Abstract

   Drought can severely damage crops, resulting in major yield losses.
   During drought, vascular land plants conserve water via stomatal
   closure. Each stomate is bordered by a pair of guard cells that shrink
   in response to drought and the associated hormone abscisic acid (ABA).
   The activation of complex intracellular signaling networks underlies
   these responses. Therefore, analysis of guard cell metabolites is
   fundamental for elucidation of guard cell signaling pathways. Brassica
   napus is an important oilseed crop for human consumption and biodiesel
   production. Here, non-targeted metabolomics utilizing gas
   chromatography mass spectrometry (GC-MS/MS) and liquid chromatography
   mass spectrometry (LC-MS/MS) were employed for the first time to
   identify metabolic signatures in response to ABA in B. napus guard cell
   protoplasts. Metabolome profiling identified 390 distinct metabolites
   in B. napus guard cells, falling into diverse classes. Of these, 77
   metabolites, comprising both primary and secondary metabolites were
   found to be significantly ABA responsive, including carbohydrates,
   fatty acids, glucosinolates, and flavonoids. Selected secondary
   metabolites, sinigrin, quercetin, campesterol, and sitosterol, were
   confirmed to regulate stomatal closure in Arabidopsis thaliana, B.
   napus or both species. Information derived from metabolite datasets can
   provide a blueprint for improvement of water use efficiency and drought
   tolerance in crops.

Introduction

   Plants are sessile organisms that are continuously subjected during
   their lifecycles to a spectrum of environmental signals and stimuli,
   including both abiotic factors such as availability of water, light,
   and nutrients, and biotic factors such as interactions with both
   beneficial and pathogenic organisms. Accordingly, plants have evolved a
   suite of molecular mechanisms for external signal perception and
   transduction, facilitating acclimation to diverse environmental
   conditions^[26]1. Water deficit is one of the major abiotic stresses
   causing severe losses in crop production^[27]2. Drought triggers the
   biosynthesis, accumulation, and redistribution of abscisic acid (ABA),
   which promotes stomatal closure, inhibits stomatal opening, and thereby
   reduces water loss^[28]3–[29]7. The pivotal role of stomata in ABA
   interactions during drought stress has resulted in extensive efforts to
   elucidate the ABA signaling pathways of guard cells, which border and
   regulate stomatal apertures. Genetic screens and recent systems biology
   studies have revealed many signaling events and molecular components
   that participate in ABA signaling^[30]8–[31]15.

   Knowledge of ABA signaling in guard cells has been largely derived from
   the model plant Arabidopsis thaliana, which has limited economic value.
   Brassica napus, also known as oilseed rape or rape, is one of the
   largest commercial sources of vegetable oil. B. napus is grown
   worldwide for both human consumption and biodiesel production. B. napus
   is susceptible to drought stress, which can cause severe reduction in
   oilseed production^[32]16. An improved understanding of molecular
   responses to ABA in B. napus guard cells will inform genetic
   engineering and breeding approaches to enhance drought tolerance in
   crops.

   Large scale B. napus guard cell protoplast isolation from B. napus
   leaves can be conducted with high purity and yield, which provides
   optimal material for –omics analyses on this single cell
   type^[33]17–[34]19. Using an iTRAQ (isobaric tag for relative and
   absolute quantitation)-based comparative proteomics approach, 66 and 38
   proteins were found to be significantly induced and suppressed by ABA
   in B. napus guard cells, respectively. These ABA responsive proteins
   participate in photosynthesis, metabolism, energy, protein synthesis,
   stress/defense (antioxidant system and glucosinolate-myrosinase
   system), membrane and transport processes, and protein
   folding/transport and degradation^[35]10. Recently, 65 thiol-based
   redox responsive proteins were identified from ABA-treated B. napus
   guard cells, which highlights redox switches as important regulatory
   mechanisms in ABA signal transduction in guard cells^[36]14.

   Metabolites are direct physiological signatures and are highly
   correlated with phenotypes^[37]20; thus, study of cellular metabolomics
   is also indispensable for complete understanding of stress responses.
   Stress responsive metabolomes have been investigated in cell culture
   and in whole plants or whole organs, but rarely in single cell
   types^[38]21,[39]22. One landmark application of metabolomics to study
   the stress regulated metabolome at the level of the single cell type
   was an investigation of the ABA responsive metabolic changes in guard
   cell protoplasts from A. thaliana wild type and heterotrimeric
   G-protein α subunit mutant, gpa1, using targeted metabolomics with
   multiple reaction monitoring (MRM)^[40]12. In targeted metabolomics,
   paired mass/charge (m/z) ratios of the precursor ion and a selected
   daughter ion along with the chromatographic retention time, as acquired
   from analysis of authentic compounds (standards), are employed to
   identify a metabolite in experimental samples. Eighty-five
   signaling-related metabolites in A. thaliana guard cells were detected
   and quantified. The abundance of nearly half of these metabolites (41
   out of 85) in wild type guard cells was significantly changed after ABA
   treatment. Interaction with other hormones, particularly
   indole-3-acetic acid (IAA), in ABA modulated stomatal movement was
   revealed, validating phytohormone crosstalk^[41]12. These targeted
   MRM-based profiles of the A. thaliana guard cell metabolome provided
   the first example of investigating dynamic metabolome changes of a
   single-cell-type in plants.

   Plant metabolomes are highly diverse and have been recognized for their
   nutritional and medicinal value for centuries^[42]23. There are an
   estimated ~200,000 metabolites produced by the plant kingdom^[43]24. To
   date, however, only ~100 metabolites have been identified in guard cell
   protoplasts or implicated in guard cell functions^[44]18. The majority
   of these metabolites were identified in the targeted metabolomics study
   of Jin et al.^[45]12, while others were identified in focused studies
   on a specific metabolite or metabolic pathway. Recently, using guard
   cell enriched epidermal peels prepared from B. napus leaves, a material
   relatively easier to obtain, the guard cell metabolite inventory has
   been expanded to a few hundred metabolites, based on discovery from
   both targeted and non-targeted metabolomics platforms^[46]25,[47]26.

   Non-targeted metabolomics provides a complementary approach to targeted
   metabolomics, with the aim to acquire not only the mass/charge ratio
   but also the tandem mass spectra of all detected precursor
   molecules^[48]20. Such information facilitates elucidation of the
   chemical structure of each molecule. Instead of selective detection of
   a pre-defined metabolite group, non-targeted metabolomics provides
   global information on the metabolome. To improve our knowledge of the
   functional guard cell metabolome, here we employed non-targeted
   metabolomics workflows utilizing two complementary platforms, i.e., gas
   chromatography (GC)-mass spectrometry (MS) and liquid chromatography
   (LC)-MS to profile the B. napus guard cell metabolome and its
   modulation by ABA, resulting in a profile of 390 non-redundant
   metabolites, 77 of which were ABA responsive. Based on these results,
   several secondary metabolites were chosen for targeted study and were
   found to show either antagonistic or additive effects on ABA-induced
   stomatal closure. Information derived from metabolite datasets will
   improve our knowledge of ABA signaling in guard cells.

Results

Physiological stomatal response to ABA in B. napus

   ABA regulated stomatal movement has been observed in a wide range of
   plant species, including A. thaliana and B. napus
   ^[49]10,[50]12,[51]27,[52]28. Here we first confirmed that 10 µM ABA, a
   concentration typically used in assays of stomatal
   responses^[53]29–[54]31, is sufficient to induce stomatal closure in
   both leaf pieces (Fig. [55]1A) and epidermal peels (Fig. [56]1A) of B.
   napus line DH12075. ABA-induced stomatal closure was observed within
   2 min and closure was complete within 30 min of treatment in both
   materials (Fig. [57]1A). These results indicate the effectiveness of
   the ABA concentration used for our subsequent metabolomics analyses on
   B. napus guard cell protoplasts (GCPs). We also confirmed that the
   solvent for ABA application, ethanol, had no effect on stomatal
   apertures (Fig. [58]1A).

Figure 1.

   [59]Figure 1
   [60]Open in a new tab

   Responses to ABA in B. napus leaves, epidermal peels, and guard cell
   protoplasts. (A) ABA (10 µM) induces stomatal closure in both leaf
   pieces (left panel) and epidermal peels (right panel) of B. napus line
   DH12075. Data are means ± standard errors of 3 independent replicates
   with 100 ± 5 stomata measured for each sample. (B) ABA-induced
   shrinkage of B. napus GCPs. Representative image (left); scale bar
   indicates 25 µm. Data (right) are means ± standard errors of 4
   independent replicates with 100 ± 5 GCPs measured for each sample. (C)
   B. napus GCPs are viable following ABA or ethanol (solvent control)
   treatment. Samples before treatment (0 min) and after treatment
   (ethanol (EtOH) 60 min and ABA 60 min) were FDA stained to assess cell
   viability. Scale bars indicate 10 µm. Asterisks in A and B indicate
   that ABA treatment differed significantly from the EtOH solvent control
   (Student’s t test; p < 0.05).

   Although epidermal peels have been used for guard cell related –omics
   studies^[61]11,[62]25,[63]26,[64]32, our metabolome analyses were
   performed on guard cell protoplasts, rather than on epidermal peels, in
   order to exclude metabolites arising from pavement cells and the
   cuticle. The viability of isolated GCPs (Fig. [65]1B) before and after
   ABA or ethanol treatment was confirmed by fluorescein diacetate (FDA)
   staining (Fig. [66]1C). ABA responsiveness of the GCPs was evaluated by
   measuring protoplast diameters over a time course of ABA treatment
   (Fig. [67]1B). GCP shrinkage was observed when GCPs were treated with
   10 µM ABA as compared to the ethanol (solvent) control (Fig. [68]1B and
   C). Together, these results confirm that viability and ABA
   responsiveness were maintained in B. napus GCPs after protoplasting and
   treatment.

Metabolome profiling of B. napus guard cells

   Two major objectives of this study were metabolome profiling and
   identification of ABA responsive metabolites in B. napus guard cells.
   We prepared a total of 226 million B. napus GCPs, obtained in ~50 GCP
   isolations from ~1500 g of B. napus leaves (fresh weight), for our
   metabolomics analyses. After protoplasting, B. napus GCPs were left
   untreated or exposed to a time course of ABA treatment (see next
   section). Pre-separation by gas chromatography (GC) or liquid
   chromatography (LC) coupled with tandem mass spectrometry are robust
   methods to generate fragmentation patterns that can yield definitive
   metabolite identification. Here we first employed both GC-MS/MS and
   LC-MS/MS to explore the B. napus guard cell metabolome. Five
   replicates, each with 4–4.5 million untreated GCPs, were prepared and
   analyzed on the two platforms. For our GC-MS/MS analysis, applying a
   requirement for metabolite presence in at least 4 out of the 5
   replicates of untreated sample (0 min) coupled with an identification
   score over 70 arising from a NIST 11 (National Institute of Standards
   and Technology, USA) mass spectral library search as the threshold, led
   to identification of a total of 53 metabolites. The majority of these
   metabolites (48 out of 53) are primary metabolites, as is expected for
   GC-MS analysis^[69]33–[70]35, and fall into the categories of
   carbohydrates, carboxylic acids, fatty acids and lipids (Fig. [71]2 and
   Supplemental Table [72]1). For our LC-MS/MS analysis, the same
   requirement for presence in at least 4 out of the 5 replicates of
   untreated sample (0 min), coupled with an identification score
   threshold of ≥0.6 in MassBank, led to an identification of a larger
   number of metabolites (Fig. [73]2 and Supplemental Table [74]1). Under
   positive ion mode, 224 metabolites were identified, with nearly 80%
   involved in secondary metabolism. In particular, a number of
   carotenoids (subgroup of terpenoids) and flavonoids (subgroup of
   phenolics) were detected in positive mode (Fig. [75]2A and Supplemental
   Table [76]1). Under negative ion mode, 168 metabolites were identified,
   of which nearly two thirds were secondary metabolites, with the
   dominant group being phenolics (58 out of 168), followed by sugar
   nucleosides/ nucleotides, carbohydrates and carboxylic acids
   (Fig. [77]2A and Supplemental Table [78]1).

Figure 2.

   [79]Figure 2
   [80]Open in a new tab

   Metabolomic profiling using complementary platforms resulted in
   identification of 390 non-redundant metabolites in B. napus GCPs. (A)
   Classification of metabolites identified from each platform, mainly
   based on structural characteristics^[81]101. The metabolite categories
   are listed in the figure caption in the clock-wise order in which they
   appear in the figure, starting with “Proteinogenic amino acids” at the
   12 o’clock position (arrow) in all four pie charts. No metabolites in
   the category of “Amines and polyamines” were identified by the LC-MS
   (−) platform. No metabolites were identified by GC-MS in the categories
   of “Cofactors”, “Non-proteinogenic amino acids”, “Alkaloids”, or
   “Sulfur-containing”. (B) Venn diagram showing the number of metabolites
   identified from each platform.

   In total, 390 non-redundant metabolites were identified by our criteria
   in untreated B. napus guard cells. Each of the three datasets from
   GC-MS/MS and LC-MS/MS positive and negative ion modes contain a unique
   subset of metabolites and thus these methods are complementary
   (Fig. [82]2B and Supplemental Table [83]1). Only three metabolites,
   phenylalanine, ferulic acid, and sinapic acid, were identified by all
   three acquisition methods (Fig. [84]2B). Only 13 metabolites were found
   in common between GC-MS/MS and LC-MS/MS, while 42 were identified by
   both positive ion mode and negative ion mode in LC-MS/MS (Fig. [85]2B).

   The KEGG (Kyoto Encyclopedia of Genes and Genomes) PATHWAY database
   ([86]http://www.genome.jp/kegg/pathway.html) contains a collection of
   manually compiled KEGG pathway maps representing molecular interaction
   and reaction networks of metabolism and other functions, derived from
   multiple organisms, including A. thaliana, B. napus, and B. rapa, a
   parental ancestor of B. napus. Of the metabolites we identified, there
   were 286 metabolites with a KEGG compound index. Of these, 124
   metabolites mapped to the metabolic pathways available in KEGG from the
   reference species B. napus. The same set of metabolites also mapped to
   the metabolic pathways of B. rapa in KEGG. Distribution of identified
   metabolites on the metabolic map indicates that a wide variety of
   metabolic pathways are reseprented in the B. napus guard cell
   metabolome (Supplemental Fig. [87]S1 and Supplemental Table [88]1).
   There are several reasons for incomplete mapping of all of our
   identified metabolites: 1) not all metabolites are indexed in KEGG. For
   example, a few of the flavonoids identified in our profiling,
   especially those in the glycosidic form, are not available in KEGG
   (Supplemental Table [89]1), which is consistent with previous
   observations on other plant metabolomes^[90]36. 2) KEGG pathway maps
   are based on experimental knowledge of metabolism, which can be far
   from complete.

   The Plant Metabolic Network (PMN, [91]http://www.plantcyc.org/) is
   another valuable compendium of plant compound information^[92]37. Of
   the metabolites we identified, 128 are included in the B. rapa compound
   list from PMN, which contains over 1700 non-redundant metabolites. For
   the 262 metabolites without a hit in the PMN B. rapa compound list, 92
   have one or more chemical derivatives in the PMN list (Supplemental
   Table [93]1). Together with those metabolites mapped to KEGG pathways,
   174 metabolites known to be present in B. rapa and/or B. napus are
   present in our metabolome profiling, suggesting considerable metabolite
   diversity in B. napus guard cells. Additionally, 29 metabolites out of
   the 390 metabolites we identified in B. napus guard cells were also
   detected in a previous, targeted study on A. thaliana guard cells that
   quantified a set of 85 specific metabolites^[94]12 (Supplemental
   Table [95]1). Misra et al.^[96]25 and Geng et al.^[97]26 recently
   reported 268 and 358 metabolites, respectively, in the metabolite
   profiling of guard cell enriched epidermal peels from B. napus, in
   studies of metabolome responses to bicarbonate and elevated CO[2],
   respectively. Our identified metabolome had an overlap of 62 and 74
   metabolites with those datasets, respectively (Supplemental
   Table [98]1). Our metabolome profile (390 metabolites; 225 of which
   were not previously identified in other guard cell metabolome profiling
   datasets^[99]12,[100]25,[101]26) thus significantly expands our
   knowledge regarding the metabolome of this specialized cell type.

Identification of ABA-responsive metabolic signatures in B. napus GCPs

   B. napus GCPs were treated with ABA at a final concentration of 10 µM
   for 0 (i.e., untreated GCPs; results discussed above), 2, 15, or
   60 min, respectively. Given that ethanol (EtOH, solvent control) did
   not regulate stomatal movement in B. napus (Fig. [102]1A) and no
   significant changes in metabolites caused by ethanol were detected in
   A. thaliana guard cells^[103]12, only GCPs treated with ethanol for
   15 min were prepared in our experiment. Five replicates of each sample,
   i.e., 0 min, ABA 2 min, ABA 15 min, ABA 60 min, and EtOH 15 min, were
   prepared and analyzed on GC-MS/MS and LC-MS/MS in parallel. Principal
   component analysis (PCA) revealed that the 0 min and EtOH 15 min groups
   cluster together, whereas ABA treated groups are distinguished from
   0 min and EtOH 15 min, indicating the ABA treatment as the major factor
   contributing to the cluster segregation (Supplemental Fig. [104]S2). To
   identify responsive metabolites, we imposed a threshold of p value
   ≤0.05 in Student’s t test and at least 20% in fold
   change^[105]38,[106]39. For these analyses, each treatment sample,
   i.e., ABA 2 min, ABA 15 min, ABA 60 min, and EtOH 15 min was compared
   to the 0 min sample. EtOH treatment for 15 min caused changes in only
   12 metabolites, with 10, 1, and 1 identified from GC-MS/MS, LC-MS/MS
   positive mode, and LC-MS/MS negative mode, respectively (Supplemental
   Table [107]2). Eleven of these metabolites, for example, xylitol and
   palatinose, were also found in our ABA responsive dataset; these
   metabolites were not designated as ABA-regulated, due to their EtOH
   responsiveness. After combining the GC dataset with the two LC
   datasets, 17, 66, and 18 metabolites were found to be ABA responsive at
   time points 2 min, 15 min, and 60 min respectively, as compared to
   0 min (Supplemental Table [108]2). The abundance changes of all these
   metabolites (77 unique metabolites in total) at different time points
   under ABA treatment are represented by heat maps (Fig. [109]3). None of
   these 77 metabolites was absent (i.e., no detection) in untreated
   (0 min) samples but present in ABA-treated samples, although some
   unidentified MS peaks appeared upon ABA treatment. Among the ABA
   responsive metabolites, 8, 27, and 48 metabolites were revealed by GC,
   LC positive mode, and LC negative mode, respectively. There was only
   one metabolite (galactinol) common to GC and LC negative mode, no
   metabolites common to GC and LC positive mode and only 5 metabolites
   common to LC positive mode and negative mode
   (5-aminoimidazole-4-carboxamide-1-ribofuranosyl 5′-monophosphate,
   2,3-diphosphoglycerate, 2′-deoxyadenosine-5′-monophosphate,
   S-lactoylglutathione, and kaempferol) (Supplemental Table [110]2),
   again illustrating the value of multiple analysis methods. The major
   groups of the ABA responsive metabolites were phenolics (mostly
   flavonoids), carbohydrates, terpenoids (mostly tetraterpenoids), sugar
   nucleosides/nucleotides, and sulfur-containing metabolites (Fig. [111]3
   and Supplemental Table [112]2).

Figure 3.

   [113]Figure 3
   [114]Open in a new tab

   Primary (A) and secondary (B) metabolites responsive to ABA at
   different time points in B. napus GCPs. At 2, 15, and 60 min heat maps
   represent log2 of fold change, i.e., the log2-transformed metabolite
   abundance (peak area) at each time point divided by the level at 0 min;
   a 0 min column is also provided for comparison. All metabolites
   depicted were significantly changed at one or more time points (2 min,
   15 min, and 60 min) of ABA treatment. Abbreviations: UDP: uridine
   diphosphate; AICAR: 5-aminoimidazole-4-carboxamide-1-ribofuranosyl.

   A pathway enrichment analysis was performed using all available KEGG
   IDs of the ABA responsive metabolites (58 out of 77) against the KEGG
   A. thaliana reference metabolome using MetaboAnalyst 3.0^[115]40.
   Figure [116]4 shows all identified pathways from pathway enrichment
   analysis, which assesses the over-representation of inquiry compounds
   in known pathways, and their pathway impact values from pathway
   topology analysis, which indicate the importance of the identified
   metabolites to that pathway^[117]41 (Supplemental Table [118]3).
   Enriched pathways with high impact include flavone and flavonol
   biosynthesis, amino sugar and nucleotide sugar metabolism, and starch
   and sucrose metabolism (Fig. [119]4). Flavone and flavonol are two
   subgroups of flavonoids that are widely distributed secondary
   metabolites in higher plants^[120]42. The majority (15 out of 17) of
   the flavonoids were upregulated by ABA at 15 min (Fig. [121]5A and B;
   Supplemental Table [122]2). Sugar metabolism is also highly impacted by
   ABA treatment (Fig. [123]4). For example, an increase in sucrose and
   glucose 1-phosphate was observed in B. napus GCPs under ABA treatment
   (Fig. [124]3). Uridine 5′-diphosphate (UDP), UDP-glucose, and
   UDP-rhamnose also showed significant upregulation by ABA treatment
   (Fig. [125]3).

Figure 4.

   Figure 4
   [126]Open in a new tab

   Metabolic pathways affected by ABA treatment in guard cells revealed by
   pathway analysis. x axis represents the impact of the identified
   metabolites on the indicated pathway. y axis indicates the extent to
   which the designated pathway is enriched in the identified metabolites.
   Values were ascertained from MetaboAnalyst. Circle colors (see color
   scale for reference) indicate pathway enrichment significance. Circle
   size indicates pathway impact.

Figure 5.

   [127]Figure 5
   [128]Open in a new tab

   Abundance changes along the time course of ABA treatment for quercetin
   and quercetin derivatives (A), non-quercetin flavonoids (B), and
   glucosinolates (C). Metabolites 1–22 are: 1:
   quercetin-3-(6″-malonyl)-glucoside; 2: quercetin; 3:
   quercetin-3-arabinoside; 4: quercetin-3,4′-O-di-beta-glucopyranoside;
   5: quercetin-4′-glucoside; 6: myricetin-3-galactoside; 7: kaempferol;
   8: kaempferol-3-O-glucoside; 9: cyanidin-3,5-di-O-glucoside; 10:
   hesperetin; 11: isosakuranetin-7-O-neohesperidoside; 12:
   cyanidin-3-sophoroside; 13: hesperidin; 14: naringin; 15:
   cyanidin-3-O-galactoside; 16: 3-hydroxy-3′,4′,5′-trimethoxyflavone; 17:
   myricetin; 18: 7-methylthioheptyl glucosinolate; 19: 8-methylthiooctyl
   glucosinolate; 20: (2 R)−2-hydroxy-2-phenethylglucosinolate; 21:
   4-methylsufinyl-3-butenyl glucosinolate. Solid data points indicate
   statistically significant changes upon ABA treatment (Student’s t test;
   p value < 0.05) compared to 0 min data.

Effects of flavonoids, glucosinolates, and sterols in stomatal responses to
ABA in A. thaliana and B. napus

   Multiple flavonoids were identified as ABA-responsive in B. napus guard
   cells (Supplemental Table [129]2, Fig. [130]3, and Fig. [131]5A and B),
   most of which were glycosylated, i.e., linked with a sugar moiety.
   Although such conjugated metabolites are generally presumed to be
   inactive^[132]43, one study found that flavonoid glycosides such as
   quercetin 3-O-glucoside and kaempferol 3-O-glucoside exhibit radical
   scavenging activities, a feature that would suppress ROS^[133]42, which
   are known to promote stomatal closure^[134]44,[135]45. We observed that
   quercetin and quercetin-3-(6″-malonyl)-glucoside were strongly induced
   by ABA (Fig. [136]5A). Two other quercetin derivatives,
   quercetin-3,4′-O-di-beta-glucopyranoside and quercetin-3-arabinoside
   were also induced while quercetin-4′-glucoside was slightly repressed
   (Fig. [137]5A and Supplemental Table [138]2). All of the other 12
   non-quercetin related flavonoids were significantly induced by ABA
   treatment at 15 min except for myricetin (Fig. [139]5B and Supplemental
   Table [140]2).

   Based on the results of our large-scale metabolite analyses and
   previous observations^[141]46,[142]47, we hypothesized that the
   identified flavonoids would modulate stomatal movements. As a test of
   this hypothesis, we applied a non-glycosidic form of one of the
   strongly ABA-upregulated flavonoids, quercetin, to investigate its
   effect on stomatal movement and its regulation by ABA in A. thaliana.
   Based on quercetin measurement in the leaves of A. thaliana
   ^[143]47,[144]48, we estimated that in vivo quercetin concentration is
   close to 1 µM. In Arabidopsis leaves, application of 1 µM quercetin
   caused a slight increase in stomatal aperture compared to solvent
   control after 90 min treatment; however, without statistical
   significance (Fig. [145]6A). On the other hand, ABA-induced stomatal
   closure was opposed by 1 µM quercetin (Fig. [146]6A), suggesting an
   antagonistic role of quercetin in the ABA signaling pathway. A
   significant effect of quercetin was observed at 90 min, implying the
   interaction of quercetin with ABA either as a late stage signaling
   event or a sustained process. We also tested the effect of quercetin in
   B. napus leaf pieces. However, an antagonistic effect of quercetin in
   ABA (10 µM) -induced stomatal closure was not observed in B. napus even
   at concentrations up to 5 µM (Fig. [147]6B). The inconsistency of
   quercetin effect between the two species might be caused by
   species-dependent sensitivity to the metabolite tested.

Figure 6.

   [148]Figure 6
   [149]Open in a new tab

   Effects of quercetin, sinigrin, campesterol, and β-sitosterol on
   stomatal apertures in A. thaliana (A,C,E and G) and B. napus (B,D,F and
   H) leaves. Data are means ± standard errors of at least 4 independent
   replicates with 100 ± 5 stomata measured for each sample. Asterisks
   indicate a significant effect of addition of the secondary metabolite
   (Student’s t test; p < 0.05).

   Several glucosinolates were also found to be responsive to ABA in our
   guard cell metabolomes, with 7-methylthioheptyl glucosinolate and
   8-methylthiooctyl glucosinolate significantly increasing after ABA
   treatment (Fig. [150]5C and Supplemental Table [151]2). The
   glucosinolate-myrosinase system is a defensive mechanism uniquely
   present in some plant families, including the Brassicaceae ^[152]49.
   Stomatal movement modulation by components in this system also has been
   recognized^[153]50–[154]52. To investigate glucosinolate regulation of
   stomatal movement, we applied an allyl-glucosinolate, sinigrin,
   separately or with ABA, to A. thaliana and B. napus leaves. Sinigrin is
   a naturally occurring metabolite in A. thaliana and B. napus, and is
   hydrolyzed by myrosinases into allyl isothiocyanate and allyl
   cyanide^[155]53,[156]54. Sinigrin-induced stomatal closure and an
   additive effect with 10 µM ABA in promotion of stomatal closure were
   observed with statistical significance in both A. thaliana
   (Fig. [157]6C) and B. napus (Fig. [158]6D).

   Several phytosterols (β-sitosterol, 5β-stigmastan-3b-ol, and
   campesterol) were downregulated in guard cells upon ABA treatment
   (Fig. [159]3 and Supplemental Table [160]2). We applied campesterol
   (2 µM), either separately or with ABA (10 µM), to A. thaliana and B.
   napus. In both species, an antagonistic effect of campesterol on
   ABA-induced stomatal closure was observed (Fig. [161]6E and F). The
   effect of another phytosterol, β-sitosterol was also tested in the
   ABA-induced stomatal closure of A. thaliana and B. napus. An
   antagonistic effect of β-sitosterol (5 µM) in ABA- induced stomatal
   closure was observed in B. napus (Fig. [162]6H) but not consistently in
   A. thaliana (Fig. [163]6G).

Discussion

Complementary GC-MS and LC-MS platforms together with simplified extraction
enhances coverage in metabolome profiling

   Largely due to the differences in the ionization techniques and columns
   for molecule separation, GC-MS and LC-MS each exhibit detection biases
   for certain classes of metabolites. Temperature gradients for
   separation and electron ionization are commonly used in GC-MS, and
   primary metabolites, such as amino acids, carbohydrates, organic acids,
   and fatty acids are the main categories of metabolites detected with
   GC-MS^[164]33. LC-MS usually separates molecules based on their
   polarity and uses soft electrospray ionization, which in practice
   covers a wider range of metabolites, including plant secondary
   metabolite groups such as alkaloids, phenolics, and flavonoids^[165]33.
   These platform biases were also observed in our study: the majority of
   metabolites identified with GC-MS were primary metabolites whereas
   LC-MS analyses detected both primary and secondary metabolites
   (Fig. [166]2 and Supplemental Table [167]1). The three analysis
   conditions are complementary and overall 390 non-redundant metabolites
   were identified, making this dataset one of the largest to date for a
   plant single cell type^[168]22,[169]55.

   Ionization in LC-MS can generally be divided into positive mode, in
   which samples are protonated, and negative mode, in which samples are
   deprotonated. Typically, certain categories of metabolites can be
   preferentially resolved under a certain mode. For example, in our
   experiments, carotenoids were only detected under positive mode
   (Supplemental Table [170]1), mostly with the MS precursor in the
   radical cation form [M]*^+ rather than the protonated form [M + H]^+, a
   phenomenon that has been observed before^[171]56. On the other hand,
   more acids were identified under negative mode, which might be due to
   the prone-to-deprotonate feature of acids (Supplemental Table [172]1).
   Therefore, as previously known and as we observed in our study, no
   single analytical instrument is entirely robust to cover the whole
   metabolome profile. For broad coverage, it is necessary to utilize
   multiple platforms for metabolome profiling.

   In addition to the choice of the instrumentation platforms, the
   selection of the metabolite extraction protocol is another key factor
   that influences metabolome coverage. Practically, only a fraction of
   the entire metabolome can be resolved, in part depending on the
   composition of the extraction solution. For example, inclusion of
   chloroform in the extraction solvent was found to be counterproductive
   in an untargeted LC-MS metabolomics workflow^[173]33. Additionally,
   metabolites, even if successfully extracted, might not be detected by
   mass spectrometry due to failure to be dissolved by the loading buffer
   prior to the analysis. Therefore, an unbiased and efficient extraction
   protocol is critically important to successful metabolome profiling.
   Although responsive to stimuli, GCPs lack cell walls, and so can be
   easily disrupted for metabolite release. Accordingly, in this study
   metabolites were extracted from GCPs simultaneously with
   derivatization, including heating and shaking steps for GC-MS/MS
   analysis, and vortexing with loading buffer for LC-MS/MS analysis. This
   procedure dramatically reduced sample processing time and potential
   sample loss caused by transfers during extraction.

Metabolite identification and annotation

   In the early days of mass spectrometry based metabolomics, m/z was used
   for identification, sometimes together with chromatographic retention
   time. A major disadvantage of this type of identification is that
   isomers and stereoisomers cannot be differentiated due to their
   identical mass and occasional co-elution on chromatography^[174]57.
   With the advent of tandem mass spectrometry, fragmentation patterns
   (MS/MS or MS^n) of a compound provide another important, and often
   defining, feature for metabolite identification^[175]58. Databases with
   mass spectral information are essential references for metabolomics