MALDI mass spectrometry-assisted molecular imaging of metabolites during nitrogen fixation in the Medicago truncatulaSinorhizobium meliloti symbiosis


For correspondence (e-mail


Symbiotic associations between leguminous plants and nitrogen-fixing rhizobia culminate in the formation of specialized organs called root nodules, in which the rhizobia fix atmospheric nitrogen and transfer it to the plant. Efficient biological nitrogen fixation depends on metabolites produced by and exchanged between both partners. The Medicago truncatulaSinorhizobium meliloti association is an excellent model for dissecting this nitrogen-fixing symbiosis because of the availability of genetic information for both symbiotic partners. Here, we employed a powerful imaging technique – matrix-assisted laser desorption/ionization (MALDI)/mass spectrometric imaging (MSI) – to study metabolite distribution in roots and root nodules of M. truncatula during nitrogen fixation. The combination of an efficient, novel MALDI matrix [1,8–bis(dimethyl-amino) naphthalene, DMAN] with a conventional matrix 2,5–dihydroxybenzoic acid (DHB) allowed detection of a large array of organic acids, amino acids, sugars, lipids, flavonoids and their conjugates with improved coverage. Ion density maps of representative metabolites are presented and correlated with the nitrogen fixation process. We demonstrate differences in metabolite distribution between roots and nodules, and also between fixing and non-fixing nodules produced by plant and bacterial mutants. Our study highlights the benefits of using MSI for detecting differences in metabolite distributions in plant biology.


Legumes develop a very efficient nitrogen-fixing symbiosis with soil bacteria collectively referred to as rhizobia. Every year, legume nodulation produces a quantity of available nitrogen that is equivalent to the amount synthesized by the fertilizer industry throughout the world (Zahran, 1999; Graham and Vance, 2003). This symbiotic association leads to development of specialized organs called root nodules in which the rhizobia reduce atmospheric dinitrogen into ammonium and transfer it to the host plant (Venkateshwaran and Ané, 2011). Biological nitrogen fixation is catalyzed by the nitrogenase complex, and is energetically expensive for the bacteria, which require an abundant carbon supply from the plant partner (Masson-Boivin et al., 2009). Moreover, the nitrogenase is irreversibly inhibited by oxygen, so root nodules create micro-aerobic conditions, partly through production of a hemeprotein called leghemoglobin (Gilles-Gonzalez et al., 1991; Ott et al., 2009). Maintaining efficient symbiosis requires a fine coordination between legume and rhizobial metabolic processes (Udvardi and Day, 1997).

Medicago truncatula (Medicago) has emerged as model legume for studies pertaining not only to plant–microbe symbioses, but also to general legume biology (Penmetsa and Cook, 2000; Benloch et al., 2003; Gallardo et al., 2003; Wang et al., 2008; Branca et al., 2011; Samac et al., 2011). Transcriptomic, proteomic and phosphoproteomic studies have been extensively pursued in this model legume (Benedito et al., 2008; Grimsrud et al., 2010; Rose et al., 2012). However, non-targeted large-scale metabolomic studies have often been limited to metabolite profiling of total extracts obtained from Medicago suspension cell cultures (Broeckling et al., 2005; Suzuki et al., 2005; Farag et al., 2007, 2008). A notable exception is a recent untargeted metabolomic study that used entire Medicago seedlings and identified oxylipins as important regulators of early symbiotic signaling (Zhang et al., 2012).

Plant metabolites have mainly been investigated using conventional techniques, such as liquid chromatography-mass spectrometry (LC–MS), gas chromatography-mass spectrometry (GC–MS) and capillary electrophoresis-mass spectrometry (CE–MS) (Desbrosses et al., 2005; Barsch et al., 2006b; Edwards et al., 2006; Harada and Fukusaki, 2009; Kueger et al., 2012). Various amino acids, organic acids, sugars and polyols were shown to be differentially represented in roots and nodules (Colebatch et al., 2004; Desbrosses et al., 2005; Barsch et al., 2006a,b; Brechenmacher et al., 2010). While impressive, these experiments used homogenized metabolite extracts, and thus were unable to resolve the spatial distribution of the metabolic components at the organ or tissue levels.

MALDI-MSI has emerged as a powerful tool to investigate the distribution of a wide range of molecules through direct analysis of plant specimens to visualize the distribution of multiple metabolites among various organs and tissues, with enormous interest in the field of plant biology (Stoeckli et al., 2001; Kaspar et al., 2011; Lee et al., 2012b). This innovative technique holds promise for a mechanistic understanding of the metabolic differentiation that defines legume nodulation. However, the matrix involved in the desorption/ionization process usually generates abundant peaks at low m/z (mass to charge ratio), thus interfering with detection of small molecules present in the samples. Matrix-free laser desorption/ionization (LDI) has therefore been investigated and shown to be applicable to image UV-absorbing metabolites from plant tissues (Hölscher et al., 2009). Moreover, various matrix-free surface-modified/functionalized techniques including metal and carbon materials, such as graphite powder, graphite solution, colloidal graphite or nanostructure-initiator functionalized chips have been developed to improve the performance of metabolite imaging applications (Northen et al., 2007; Cha et al., 2008; Woo et al., 2008; Jun et al., 2010). Alternative ionization techniques have been applied to the field of small-molecule MSI. Secondary ion mass spectrometry, which provides comparatively good resolution, has been coupled to MSI to investigate the distribution of metabolites, but is limited by its poor detection efficiency in the low mass range due to relatively low yields of secondary ion production (Pachuta and Cooks, 1987). Some ambient ion sources have been developed recently, including atmospheric pressure infrared MALDI (Li et al., 2007, 2008), laser ablation electrospray ionization (Nemes and Vertes, 2007), direct analysis in real time (Vaclavik et al., 2009) and direct electrospray ionization (Takáts et al., 2005; Wiseman et al., 2006). However, all of these ambient ion sources require a modified instrument set-up and provide limited spatial resolution, while MALDI-MSI instruments have been widely employed and extensively developed for high spatial resolution tissue imaging with excellent ionization efficiency for metabolites. Therefore, alternatives to conventional matrices have been sought that alleviate the matrix-related interferences. Based on the Bronsted–Lowry acid–base theory, which states that extremely basic or acidic organic molecules do not produce matrix-related ions (Shroff and Svatoš, 2009), a strong base 1,8–bis(dimethyl-amino) naphthalene (DMAN), also referred to as ‘proton sponge’, was used for metabolite detection. Spectra lacking of matrix-related peaks were obtained in low mass regions, enabling matrix-free metabolite detection with routine MALDI-MSI operations (Shroff et al., 2009).

Here, we investigated utilization of the novel matrix DMAN, together with the conventional matrix 2,5–dihydroxybenzoic acid (DHB), for MALDI-MSI applications to study Medicago root and nodule metabolome during nitrogen fixation. Metabolites of various chemical species, including amino acids, sugars, organic acids, lipids, flavonoids and their conjugates, were characterized and mapped on Medicago roots and nodules. In addition to wild-type (WT) plants and rhizobia, we utilized plant and bacterial mutants that are defective in nitrogen fixation to detect the metabolic differences relevant to nitrogen fixation, generating valuable information for understanding of the underlying mechanism of nitrogen-fixing process.


In situ profiling of metabolites in Medicago roots and nodules in positive mode

Medicago roots with nodules were sampled 3 weeks after inoculation with S. meliloti. The nodules were excised, embedded in gelatin and sectioned followed by matrix application prior to MSI analyses (Figure 1). To study the metabolic distribution in the roots and nodules, in situ profiling was performed in positive mode on the cross-sections of Medicago root and nodule tissues. ‘Snapshots’ of the metabolic profiles of the two distinct organs were compared with the MS profile obtained from the matrix alone (Figure 2a). The matrix-derived trace indicates that DHB gives rise to a number of abundant peaks, especially in the mass range of 150–400 Da, complicating metabolite detection within this range. However, the matrix background is relatively clean outside this range, making it possible to examine the endogenous metabolites present in Medicago. The usefulness of DHB is exemplified by enlargement of the overlaid spectrum between m/z 100 and 160. A number of ions were detected above the matrix background (Figure 2b). These masses were searched against the database and assigned to putative metabolites. The reliability of the peak assignments was indicated as the absolute deviation of observed and theoretical monoisotopic mass values (Δm). Moreover, the assignments that agree with previous metabolite studies in legumes were given higher fidelity. Multiple peaks in Figure 2(b) were assigned as amino acids with a threshold of Δm = 0.03 Da based on this method. For example, the m/z 118.09 and the m/z 132.10 ions were assigned as two non-polar amino acids, valine and leucine, respectively, both with Δm = 0. The polar amino acids proline, asparagine and glutamine were assigned to the peaks at m/z 116.07 (Δm = 0), 133.06 (Δm = 0) and 147.09 (Δm = 0.01 Da), respectively. In the legume–rhizobia symbiosis, free amino acids are not only the reduction products of atmospheric nitrogen catalyzed by nitrogenase in bacteroids, but are also involved in amino acid cycles to balance the mutualism. Therefore, it is not surprising that free amino acids, a major class of primary metabolites, appear prominently in profiling spectra of Medicago roots and nodules. In addition, several other free amino acids, including arginine and histidine, were also detected among the matrix clusters due to their relatively high abundance in Medicago. Other than standard amino acids, an osmoprotectant derived from proline, proline betaine, was assigned to the peak at m/z 144.10. In addition to amino acids, the peak at m/z 136.07 was identified as adenine, a nucleotide base with a variety of biological functions, such as respiration and protein synthesis (Lodwig et al., 2003; Borowiec et al., 2006). Its detection in nodulated alfalfa by MS reinforced our confidence in this assignment (Barsch et al., 2006b).

Figure 1.

MALDI-MSI of metabolites in Medicago roots and nodules. (a, b) Optical images of a longitudinal section of a nodule stained with methylene blue (a), and a thaw-mounted Medicago section on a MALDI plate (b). (c) Workflow for mapping metabolites in Medicago by MALDI-MSI.

Figure 2.

Profiling spectra of Medicago roots and nodules. (a) Overlaid profiling spectra from the nodule region (red), the root region (green) and the matrix DHB (blue). (b, c) Two enlarged spectra from (a). Metabolites detected on MALDI-TOF/TOF are annotated with their accurate masses. (d–f) MS/MS spectra of selected precursors acquired in positive mode for molecular identification.

Accurate mass and tandem mass spectrometry-assisted identification of metabolites

In order to confirm the mass-based identification obtained by MALDI-TOF/TOF, extracts of roots with nodules were directly infused into an electrospray ionization (ESI) ultra-high-resolution-quadrupole-time-of-flight (UHR-Q-TOF) mass spectrometer (mass error <5 ppm). In addition to the benefits of high resolution and accuracy provided by UHR-Q-TOF, the matrix-free ionization method, ESI, excludes the possibility of introducing interfering matrix peaks to the resulting spectrum, confirming the existence of the detected ions in the extract. The extracts were also separated and analyzed by coupling C18 reversed-phase liquid chromatography (RPLC) with an ESI-Q-TOF mass spectrometer to determine the complex isobaric secondary metabolites in the extracts. As shown in Table 1, the high-accuracy mass measurements acquired on the UHR-Q-TOF mass spectrometer confirmed the MALDI-TOF/TOF results, facilitating the molecular identification of metabolites. For example, the high mass resolution and accuracy of the UHR-Q-TOF mass spectrometer were manifested by differentiation of two close peaks at m/z 104.0706 and 104.1071, assigned as γ–aminobutyric acid (GABA) (Δm = 0 mDa) and choline (Δm = 0.1 mDa), respectively. In contrast, the modest resolution of MALDI-TOF/TOF produced a convoluted peak at m/z 104.10, resulting a Δm of 0.03 Da for GABA and a Δm of −0.01 Da for choline. The presence of GABA may have been overlooked if relying on MALDI-TOF/TOF. Despite the interference from matrix ions in the low-mass region, DHB is an excellent matrix for MALDI MSI, promoting ionization of analyte molecules in the higher mass range. Figure 2(c) shows two abundant peaks at m/z 616.18 and 664.12 in the enlarged nodule-specific spectrum. By a database search, the m/z 616.18 ion corresponds to heme and the m/z 664.10 ion is likely to be nicotinamide adenine dinucleotide (NAD). Unfortunately, the m/z value of putative metabolites. obtained on a moderate resolution MS may be close or equal to several potential candidate metabolites. Therefore, tandem mass (MS/MS) experiments were performed on selected precursor ions for unequivocal assignment directly from the regions that were rich in the metabolites of interest by MALDI-TOF/TOF. Figure 2(d–f) demonstrates utilization of in situ MS/MS for structural elucidation and therefore molecular identification. The fragments of the m/z 664.12 precursor ion sampled from the Medicago nodule region matched well with the theoretical fragments of NAD, indicating that the majority of the m/z 664.10 peak corresponds to NAD. Similarly, the MS/MS spectrum of the m/z 616.18 ion obtained from the nodule region gave rise to fragments that correspond to sequential losses of –CH2COOH from heme (Shimma and Seto, 2007) (Figure 2e). In Figure 2(f), MS/MS again assisted in identification of the root-specific m/z 517.13 peak, which was mass-matched to an isoflavonoid conjugate, formononetin glucoside malonate (MalGlc). Its signature fragment at m/z 269.08 agreed well with what was observed on an ESI-Q-TOF mass spectrometer (Figure S1a), increasing our confidence in MALDI-MS-based molecular assignments. Figure S1(b,c) show the tandem mass spectra of the isoflavonoid formononetin and the flavonoid conjugate aformonisin MalGlc at m/z 269.08 and 547.14, acquired via a conventional LC-MS/MS approach. The retention time and MS/MS spectra of putative metabolites observed by LC-ESI-Q-TOF further confirmed the identities assigned by MALDI-MS mass matching. In summary, combined use of MALDI-MS with a conventional metabolite characterization tool such as ESI-Q-TOF enables mapping of multiple metabolites from nodulated Medicago with unambiguous assignment (Table 1).

Table 1. Peak assignments in positive-mode profiling mass spectra of Medicago root and nodules
Name of metaboliteTheoretical [M+H]+UHR-Q-TOF measured [M+H]+UHR-Q-TOF measurement error; Δm (mDa)TOF/TOF measured [M+H]+
  1. a

    Tandem MS fragmentation was performed for identification.

  2. *[Correction added on 24.05.2013, after first online publication: a calculation error for mass measurement accuracy has been corrected which does not change any conclusions of this paper]

γ–aminobutyric acida104.0706104.07060104.10
Proline betainea144.1019144.10240.5144.10
Sucrose + K381.0794381.0791−0.3381.05
Chrysoseriol GlcAa477.1028477.1008−2.0477.10
Formononetin MalGlca517.1341517.1337−0.4517.13
Afrormosin MalGlca547.1446547.1427−1.9547.16

ClinProTools-processed profiling of root and nodule spectra

By comparing the root and nodule MS profiling, we clearly observed that a number of ions were differentially represented in the two regions, indicating that a characteristic metabolic pattern may exist for roots and nodules. In order to quickly extract the signature metabolites from the complex dataset, which consists of 12 root and nodule samples (three mass spectra acquired from each root and nodule regions derived from four Medicago sections), the raw spectra from the root and nodule regions were divided into two classes and loaded into ClinProTools (Bruker Daltonics, sequentially. Using ClinProTools, we were able to process and analyze raw spectra, including spectra recalibration and average spectra calculation, select peaks and calculate peak statistics. Processed spectra of the two loaded classes arranged in a ‘gel’ view with nodules displayed at the bottom of the view and roots above it are shown in Figure 3(a). Principal component analysis (PCA), a multivariate data-mining technique, was then applied to spectra of the two classes. Peak masses and peak heights from each technical replicate were averaged and translated to a point (x, y and z) in a 3D coordinate. Therefore, each point shown in the 3D PCA score plot represents a biological replicate averaged from three technical replicates, and each class consists of four points. The four biological replicates of nodules (red point) and the four biological replicates of roots (green point) segregate into two distinct groups as highlighted in the dashed circles (Figure 3b). To determine the significance of this segregation, an unsupervised clustering using a hierarchical clustering algorithm was then performed on PCA-transformed spectra, and a dendrogram was generated for all the data points (Figure S2). Two clades are shown, each corresponding to spectra with root or nodule origin, providing statistical support to the differentiation between roots and nodule metabolite spectra. Semi-quantitative results were obtained utilizing the peak statistics function embedded in ClinProTools. The peaks of m/z 137.02, 155.03, 177.02 and 214.97 that correspond to [DHB-H2O+H]+, [DHB+H]+, [DHB+Na]+ and [DHB-H+K+Na]+ show similar intensity between the nodule and root classes, demonstrating that the ionization efficiency of the analytes from the roots and nodules is comparable. Additionally, the signal intensity of metabolites probably reflects their presence in root and nodule regions and allows regional comparison. The metabolites showing distinct intensities in the two classes were thereby recorded, which potentially contributes to the metabolic differences between Medicago roots and nodules. As indicated in Figure 3(d), representative metabolites at m/z 147.09, 381.05, 616.18 and 664.12 were abundant in nodules with significant differences (< 0.01), whereas the intensities of the metabolites at m/z 517.12 and 849.47 were higher in roots than in nodules (< 0.01 and 0.05, respectively). The identities of these marker metabolites were assigned as glutamine, potassiated sucrose, heme, NAD, formononetin MalGlc and the glycerophospholipid GP(36:5), respectively (Table 1).

Figure 3.

Statistical analysis of root and nodule profiling spectra. (a) ClinProTools-processed profiling spectra loaded as two classes arranged in a ‘gel’ view with nodules displayed at the bottom and roots above. (b) PCA score plot of the nodules (red) and roots (green) classes. (c) Averaged signal intensities of the matrix-derived peaks between the nodule and root classes. (d) Averaged signal intensities of representative metabolites in Medicago. The peaks obtained for in nodule regions are shown in red and those obtained in root regions are shown in green. Asterisks indicates statistically significant differences compared with the root-specific spectra (**< 0.01, *< 0.05; Student's t test).

Imaging of metabolites in Medicago roots and nodules in the positive mode

Although MALDI-MS-based metabolite profiling reveals a differential distribution pattern of metabolites in Medicago roots and nodules, spatial resolution was sacrificed. To visualize the organ-specific patterns, MSI was performed on Medicago sections. Representative metabolites that exemplified distinct distribution patterns in roots and nodules are shown in Figure 4(a–h). The molecule at m/z 616.18, previously assigned as heme, was localized in the nodule region and particularly concentrated in the fixation zone (Figure 4a). Moreover, the peak of m/z 664.12 identified as NAD was also abundant in the nodule region (Figure 4b). This observation agrees with the higher abundance of these two peaks in nodules (Figure 3d), demonstrating the consistency between profiling and imaging results. On the other hand, the m/z 144.10 peak previously identified as proline betaine was more abundant in roots than in nodules (Figure 4c). In addition, the ion of m/z 849.47, putatively GP(36:5), displays a ubiquitous distribution pattern over the entire root/nodule section, but with relatively higher abundance in the root (Figure 4d).

Figure 4.

Representative metabolite distribution in a Medicago section as revealed by MALDI-MSI. (a–g) m/z 616.15 as a heme moiety (a), m/z 664.10 as NAD (b), m/z 144.10 as proline betaine (c), m/z 849.47 as a putative sodiated lipid (d), m/z 269.08 as formononetin (e), m/z 517.13 as formononetin MalGlc (f), and m/z 547.16 as afrormosin MalGlc (g), showing distinct distribution patterns in roots and nodules. (h) An overlaid image of (a) and (f). (i) PCA score plot of the two groups of MSI results (nodule and root) obtained by selecting the nodule as the region of interest 1 (ROI 1) and the root as the region of interest 2 (ROI 2). The pink points representing nodules plotted separately from the green points representing roots, demonstrating significant differences between the two ROIs. (j) Loading plot of PC1 and PC2. Each point within the plot represents a peak detected in the MSI dataset, whereas points that are far away from the center are responsible for the variance. The highlighted points corresponding to metabolites at m/z 547.16, 616.18 and 664.12 are furthest away from zero on PC2, thereby indicating their significant contribution to segregation of ROI 1 and ROI 2 on PC2.

An interesting class of secondary metabolites, flavonoids and their conjugates, were mapped with high resolution in nodulated Medicago. The m/z 269.08 ion assigned as formononetin exhibited ubiquitous distribution on the entire section, with high abundance in the root region (Figure 4e), whereas its glucoconjugate, formononetin MalGlc, at m/z 517.12 exhibited higher abundance in roots compared to nodules (Figure 4f). Flavonoids and their conjugates are a group of diverse molecules and have various functions in legume nodulation. Therefore, it is not surprising to find that the ion assigned as afrormosin MalGlc at m/z 547.16 showed a localization distinct from that of formononetin MalGlc. The metabolites that defined the nodule and root regions by localization were overlaid for visual contrast in Figure 4(h), with heme (m/z 616.18) shown in red and formononetin MalGlc (m/z 517.13) shown in green. PCA was applied to the MSI results, showing significant differences between nodules and roots (Figure 4i). Figure 4(j) shows the loading plots of PC1 and PC2. The highlighted points (Figure 4j) corresponding to the peaks at m/z 547.16, 616.18 and 664.12 contribute significantly to the segregation of nodule and root regions on PC2, agreeing well with their nodule-concentrated localization. By performing MSI with corroborative statistical analyses, detailed localization of metabolites was obtained without prior knowledge, enabling dissection of nodule chemistry and related mechanistic studies.

Although metabolic differences in the root and nodule regions were visualized by MSI in Medicago, identifying the metabolites that are differentially regulated in fixing nodules over non-fixing nodules may further facilitate our investigation of the key metabolites involved in nitrogen fixation. Here, we utilized mutants that are defective in nitrogen fixation, and performed MSI on the sections of the WT and mutant plants, respectively. Medicago defective in nitrogen fixation 1 (dnf1) mutants are affected in bacteroid and symbiosome development, hence dnf1 mutants form non-functional nodules that are unable to fix nitrogen (Mitra and Long, 2004; Wang et al., 2010). Similarly, fixJ mutants of S. meliloti may differentiate into elongated bacteroids, but quickly degenerate, leading to nodule senescence. These fixJ mutants cannot encode nitrogenase enzyme, and hence form non-fixing nodules. Thus, we utilized WT plants and WT rhizobia in combination with mutant plants and rhizobia (dnf1–1 and fixJ) to perform differential MSI experiments. Nitrogen-fixing nodules formed on WT plants interacting with WT rhizobia were pinkish and elongated (Figure 5a), but the non-fixing nodules formed on WT/fixJ, dnf1/WT and dnf1/fixJ were pale, small and more globular in shape (Figure 5b–d). No significant difference was observed in the rate of nodulation and amount of nodule tissue per root in three of the four combinations of plants and rhizobia tested (WT/WT, WT/fixJ and dnf1/WT) (Figure S3). As expected, the nodules formed in the WT/WT interaction showed high nitrogen fixation activity in the acetylene reduction assay, whereas nodules formed in the WT/fixJ, dnf1/WT and dnf1/fixJ interactions showed negligible to nil nitrogen fixation activity (Figure 5e). Distinct metabolite distributions were observed in the nodules formed in the WT/WT interaction compared with the nodules formed on other combinations (WT/fixJ, dnf1/WT and dnf1/fixJ; Figure 5f–i). The m/z 616.18 ion corresponding to the heme moiety, possibly of leghemoglobin, was present at high abundance in the fixation zone of nodules formed on WT/WT samples, but was completely absent from the nitrogen fixation-defective nodules. In contrast, the ion at m/z 517.13, formononetin MalGlc, was particularly concentrated in the roots of WT/WT Medicago, showing negligible differences with the other mutant samples (Figure 5j–m).

Figure 5.

Comparison of the metabolite distribution in nitrogen fixing and non-fixing Medicago nodules. (a) Nodules on a WT Medicago plant inoculated with the WT strain of rhizobia (WT/WT). (b) Nodules on a WT plant inoculated with the fixJ mutant of rhizobia (WT/fixJ). (c) Nodules on a Medicago dnf1–1 mutant inoculated with the WT strain of rhizobia (dnf1/WT). (d) Nodules on a Medicago dnf1–1 mutant inoculated with the fixJ mutant of rhizobia (dnf1/fixJ). (e) Nitrogen fixation activity. Each bar corresponds to the mean value of acetylene reduction for eight plants assayed 3 weeks after inoculation. Error bars indicate standard error (SE). (f–i)Distribution of the heme moiety at m/z 616.18 in the four different nodule types: WT/WT (f), WT/fixJ (g), dnf1/WT (h) and dnf1/fixJ (i). The heme moiety was detected only in fixing nodules in (f). (j–m) Distribution of the metabolite formononetin MalGlc at m/z 517.13 on WT/WT (j), WT/fixJ (k), dnf1/WT (l) and dnf1/fixJ (m), showing great similarity in contrast to (f)–(i). Scale bars = 1 mm.

Profiling and imaging of metabolites in Medicago roots and nodules using DMAN in negative mode

Although DHB has shown great capability in ionizing metabolites in the positive mode, the high-abundance matrix-derived peaks made endogenous metabolite detection more challenging. To overcome this problem, we utilized DMAN and developed its application for metabolite imaging in the negative mode. As the ionization efficiency of DMAN has been shown to be effective using metabolite standards (Figure S4), its utility for MALDI-MSI on biological tissues was investigated. Spectra acquired from the nodule region, the root region and the matrix alone using DMAN were compared, revealing numerous putative Medicago-specific metabolites with clean background (Figure 6). Based on the mass matching and MS/MS-assisted identification approach described previously, various classes of molecules were identified, including six amino acids, 14 organic acids and four carbohydrate-containing compounds. In contrast to the amino acids identified in positive mode, negatively charged side chain groups, such as aspartate at m/z 132.03 and glutamate at m/z 146.01, were identified. Several other amino acids with relatively low isoelectric points (pI) were also detected, including alanine, serine and threonine. In addition, GABA was detected in negative mode with no interference, circumventing the issues of potential mis-identification and overlapping with choline in the positive mode. All of the amino acid masses obtained on the UHR-Q-TOF mass spectrometer were within Δm = 1.5 mDa. In addition to amino acids, organic acids are critical for nitrogen fixation, serving as a source of carbon and energy delivered from the host plant cells to the bacteroids. Multiple organic acids were detected using DMAN rather than DHB, due to its high proton affinity. As detailed in Table 2, a significant number of organic acids were detected in negative mode, greatly expanding the metabolite coverage compared with DHB alone. Carbohydrates are another class of molecules that are pivotal to the mutual exchange of nutrients between legumes and rhizobia. In positive mode, sucrose is usually detected in its sodiated form at m/z 365.11 and its potassiated form at m/z 381.08. In contrast, sucrose was detected in its deprotonated form at m/z 341.07 (Δm = −0.04 Da) in negative mode, providing straightforward evidence for its presence in Medicago roots and nodules. Moreover, ions corresponding to the hexoses glucose and fructose were observed at m/z 179.05 (Δm = −0.01 Da) in deprotonated form. The ion at m/z 149.04 was identified as pentose, with an accurate mass of 149.0439 (Δm = −0.6 mDa) further provided by UHR-Q-TOF. Sugar phosphates are also used in biological systems to store or transfer energy. For example, hexose-6–phosphate was deprotonated and gave rise to a peak at m/z 259.04 in negative mode using DMAN. Its detection on the UHR-Q-TOF mass spectrometer further increased the confidence of assignment.

Table 2. Peak assignments in negative-mode profiling mass spectra of Medicago root and nodules
Name of metaboliteMonoisotopic [M-H]UHR-Q-TOF measured [M-H]UHR-Q-TOF measurement error; Δm (mDa)TOF/TOF measured [M-H]
  1. a

    Tandem MS fragmentation was performed for identification.

  2. Bold type indicates major metabolites in the tricarboxylic acid cycle.*[Correction added on 24.05.2013, after first online publication: a calculation error for mass measurement accuracy has been corrected which does not change any conclusions of this paper]†[Correction added on 24.05.2013, after first online publication: a calculation error for mass measurement accuracy has been corrected which does not change any conclusions of this paper]

Pyruvic acid 87.0077NANA87.00
Lactic acida89.023389.02441.189.01
Phosphoric acid96.968596.9671−1.496.96
2–ketobutyric acid101.0233101.0227−0.6101.02
γ–aminobutyric acida102.0550102.05611.1102.04
Maleic/fumaric acid a 115.0026115.00270.1115.03
Succinic acid a 117.0182117.01840.2117.01
Oxalacetic acid 130.9975130.9965−1.0131.03
Aspartic acid a 132.0291132.03051.4132.03
Malic acid a 133.0132133.0131−0.1133.02
Salicyclic acid137.0233137.0231−0.2137.02
α–ketoglutaric acid 145.0132145.0127−0.5145.02
Glutamic acida146.0448146.0435−1.3146.01
Aconitic acid a 173.0081173.00830.2173.03
Ascorbic acida175.0237175.0212−2.5175.05
Citric/isocitric acid a 191.0186191.01860191.02
Palmitic acid255.2319255.2318−0.1255.22
Stearic acid283.2632283.26390.7283.26
Figure 6.

Comparison of in situ profiled mass spectra from Medicago roots, nodules and DMAN matrix in negative mode. The inset shows an enlarged view of the nodule-specific spectrum. The identified peaks are annotated with masses and assigned identities. The asterisks indicate the major peaks arising from DMAN.

Similar to the positive-mode workflow, MS/MS fragmentation patterns assist in metabolite identification (Figure 7). The presence of diagnostic fragment peaks at m/z 27 and 71 shown in the top panel of Figure 7(a) suggests an identity of maleic acid when compared to the MS/MS fragments of the standard in the bottom panel. Similarly, the major fragment ions at m/z 59, 87, 129 and 173 from the citric acid standard and the peaks at m/z 59, 71, 89, 119 and 179 from the sucrose standard confirmed the identities of the metabolites that were in situ fragmented from the Medicago sections as shown in Figure 7(b,c). The MS/MS spectra confirm the metabolite identification based on accurate masses. Moreover, the utility of DMAN in generating MS/MS spectra of quality for metabolite identification is also demonstrated.

Figure 7.

In situ MS/MS spectra of metabolites. (a–c) In situ tandem MS spectra of maleic acid (a), citric acid (b) and sucrose(c) obtained from Medicago sections in negative mode. (d–f) MS/MS spectra of maleic acid (d), citric acid (e) and sucrose standards (f). The possible fragmentation pathways are shown.

DMAN has been applied here to MSI of metabolites that cannot be achieved otherwise. Figure 8 shows representative MS images of deprotonated metabolite peaks, including maleic acid (m/z 115.03), citric acid (m/z 191.02) and sucrose (m/z 341.07). Maleic acid exhibited a more universal distribution over the entire section, with higher concentration in the nitrogen-fixation zone, whereas citric acid and sucrose were localized more to the root area and the fixation zone, but displayed the highest concentration at the meristematic zone.

Figure 8.

Spatial distribution of metabolites in Medicago nodules. Optical image of a Medicago section (a) and representative negative-mode MS images of deprotonated metabolite peaks, including (b) maleic acid (m/z 115.03), (c) citric acid (m/z 191.02) and (d) sucrose (m/z 341.07).


To date, MALDI-MSI is one of the most promising imaging techniques, with ever-increasing sensitivity and specificity. Previous applications of MSI were dominated by the investigation of large molecules, until recently when MALDI-MSI of metabolites has received increased interest in the field of plant biology. Here, we have adapted MSI to describe the metabolome in Medicago roots and nodules. A number of metabolites were identified based on accurate mass measurements and MS/MS fragmentation, including amino acids, carbohydrates, organic acids and members of the flavonoid family. With statistical support by ClinProTools, we compared the differential abundances of metabolites in Medicago root and nodule. For example, glutamine and sucrose showed significantly higher intensities in the Medicago nodule. Interestingly, an independent study using GC–MS in Medicago sativa also reported significant increase in the abundance of glutamine and sucrose in nodules compared to roots (Barsch et al., 2006b). Sucrose is the primary source of carbon energy, which is synthesized in the leaves and exported to sinks such as nodules via the phloem, where it is metabolized by the glycolytic pathway into phosphoenol pyruvate or malate, the likely substrates of bacteroids to fuel biological nitrogen fixation (conversion of atmospheric dinitrogen into ammonia). The fixed ammonia is assimilated into glutamine, which is subsequently utilized in the synthesis of amides (in the case of Medicago) or ureides, and transported via xylem to the aerial parts (Gordon et al., 1995). As nodules are considered the main nitrogen source in nodulated legume plants, the elevated levels of glutamine together with other amino acids probably reflects the efficient nitrogen metabolism occurring inside the nodule. The power of MALDI-MSI in identifying the region-specific distribution of metabolites was further supported by detection of the heme moiety in mature fixing nodules, where it constitutes a key component of both leghemoglobin of plant origin and FixL of rhizobial origin (Gilles-Gonzalez et al., 1991; Lois et al., 1993; Garrocho-Villegas et al., 2007). Leghemoglobins are present in legume nodules at concentrations of 2–3 mm in the host cell cytoplasm, and help to maintain free oxygen concentrations of approximately 20–40 nm in the cytosol (Becana and Klucas, 1992). Such microaerobic conditions allow an adequate supply of ATP through respiration for nitrogen fixation, but prevent inactivation of the nitrogenase complex (Bergersen et al., 1982). By comparing MS images obtained from nodulated roots of Medicago (WT and dnf1) inoculated with rhizobia (WT and fixJ), we showed that the heme moiety is exclusively present in fixing nodules formed on WT plants inoculated with WT rhizobia. In contrast, nodules formed by plant or bacterial mutants were non-fixing and lacked the heme moiety. In line with our observations, Starker et al. (2006) observed a complete absence of nitrogen fixation in dnf1–1 mutant compared with WT plants. Medicago dnf1 mutants form infection threads similar to WT plants, but bacteroid differentiation is arrested at stage 1 (Starker et al., 2006; Wang et al., 2010). Similarly, fixJ rhizobium mutants cannot produce the nitrogenase enzyme, but are released into plant cells where they differentiate into elongated bacteroids. However, these bacteria typically degenerate and the nodule senesces prematurely (Vasse et al., 1990; Mitra and Long, 2004). In comparison, the heme moiety, possibly of leghemoglobin, was found to be highly enriched in the nitrogen fixation zone (Figure 4a), where the differentiation of bacteroids occurs in fixing nodules. In accordance with our observation, it was shown that the transcript for coproporphyrinogen III oxidase, one of the terminal enzymes of heme synthesis, is highly elevated in Glycine max (soybean) and Pisum sativum (pea) nodules compared with roots, particularly in the infected cells of the nodules where the protein moiety of leghemoglobin is also synthesized (Santana et al., 1998).

It is also possible that the heme moiety may belong to FixL of rhizobial origin, an oxygen-binding hemoprotein with kinase and phosphatase activities, which senses the oxygen levels directly and transmits this signal to FixJ via phosphorylation/dephosphorylation reactions (Gilles-Gonzalez et al., 1991). FixJ controls the expression of other regulatory genes, including nifA and fixK, that regulate the transcription of genes required for symbiotic nitrogen fixation (Reyrat et al., 1993). With the current techniques and knowledge, we were unable to differentiate the heme moiety of leghemoglobin from that of FixL.

Proline and glycine betaines are nitrogenous osmolytes that accumulate under osmotic stress conditions in plants (Hasegawa et al., 2000; Trinchant et al., 2004). It was shown that S. meliloti uses proline betaine as a carbon and nitrogen source or as an osmoprotectant (Gloux and Lerudulier, 1989; Goldmann et al., 1991). Given the competition among soil bacteria for plant carbon sources, the ability of S. meliloti to utilize proline betaine provides a selective advantage in colonization of legume roots (Phillips et al., 1992). In addition, proline betaine may also act as an inducer of nodulation (nod) genes in S. meliloti (Phillips et al., 1992). Our MALDI-MSI strategy identified a strong accumulation of proline betaine in root tissues compared to nodules. It is possible that a higher rate of proline betaine catabolism (as an energy source) by S. meliloti inside the nodule may be responsible for the lower proline betaine content in nodule tissues compared with root tissues.

Several metabolites, such as NAD or adenine, were also detected using DHB in the positive mode. However, these ions were not observed on the UHR-Q-TOF mass spectrometer used for this study. Despite the high concentration of these ions in nodules, their dilution in the homogenized root tissues probably prevented us from detecting them on an ESI-MS instrument. Therefore, the in situ tandem MS spectra of selected metabolites acquired on MALDI-TOF/TOF are critical to the determination of their identities. A comparison between the MS/MS spectra acquired on MALDI-TOF/TOF and conventional LC-ESI-Q-TOF showed identical fragments, as in Figure 2(f) and Figure S1(a), increasing our confidence in assignments based on MALDI-TOF/TOF-generated spectra. Moreover, the MALDI-TOF/TOF-generated MS/MS fragments were all well-matched to the theoretical structures (Figure 2), further supporting their identifications.

Flavonoids are polyphenolic compounds that constitute one of the most diverse classes of metabolites in higher plants. Here, we directly localized the distribution of flavonoids and their conjugates, formononetin and formononetin MalGlc, in Medicago roots and nodules. Flavonoids and isoflavonoids play essential roles at various steps of legume nodulation, from recruitment of compatible rhizobia to the regulation of nodule development (Subramanian et al., 2007). Flavonoids and isoflavonoids are released by the host plants, which act as chemo-attractants for the rhizobia to recognize their host (Peters et al., 1986; Maxwell et al., 1989; Hartwig et al., 1990a). These compounds activate the expression of several nod genes in rhizobia, leading to Nod factor production and secretion (Hartwig et al., 1990b; D'Haeze and Holsters, 2002). Localized inhibition of auxin transport mediated by flavonoids is also necessary for nodule development in the roots of legumes with indeterminate nodules, such as Medicago (Subramanian et al., 2007). Auxin transport is arrested at the site of infection by rhizobia, which is mediated by flavonoids (Subramanian et al., 2006). RNAi silencing of the isoflavone biosynthesis pathway resulted in increased auxin transport and defective nodule formation (Wasson et al., 2006). Although the function of isoflavonoid conjugates remains uncertain (Kessmann et al., 1990), the higher level of afrormosin MalGlc in nodules compared with roots suggests a distinct role from formononetin MalGlc in Medicago.

Although conventional methods may provide good coverage of metabolites detected from tissue extracts, MSI enables visualization of metabolites by direct analysis of tissue sections, as demonstrated here with Medicago roots and nodules. In this study, we pioneered use of a novel matrix, DMAN, for exploring the metabolome of plant tissues, and showed excellent compatibility with MS/MS and MSI. The interference-free matrix enabled direct mapping of acidic metabolites by MALDI-TOF/TOF in the negative mode, significantly expanding the coverage of the metabolites detected. A representative acidic molecule is malic acid, which was only detected in the negative mode using DMAN and is involved in the tricarboxylic acid (TCA) cycle, the glyoxylate and dicarboxylate metabolic pathway, and carbon fixation in photosynthetic organisms. This dicarboxylic acid is also the major energy source for the bacteroids, and is used as a carbon skeleton in the glutamine synthetase/glutamate synthase pathway (Stitt et al., 2002; Sulieman, 2011). Another organic acid, pyruvic acid, is the end product of glycolysis and the starting point of gluconeogenesis. Similarly, succinic acid was found to be an excellent substrate that supports the highest level of nitrogen fixation by isolated bacteroids (Kahn et al., 1985; Sulieman and Schulze, 2010). In addition to being intermediates in the key metabolic pathways, organic acids have also been proposed to be involved in many processes, including nutrient acquisition and uptake, metal detoxification, alleviation of anaerobic stress in roots, mineral weathering, and interaction with microbes in the rhizosphere (Jones, 1998). Most key organic acids involved in the TCA cycle were detected using DMAN, as illustrated in Figure S5. This observation will help our understanding not only of the legume–rhizobia symbiosis, but also of many other metabolic processes in plants (Shroff et al., 2009). Nevertheless, the MSI-based approach often lacks the selectivity to separate the metabolites desorbed simultaneously from Medicago tissue. Therefore, caution must be taken when evaluating the distribution images of metabolites that have isobaric ions, such as sucrose and anhydrous trehalose. Given their identical m/z and highly similar collision-induced dissociation fragmentation patterns, it is possible that both species were observed as one peak by MALDI-TOF/TOF in this study. Nevertheless, these two isobaric disaccharide species may be distinguished by a combination of ion mobility spectrometry and vacuum ultraviolet photodissociation techniques, as described recently (Lee et al., 2012a). In future investigation, we will employ ion mobility spectrometry to further separate these structural isomers to allow definitive characterization. In addition to isomers, the isotopic contribution of metabolites with adjacent masses may also result in merging of two species into one peak, such as the second isotopic peak of oxalacetate [C4H4O5-H] at m/z 132.0009 and the third isotopic peak of aspartate [C4H7NO4-H] at m/z 132.0291 on MALDI-TOF/TOF, whereas a high resolution MALDI-MS instrument may fully resolve the almost isobaric peaks and produce their corresponding distribution images unambiguously. Exploration of metabolic differences in nodule chemistry will be attempted with high-resolution MALDI-MS in the near future.

In conclusion, we have demonstrated the benefits of using MALDI-MSI to obtain unique and valuable information on the identity and spatial distribution of plant metabolites that cannot be obtained. The conventional matrix DHB and the novel matrix DMAN were complementary for profiling and imaging metabolites directly from tissue sections. Furthermore, MS/MS fragmentations were performed to facilitate the identification of region-specific metabolites. The knowledge obtained through comparison of metabolite profiles and molecular ion images obtained from nitrogen-fixing and non-fixing nodules highlighted the benefits of MALDI-MSI in understanding the roles of metabolites in symbiotic nitrogen fixation.

Experimental Procedures

Plant growth and inoculation with rhizobia

Medicago seeds were acid scarified, surface-sterilized and germinated as described previously (Catoira et al., 2000). One-day-old seedlings were placed on plates containing nitrogen-free modified Fahraeus medium (Catoira et al., 2000) overlaid with sterile germination paper, and grown for 7 days at 22°C, 16 h light (130–200 mmol m−2 sec−1) and 8 h dark in a growth chamber. The roots were grown in the dark by covering the root portion by aluminum foil. The roots were inoculated with S. meliloti Rm1021 (WT) and the fixJ mutant using a rhizobial suspension with an OD600 of 0.01, and incubated in the growth chamber for 3 weeks for nodule development. At approximately 3 weeks old, nodules were selected from each sample for metabolite imaging. Information on plant lines and bacterial strains is provided in Methods S1. The experimental set-up is shown in Figure S6.

Acetylene reduction assays

Acetylene reduction assays (Turner and Gibson, 1980) were performed to quantify the amount of nitrogen fixed in the nodules formed on Medicago roots. The details are described in Methods S2.

Tissue extraction

Nodulated roots were detached from the plant, flash-frozen, ground to powder in liquid nitrogen, and extracted using acidified ice-cold methanol (92:7.9:0.1 v/v/v MeOH/water/formic acid). The detailed procedure is provided in Methods S3.

Sample preparation for imaging

The nodules were excised, along with flanking root tissues, embedded in 100 mg/ml gelatin aqueous solution, and flash-frozen in liquid nitrogen. The frozen tissue was then sectioned into 12 μm slices at −20°C. The 12 μm thick intact longitudinal sections of Medicago nodules, together with the root tissue, were subjected to metabolite MSI. Corresponding methylene blue-stained nodule sections (Figure 1a) were used to correlate the MS images acquired. The sections for MSI were thaw-mounted either onto a MALDI plate or ITO glass slides, depending on specific instruments. A representative optical image of a cross-section thaw-mounted on a MALDI plate is shown in Figure 1(b). Subsequently, these sections were dehydrated, matrix-coated and subjected to MS imaging/profiling as illustrated in Figure 1(c). Different matrix application procedures were utilized for positive- and negative-mode MSI experiments. The experimental details are provided in Methods S4.


A Applied Biosystems ( 4800 MALDI-tandem-time-of-flight (TOF/TOF) analyzer was equipped with a 200 Hz, 355 nm Nd:YAG laser (spot diameter 50 μm) for negative-mode MS profiling and imaging. Positive-mode MSI was performed using an autoflex III MALDI-TOF/TOF (Bruker Daltonics). The instrument details are provided in Methods S5.

Image files obtained on the Applied Biosystems 4800 TOF/TOF mass spectrometer were processed, and extracted ion images were created using TissueView (Applied Biosystems). Image files obtained on the autoflex III were processed using flexImaging (Bruker Daltonics).


Direct infusion data were acquired by reconstituting the lyophilized Medicago root nodule extract, followed by injection on a UHR-Q-TOF mass spectrometer (Bruker Daltonics) in positive and negative modes. To reduce sample complexity and obtain tandem MS spectra for isomeric metabolites, the lyophilized Medicago root nodule extract was reconstituted in 0.1% formic acid and injected into a Waters NanoAcquity Ultra Performance capillary LC system ( coupled to a ESI-Q-TOF SYNAPT G2 mass spectrometer. The detailed experimental set-up is described in Methods S6.

Molecular identification of metabolites

Molecular identification of the metabolites involves comparing the masses acquired on MALDI-TOF/TOF from Medicago sections with those recorded by UHR-Q-TOF from Medicago extracts, and searching the matched masses against the METLIN (, KNApSAcK ( and human metabolome database ( and human metabolome databases. The metabolite list generated by mass matching was further confirmed by comparison with purchased standards, the databases or previous literature. The detailed experimental set-up is described in Methods S7.


This work was supported by a grant from the US National Science Foundation (NSF#0701846) to J.M.A., and by funding from the University of Wisconsin Graduate School, the Wisconsin Alumni Research Foundation and the Romnes Faculty Research Fellowship program to L.L.