosynthetically active radiation (µEinsteins·m⁻²·s⁻¹), 20°C, and 60%–70% relative humidity. Seeds were sown in 8 cm pots with a 1:3 perlite:BX soil mixture (Promix) and incubated at 4°C for 3 d before transfer to the growth chamber. Preflowering rosette stage plants were used in labeling assays 6–8 weeks after germination. ¹³CO₂ isotopic labeling assays, flash freezing, grinding, and lyophilization of plant rosette tissues were performed as previously described (Evans et al., 2022; González-Cabanelas et al., 2015; Bergman et al., 2021b). Unless otherwise noted, all chemical reagents were obtained from Millipore Sigma. For phytohormone treatments, plants were sprayed with either 100 µM MJ in water, 100 µM SA in water, or water only (control) ∼24 h prior to adaptation in a dynamic flow cuvette and labeling in 400 µL·L^{−1 13}CO₂.

Extraction and derivatization of plant tissue

Methanolic extraction of ground, lyophilized rosette tissue and chemical derivatization of polar functional groups for GCMS analysis were performed as described previously (Roessner et al., 2000; González-Cabanelas et al., 2015) with the following modifications. To each 5 mg tissue aliquot, an internal standard (IS) cocktail was added which consisted of 6 μg tropic acid, 10 μg ribitol, 50 μg lyxose, 50 μg chlorpyrifos, and 5 μg octyl-β-d-glucoside. Samples were analyzed by GCMS within 24 h of derivatization or stored at −80°C for a maximum of 72 h.

GCMS analysis

Isotopolog analysis of derivatized plant extracts was carried out on an Agilent Technologies 7890B GC system coupled to a 5977C single quadrupole detector fitted with a CI source. The carrier gas was He (Linde Gas) running at a constant flow of 1.1 mL·min⁻¹. A VF-5ms narrowbore capillary column (Agilent Technologies, 30 m × 0.25 mm i.d., 0.25-μm film thickness) served as stationary phase, and the 57 min standard run consisted of the following gradient: 70°C isothermal for 1 min, then 5°C·min⁻¹ to 325°C with a final 5 min hold time. A 1 μL aliquot of each derivatized sample was injected in split mode with a 1:10 split ratio with the injection port held constant at 225°C. For analysis of photorespiratory intermediates, the following shortened 26 min oven gradient was used: 70°C isocratic for 1 min, then 15°C·min⁻¹ to 325°C with a final 8-min hold time.

PCI was conducted with ultrahigh purity methane and ammonia reagent gases (Linde Gas). The CI source was tuned with perfluoro-5,8-dimethyl-3,6,9-trioxidodecane using methane reagent gas to a target peak width of 0.5 m/z, and the same analyzer settings were used for analysis with ammonia. Ion source conditions for ammonia CI were as follows: 150°C source temperature, ionization energy 175 eV, current 200 µA, and 1.25 mL·min⁻¹ ammonia gas flow, while for methane CI these same values were 150°C, 220 eV, 300 µA, and 1.0 mL·min⁻¹. The gas regulator, tubing, and foreline dry scroll pump (Agilent Technologies, Santa Clara, CA, USA; G3399A) were all constructed from corrosion-resistant stainless-steel components.

Standards and control samples were also analyzed by EI (70 eV ionization energy, source temperature 230°C, and a scan range of m/z 50–550) under identical chromatographic conditions to facilitate peak annotation. Mass calibration of the quadrupole analyzer in NH₃-PCI mode was confirmed daily by monitoring the pseudomolecular ion of a chlorpyrifos standard (C₉H₁₁Cl₃NO₃PS; [M + H]⁺ = m/z 349.9). To determine the isotopic discriminating power of the system, we compared the observed distribution of chlorpyrifos natural isotopes (M + 0 through M + 6) to theoretical values and calculated the trueness of measurement (mean observed relative abundance of each isotopolog minus the theoretical relative abundance). To calculate analyte RI of peaks run in NH₃-PCI mode and facilitate peak alignment, each batch run with the CI source included a C₁₀–C₄₀ alkane mixture run in methane PCI mode. The same alkane series was analyzed in each batch run in EI mode.

PCI mass data were acquired with two different methods: first in scan mode and again using a SIM method. For the latter, we monitored the monoisotopic peaks (M + 0) and labeled isotopologs (M + 1, 2,… n) of select metabolites (summarized in Table 1). Full scan data were acquired initially from m/z 55–950 to observe any trimethylsilyl-NH₃ fragment ions (m/z 90) and then routinely from m/z 150–950. Scan speed was set to 1,562 Da/s with a cycle time of 438.6 ms (2.3 Hz). SIM data were acquired with a dwell time of 20 ms. Five-point calibration curves were constructed for absolute quantification and corrected with IS areas. It should be noted that use of ammonia reagent gas creates substantial laboratory safety issues. Strict leak detection and exhaust protocols must be established to work safely with this gas due to its corrosive properties, including installation of gas sensors, hood ventilation ducts, and corrosion-resistant parts for all segments of the chain of contact.

Data analysis

Mass data were analyzed using Agilent MassHunter Qualitative Analysis (version 10.0) and integrated with the Agile2 function. Peak annotation was accomplished by searching EI mass spectra against the NIST14 database and comparison of the same peak in NH₃-PCI, where the analyte nominal mass could be unambiguously obtained from examination of the [M + H]⁺ and [M + NH₄]⁺ ion clusters. RI values were compared with those of MeOX/TMS derivatized plant compounds on the Golm Metabolome Database (http://gmd.mpimp-golm.mpg.de/). Peak picking, alignment, and % atom (fractional) labeling were accomplished with custom Python scripts that sorted complete, background subtracted spectra and extracted the pseudomolecular ion cluster of a user defined target list which further matched all peaks to RI established with an alkane series run in the same sequence. No tissue control (NTC) samples were subjected to the same sample preparation procedure to identify background peaks from the extraction matrix, and peaks identified in these controls were subsequently removed from the peak table. Metabolite labeling was calculated with the Equation (1/N) $\sum _{i=1}^N{M}_i\mathrm{ }x\mathrm{ }i$, where N is the number of carbon atoms in the molecule and M_i is the fractional abundance of the ith isotopolog. Naturally occurring secondary isotope abundance was subtracted from labeled samples using the average of six unlabeled wild-type controls. Limits of detection and quantification were calculated for each analyte as 3.3× SD/σ and 10× SD/σ, respectively, where SD is the standard deviation of the extracted ion signal before and after the analyte peak and σ is the slope obtained from the linear regression of peak intensity versus the mass (µg) of standard added. To accommodate quantification of highly labeled analytes, peak area was determined using a combined extracted ion chromatogram for M + 0 through M + n, where n is the number of carbons in the underivatized metabolite. Each IS was used to normalize analyte peak area individually and a separate calibration curve was constructed for each. Absolute concentration of each feature in plant samples was calculated as the average value obtained from four IS compounds normalized to the mass of tissue used in the extraction.

Accession numbers

Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers At4G39980 (DHS1), At4G33510 (DHS2), At3G14415, At3G14420, and At4gG18360 (GO), At2G13360 (SGAT), At1G23310 (GGAT), and At5G0658 (GDH).

Supplemental data

The following materials are available in the online version of this article.

Supplemental Figure S1. Representative low mass scans (m/z 55–650) of methyloxime/trimethylsilyl derivatized Arabidopsis leaf metabolites.

Supplemental Figure S2. Substrate commitment into intermediates of the photorespiratory pathway in illuminated Arabidopsis rosettes.

Supplemental Figure S3. Effects of exogenous MJ treatment on ¹³C label incorporation rates into sugars, amino acids, and organic acids of central metabolism.

Supplemental Table S1. Complete list of metabolic features detected in leaf extracts by NH3-PCI-GCMS.

Supplemental Table S2. Theoretical and experimental values for secondary isotopes in chlorpyrifos obtained with NH3-PCI-GCMS.

Supplemental Table S3. Comparison of metabolite concentrations between control, MJ-, and SA-treated plants.

Supplemental Data set 1. Ammonia positive chemical ionization mass spectra of A. thaliana derivatized leaf metabolites.

Supplemental Data set 2. Complete NH3-PCI peak list and tabulated spectra.

 

M.E.B. performed the research, contributed the computational tools, and analyzed the data. B.D., S.E.E., R.H., I.B., A.J., and A.K.C. performed the research. M.A.P. designed the research, contributed analytical tools, and wrote the manuscript.

The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (https://academic.oup.com/plphys/pages/general-instructions) is: Michael A. Phillips (michaelandrew.phillips@utoronto.ca)