The present paper describes matrix-free laser desorption/ionisation mass spectrometric imaging (LDI-MSI) of highly localized UV-absorbing secondary metabolites in plant tissues at single-cell resolution. The scope and limitations of the method are discussed with regard to plants of the genus Hypericum. Naphthodianthrones such as hypericin and pseudohypericin are traceable in dark glands on Hypericum leaves, placenta, stamens and styli; biflavonoids are also traceable in the pollen of this important phytomedical plant. The highest spatial resolution achieved, 10 μm, was much higher than that achieved by commonly used matrix-assisted laser desorption/ionization (MALDI) imaging protocols. The data from imaging experiments were supported by independent LDI-TOF/MS analysis of cryo-sectioned, laser-microdissected and freshly cut plant material. The results confirmed the suitability of combining laser microdissection (LMD) and LDI-TOF/MS or LDI-MSI to analyse localized plant secondary metabolites. Furthermore, Arabidopsis thaliana was analysed to demonstrate the feasibility of LDI-MSI for other commonly occurring compounds such as flavonoids. The organ-specific distribution of kaempferol, quercetin and isorhamnetin, and their glycosides, was imaged at the cellular level.
Recently, a variety of technologies have been successfully used to obtain data from individual plant cells (Lange, 2005). Both laser microdissection (LMD) (Emmert-Buck et al., 1996; Hölscher and Schneider, 2007; Li et al., 2007a) and cell sap sampling using microcapillaries (Brandt et al., 2002) have been reported to improve access to the contents of individual cells; these reports were based on analyses using both post-genomic bioanalytical technologies and spectroscopic methods (Lange, 2005). Techniques such as LMD allow researchers to avoid the averaging effects that occur when heterogeneous tissues, which represent the most abundant cell types, are pooled; however, as amplification methods such as commonly used for DNA and RNA are not available, highly sensitive detection methods are required for metabolites. Methods based on mass spectrometry offer the required high sensitivity and specificity (Sumner et al., 2003).
MS-dependent imaging techniques have been developed in the past to study the localization on a small scale of compounds from complex biological systems. Recently, a method for MS imaging has been developed that offers information about the distribution of proteins, peptides or metabolites with 100–300 μm spatial resolution (Caprioli et al., 1997; Cooks et al., 2006; Li et al., 2008; Seeley and Caprioli, 2008). Furthermore, when atmospheric pressure infrared matrix-assisted laser desorption/ionization mass spectrometry was used to study plant metabolites in the white lily (Lilium candidum L.) and other plants, over 50 small metabolites were discovered; these were involved in flavonoid biosynthesis and had a spatial resolution of 180–640 μm (Li et al., 2008). MALDI imaging involves use of a conventional MALDI source to desorb ions of interest from the sample covered by a sprayed (Rubakhin et al., 2005) or spotted matrix (Aerni et al., 2006). MALDI imaging has been used to study the distribution of a wide variety of metabolites, including drugs, peptides and proteins, in animal tissues (Reyzer and Caprioli, 2007; Goodwin et al., 2008), and herbicides (Mullen et al., 2005), peptides (Kondo et al., 2006) and sugars (Burrell et al., 2007; Li et al., 2007a,b) in plants. Recently, ion intensity maps were constructed from MALDI-TOF mass spectra to measure the spatial distribution of some secondary plant metabolites. The matrix 9-aminoacridine was evenly applied to the leaves of Arabidopsis thaliana (Col-0) in order to detect glucosinolates (Shroff et al., 2008). However, applying MALDI matrices to the tissues complicates tissue preparation for imaging and can disturb the native distribution of the studied metabolites (Shroff et al., 2008). Low-molecular-weight metabolites have been profiled and localized using colloidal graphite-assisted LDI (GALDI) MS in A. thaliana (Cha et al., 2008). Profiles and spatially resolved images of phospholipids, cerebrosides, oligosaccharides, flavonoids and other secondary metabolites were obtained in negative and positive ion MS modes. Certain mass imaging methods, such as laser-assisted electrospray ionization (LAESI) (Vertes et al., 2008), can reduce sample handling prior to analysis. However, the infrared laser currently used does not provide cell-like resolution for the obtained MS images.
The main secondary plant products of members of the genus Hypericum of the Hypericaceae plant family (Mabberley, 2008) are flavonoids (e.g. quercetin, hyperoside, rutin, quercitrin, isoquercitrin), xanthones (e.g. 1,3,6,7,-tetrahydroxyxanthone), prenylated phloroglucinols such as hyperforin and adhyperforin, biflavonoids [I3,II8-biapigenin and I3′,II8-biapigenin (amentoflavone)] and naphthodianthrones (Brockmann et al., 1942; Wolfender et al., 2003; Charchogylan et al., 2007; Smelcerovic et al., 2008) (Figure 1 and Table 1). Multi-cellular, globular or tunnel-shaped aggregates such as secretory canals, and dark and translucent glands have been reported to contain the secondary metabolites (Ciccarelli et al., 2001; Robson, 2003). Dark and pale glands often occur in the stem, leaf, sepal, petal or anthers. In Hypericum, the stamens possess a connective terminating in a dark gland (Figure 2). The gynoecium of H. perforatum has three styles and contains relatively small red stigmata (Figure 2). There is a large variation in the glandularity of sepals in Hypericum, and some species (such as H. reflexum) have stalked glands (Robson, 2003). These glandular trichomes are often close neighbours of dark glands, which are located near the margins of the sepals (Esau, 1965; Piovan et al., 2004). The short distance (<50 μm) between these glands makes the use of (MA)LDI techniques to acquire information from each dark gland extremely challenging.
Table 1. Names, deprotonated ion mass and UV adsorption wavelengths of the Hypericum sp. natural constituents (see Figure 1 for structural formula) (Jürgenliemk, 2001)
Numerous studies have been published describing the various approaches to analysing the constituents of H. perforatum, commonly known as St John’s wort. One of the best-selling herbal medicinal plants worldwide, this species is planted in several hundred hectares in Europe (Gaudin et al., 2003). HPLC and RPLC separations have been performed to identify and quantify active compounds in extracts of St John’s wort (Li and Fitzloff, 2001; Seger et al., 2004; Pages et al., 2006). Other publications reported analysis of the major secondary metabolites of H. perforatum based on measurements using LC/MS2 (Brolis et al., 1998; Ganzera et al., 2002) and LC/NMR/MS (Hansen et al., 1999). Previously, Tatsis et al. (2007) combined liquid chromatography/diode array detector/solid phase extraction/NMR spectroscopy and liquid chromatography/ultraviolet/tandem MS spectrometry. Generally speaking, modern techniques for identifying compounds are combined with a classical approach, namely extracting whole plant material such as leaves or flowers using organic solvents.
The model plant A. thaliana was chosen to investigate the tissue-specific accumulation of flavonoids in its flowers, sepals, petals and leaves, and thus to demonstrate the broader applicability of the matrix-free LDI-MSI method. Recently, flavonoids and cuticular waxes have been profiled and imaged from various plant surfaces, and cross-sections of A. thaliana were successfully probed by colloidal graphite-assisted LDI (GALDI) MS imaging (Cha et al., 2008). The size and fragility of A. thaliana presented challenges when using the GALDI-MSI method to analyse the metabolome.
The present paper reports on the use of matrix-free laser desorption/ionization mass spectrometric imaging (LDI-MSI) to study highly localized UV-absorbing secondary metabolites in plant tissues. The applicability of the method is demonstrated on the phytochemical contents of members of the plant genera Arabidopsis and Hypericum, and verified by mass spectrometry of the areas isolated using LMD in Hypericum.
Laser desorption/ionization time of flight (LDI-TOF) detection of Hypericum sp. and A. thaliana secondary metabolites
To prepare samples from various tissues of Hypericum sp. for LDI-TOF/MS analysis, two approaches were used: cell sap sampling using microcapillaries and LMD. The thickness of the plant cell walls and the fragility of the specially prepared fused silica capillaries (Figure 3c) (Kelly et al., 2006; Sproß, 2007) made it difficult to obtain sufficient amounts of the dark glands.
Dark glands from petals (Figure 3b) and dark and translucent glands from leaves (Figure 3a,d) were successfully separated by LMD. As a control, green parenchyma tissue from the leaves of H. perforatum (Figure 3a) was collected and extracted using methanol. Aliquots (1 μl) of the individual sample solutions were dried on the stainless steel plate typically used in MALDI instruments. Because phenolic structures deprotonate easily, the MALDI analysis was carried out in negative ion mode using 9-aminacridine as the matrix. Moreover, these analytes did not require a chemical matrix for ionization, probably due to their strong UV absorption. They ionized very well under UV irradiation with a conventional nitrogen laser (λ = 337 nm) fitted to the MALDI instrument. A comparison of the MALDI (matrix 9-aminoacridine; Shroff et al., 2007) and LDI spectra of hypericin and pseudohypericin showed no difference in sensitivity. Moreover, the LDI spectra were free of matrix peaks. Both MALDI and LDI methods showed comparable limit of detection for authentic compounds (5 pg).
We compared our data with the results of several phytochemical investigations of Hypericum secondary metabolites (Brockmann et al., 1942; Wolfender et al., 2003; Charchogylan et al., 2007; Tatsis et al., 2007; Smelcerovic et al., 2008). In the dark glands of the petals of H. perforatum, we detected the naphthodianthrones hypericin (7) (m/z 503.07, [M-H]−) and pseudohypericin (8) (m/z 519.07, [M-H]−) and the flavonoids quercetin (1) (m/z 301.03, [M-H]−), quercitrin (2) (m/z 447.07, [M-H]−) and rutin (6) (m/z 609.14, [M-H]−), as well as isoquercitrin (3) (m/z 463.08, [M-H]−) and/or hyperoside (4) (m/z 463.08, [M-H]−) (indistinguishable by our methods) (Figure 4a) (Tatsis et al., 2007). Compared to the dark glands of the petals, the black nodules of the leaves show two additional compounds: protohypericin (9) (m/z 505.09, [M-H]−) and protopseudophypericin (10) (m/z 521.08, [M-H]−) (Figure 4b). The LDI-TOF/MS analysis of the translucent glands of the leaves showed signals for three phloroglucinols: hypeforin (11) (m/z 535.17, [M-H]−), adhyperfirin (12) (m/z 481.33, [M-H]−) and hyperfirin (13) (m/z 467.31, [M-H]−) (Figure 4c) (Tatsis et al., 2007). The LMD-prepared dark glandular cells of the connective tissue between the theca of the stamens (Figure 2) showed a signal profile comprising hypericin (7) (m/z 503.07, [M-H]−), pseudohypericin (8) (m/z 519.07, [M-H]−), protohypericin (9) (m/z 505.09, [M-H]−) and protopseudophypericin (10), (m/z 521.08, [M-H]−) and the flavonoid quercetin (1) (m/z 301.03, [M-H]−). Furthermore, the most intense signal of this spectrum was produced by the biflavonoid biapigenin (14) (m/z 537.08, [M-H]−) (Figure 4d). In previous reports, only traces of isobaric amentoflavones were found in H. perforatum (Repčák and Martonfi, 1997; Nebelmeir, 2006). The prepared pollen showed the biapigenin signal almost exclusively (14) (m/z 537.08, [M-H]−), and only traces of the naphthodianthrones hypericin (7) (m/z 503.07, [M-H]−), pseudohypericin (8) (m/z 519.07, [M-H]−), protohypericin (9) (m/z 505.09, [M-H]−) and protopseudohypericin (10) (m/z 521.08, [M-H]−) (Figure 4e). The spectra of the lighter red papillae of the styli showed signals for hypericin (7) (m/z 503.07, [M-H]−), pseudohypericin (8) (m/z 519.07, [M-H]−), protohypericin (9) (m/z 505.09, [M-H]−), protopseudophypericin (10) (m/z 521.08, [M-H]−), the flavonoids quercetin (1) (m/z 301.03, [M-H]−), isoquercitrin (3) (m/z 463.08, [M-H]−) and/or hyperoside (4) (m/z 463.08, [M-H]−), and the phloroglucinol hyperforin (11) (m/z 535.37, [M-H]−). Once again, the strongest signal was produced by the biflavonoid biapigenin (14) (m/z 537.08, [M-H]−) (Figure 4f). Investigation of the non-secretory parenchyma tissue showed only traces of the flavonoids quercetin (1) (m/z 301.03, [M-H]−) and rutin (6) (m/z 609.14, [M-H]−) and the phloroglucinol hyperforin (11) (m/z 535.37, [M-H]−).
Thus, use of LMD and MALDI- or LDI-MS has proven to be a powerful combination for obtaining information about the phytochemical profile of specialized plant areas. The ability to ionize the investigated hypericins without the use of matrix during the LDI process simplifies sample preparation and enables direct profiling of secondary plant metabolites by MSI.
Direct profiling/imaging of flavonoids, biflavonoids and hypericins by LDI-MSI
The commonly used MS imaging technique, which involves application of a matrix, has certain disadvantages. If the tissue is wetted when the matrix is applied, significant analyte de-localization may occur, leading to artefacts. Additionally, the size of matrix crystals formed after multiple spraying is highly variable and sometimes larger than the required spatial resolution. Hence, the best solution is to analyse the tissue without any chemical treatment in order to observe the native distribution of compounds. However, this is generally not feasible, as most analytes do not strongly absorb UV light and require a matrix to be desorbed. However, the efficient UV absorption of hypericins suggested that the LDI process could be used for imaging these compounds in H. perforatum tissues without the need for a matrix.
Preliminary experiments were performed using a MALDI micro instrument (Shroff et al., 2008); however, the laser beam (approximately 50 μm) could not be focused to allow spatial information to be differentiated on a fine scale. Higher resolution could be achieved using the Ultraflex III®, in which the laser can be focused on spots of <10 μm. We were able to resolve the localized distribution of hypericin and pseudohypericin in both secretory cavities, separated from each other by the peduncle of the glandular trichomes (Figure 5), in H. reflexum leaves. Additionally, LDI-MSI made it possible to detect hypericin, pseudohypericin, quercetin and rutin in the small red appendices of the placenta; supposedly from one cell, these appendices are easily prepared from cryo-sectioned flower material (Figure 6). Difficulties in performing LDI-MSI experiments with the three-dimensional bulky stamens and styli were overcome by LMD. The filament was removed and samples suitable for measurement were obtained (Figure 7a,b). LDI-MSI allowed hypericin (m/z 503.07, [M-H]−) (Figure 7c,e) and pseudohypericin (m/z 519.07, [M-H]−) (Figure 7d,f) to be detected in the dark connective theca stamens. Signals for hypericins and biflavonoids were detectable on the stigma as a result of the presence of pollen grains on the papillae (image not shown, and Figure 8a). Signals typical for the biflavonoids biapigenin (m/z 537.08, [M-H]−) and amentoflavone (m/z 537.08, [M-H]−) were found in the pollen (Figure 8b). A laser intensity of 283 μJ mm−2 was sufficient for efficient desorption/ionization of the analytes. Images constructed for the whole leaf were composed of approximately 9000 pixels, and the images constructed for smaller structures such as the stamens, styli and pollen were all more than 1000 pixels. Although a spatial resolution of 10 μm was required to differentiate highly localized regions, the high frequency (200 Hz) of the Nd:YAG laser made it possible for the measurements to be performed in a reasonable time span.
The ease with which biflavonoids could be analysed using LDI inspired us to study whether other phenolic compounds could be desorbed from plant material. GALDI imaging has shown the distribution of three flavonoids (kaempferol, quercetin and isorhamnetin) and their glycosides at low spatial resolution in sepals and petals from A. thaliana (Cha et al., 2008). Mounting detached A. thaliana Col-0 petals on conductive tape (Figure 9a,f) and LDI-MSI imaging using the Ultraflex III® at 10 μm resolution within the scanning range m/z 100–800 we were able to obtain highly resolved ion images with cell-like structures (Figure 9). Very strong signals for many putative flavonoids and corresponding glycosides were visible, with particular patterns for individual ions. Using CID spectra (TOF/TOF) and published data (Cha et al., 2008), we were able to identify individual ions. The upper two-thirds of most petal surfaces contain kaempferol (m/z 285.02, Figure 9b) and kaempferol-rhamnoside (m/z 431.04, Figure 9d). Upon closer inspection, petal veins show very low concentrations of these compounds. An identical distribution was observed for highly glycosylated kaempferols (data not shown), e.g. m/z 577 (kaempferol-dirhamnoside), m/z 593 (kaempferol-rhamnoside-glucoside) and m/z 739 (kaempferol-dirhamnoside-glucoside).
Quercetin and isorhamnetin (m/z 301, Figure 9c and m/z 315, Figure 9e, respectively) are co-localized in the lower third of the petal, and corresponding glycosylated forms showed an identical pattern, e.g. m/z 447 (quercetin-rhamnoside), m/z 461 (isorhamnetin-rhamnoside) and m/z 609 (quercetin-rhamnoside-glucoside). In contrast to the kaempferol family, quercetin and isorhamnetin were detected in petal veins. If the images from both distribution patterns are superimposed, all compounds are seen to be co-localized in the lower petal mid-vein, suggesting that they share a common means of transport. In sepals, kaempferol/quercetin and isorhamnetin show a somewhat less distinct distribution than in petals (Figure 9f–j). Kaempferol (Figure 9g) and its rhamnoside (Figure 9i) are more abundant in the internal parts of the plant, whereas quercetin (Figure 9h) and isorhamnetin (Figure 9j) occupy the whole leaf surface.
The detection of variations in the molecular content of cell populations requires analytical methods based on single cells to avoid the pooling of data that occurs when constituents are averaged over multiple cells. Therefore, we used various strategies to acquire information about the distribution of secondary natural products in cryo-sectioned, laser-microdissected or freshly cut plant samples of Hypericum sp. The results show the suitability of combining LMD and LDI-TOF/MS and LDI-MSI to analyse the localization of secondary metabolites of H. perforatum. This approach should minimize the negative effects caused by application of a matrix. No additional chemical interactions, no additional data-distorting diffusion processes, and no additional changes of the matrix-treated surface of the sample are expected.
The phytochemical constituents of members of the genus Hypericum, and especially H. perforatum, have been described frequently, allowing us to compare our molecular data for the secondary metabolites with published data. Recent publications that provide an overview of the phytochemical profiles of various parts of leaves and flowers of H. perforatum include those by Berghöfer and Hölzl (1989), Nebelmeir (2006) and Tatsis et al. (2007). Our methods permitted identification of most compounds described in these reports, but could not distinguish the biflavonoids biapigenin and amentoflavone from the flavonoids hyperoside and isoquercitrin, which would have required the use of LC-MS equipment.
Naphthodianthrones were detectable in the dark glands of leaves and petals and in certain parts of the connective tissue of the theca of the stamens. With the higher resolution afforded by use of the Ultraflex III® apparatus, we were able to analyse the contents of closely adjacent anatomical structures on H. reflexum leaves that were separated from each other by only 50 μm; our analysis revealed highly localized hypericins in the glandular trichomes and dark glands of H. reflexum, and avoided the time-consuming steps required to mechanically isolate glandular trichomes and determine their natural products (Piovan et al., 2004). Furthermore, the LDI-MSI results can be easily re-checked by LMD and LDI-TOF/MS or LC/MS/MS analysis.
The spatial resolution of the phytochemical profile of neighbouring secretory cavities of Hypericum was excellent, enabling us to investigate even smaller areas such as the small appendices of the placenta – to do this requires collection of data on a scale <10 μm. Proof of the presence of naphthodianthrones and flavonoids in the small appendices of the placenta was provided using the Smartbeam® optics of the Ultraflex III® equipment, which are able to acquire data from localized secondary metabolites.
Recently, hyperforin was detected in disc-shaped leaf material containing translucent glands of H. perforatum that was manually isolated using modified syringes (Soelberg et al., 2007). Our approach, namely combining LMD with LDI-TOF/MS and LDI-MSI imaging, allowed phloroglucinols to be detected in individual translucent glands, and confirmed that hyperfirin and adhyperfirin accumulate in these secretory cavities (Soelberg et al., 2007; Tatsis et al., 2007). Surprisingly, adhyperfirin was not detected in our samples. This variability of the phytochemical profile could be due to different environmental factors (Nebelmeir, 2006). LMD was also used to prepare the papillae-containing styli of the stigma. In addition to hypericins, MS signals typical for the biflavonoids biapigenin and amentoflavone were detected. Both compounds are thought to be localized in the stamens (Repčák and Martonfi, 1997; Nebelmeir, 2006), and, in comparison to biapigenin, only traces of amentoflavone are verifiable. Evidence of biflavonoids in the papillae comes from the presence of pollen on the styli. The compounds were detected and separated using LC-MS techniques on extracts of LMD samples. The presence of flavonoids appears to be connected with hypericin-containing secretory structures. Only traces of quercetin and rutin were detectable in non-secretory parenchyma cells from H. perforatum.
The well-known UV demarcations of the flower of H. perforatum result from two categories of pigments, flavonoids and de-aromatized isoprenylated phloroglucinols (Gronquist et al., 2001). Our investigation did not reveal any de-aromatized isoprenylated phloroglucinols. Future investigations will focus on the presence of secondary metabolites during the anatomical development of plant parts such as the leaves, stamens or placenta of Hypericum sp. The ability to analyse cell-specific metabolite patterns makes it possible to study the location of the biosynthesis of secondary metabolites and putative translocations. Furthermore, LDI-MSI will allow us to investigate the influence of genetic and environmental factors on the accumulation of secondary metabolites of Hypericum sp. and other plants without using a matrix.
Natural populations and breeding lines of H. perforatum possess a variety of bioactive constituents. LDI-MSI can be used on this important phytomedical plant to address harvest quality, the drying process, and storage of the plant material. Further experimental developments should enable us to analyse active compounds in species at various vegetative stages and in specific parts of the plant. Such analysis will help to define the optimal harvest time for obtaining high-quality raw material.
Anthracnose disease of H. perforatum caused by Colletotrichum gloeosporioides Penz. leads to the development of symptoms such as the reddish colour of infected plants (Gaudin et al., 2003). LDI-MSI will be useful for studying the role of secondary plant metabolites such as phytoalexins, whose concentration increases after pathogen attack; it is an elegant tool for investigating the uptake, metabolism and distribution of polycyclic aromatic plant metabolites in host–pathogen interactions.
Using LDI-MSI on A. thaliana Col-0 petals and sepals provided clear MS images (Figure 9) at cellular resolution (approximately 10 μm), and indicates the broad scope of the method. The previously determined distribution of flavonoids reported by Cha et al. (2008) was re-examined. We found that all the compounds are co-localized in the main basal petal vein, and only glycosides of the quercetin and isorhamnetin type occur in small veins. The mechanistic cause of this pattern cannot be determined from the current dataset, but the transport from other plant parts is likely. A possible function of the observed distribution could be that the localization of highly UV-absorbing pigments (the quercetin and isorhamnetin series) in the petal base directs pollinators to the nectar source. A related UV-absorbing guide mark has been isolated from the petal base of the related species Brassica rapa, and identified as isorhamnetin-3,7-O-di-β-d-glucopyranoside (Sasaki and Takahashi, 2002).
In principle, any UV-absorbing compound could be imaged on the scale of individual cells as performed here. All compounds with condensed aromatic rings, such as flavonoids, flavanes, anthocyanins or similar plant pigments, or indolic alkaloids and related structures, are targets for future application of the LDI-MSI method illustrated here. It is encouraging that plants can be imaged without cryo-sectioning, and that organs/glands in the body of the plant tissues are suitable for MSI. The UV laser is able to penetrate a few micrometres into the sample, enough to desorb and ionize the compounds of interest. This method could also be applied to samples of animal or microbial origin.
Plants of H. perforatum L. were obtained from Agrarprodukte Ludwigshof eG and Martin Bauer GmbH Co. KG (http://www.martin-bauer.de). They were grown outside close to the greenhouse of the Max Planck Institute for Chemical Ecology. Plants of H. reflexum L. were obtained from the Botanical Garden of Martin Luther University (Halle-Wittenberg, Germany). These plants were grown in either sand or soil (Klasmann Erden (http://www.klasmann.deilmann.de) clay and sand in the ratio 1:2) under greenhouse conditions (day 20–22°C, night 18–20°C; 30–55% relative humidity; the natural photoperiod was supplemented by 8 h of light from a Philips Sun-T Agro 400 W sodium light, http://www.philips.com).
A. thaliana plants of Columbia accession (Col-0) were grown in a soil:vermiculite mixture (3:1) in a controlled environment chamber at 21°C, 55% relative humidity, 100 μmol m−2 sec−1 photosynthetically active radiation from a mixture of Fluora and Cool White lamps (Osram, http://www.osram.de), and a diurnal cycle of 10 h light/14 h dark.
Laser microdissection of secretory cavities
Using to the procedure described by Hölscher and Schneider (2007), leaves, sepals and petals of H. perforatum or H. reflexum were fixed between a thin glass slide (0.6 mm thickness, Menzel Gläser, http://www.menzel.de) and a specially manufactured metal frame. The dissection of the secretory cavities was made possible by use of a nitrogen solid-state diode laser of a Leica LMD6000 system (http://www.leica-microsystems.com/) with short pulse duration (355 nm). The microdissected secretory cavities were collected in the lid of an Eppendorf tube by the Leica LMD6000 system. The microtube was briefly centrifuged (1000 g, 25°C, 1 min) to settle the contents. Methanol was added and the mixture was sonicated for 1 min, and then transferred to a MALDI plate for (MA)LDI-TOF/MS investigation.
Sap sampling using microcapillaries
Fused silica capillaries were used as microcapillaries. Their etching using hydrofluoric acid was performed as described previously (Kelly et al., 2006; Sproß, 2007). The micro-injection apparatus consists of a stereo microscope (Zeiss Stemi SV11, http://www.zeiss.com/) with a magnification range of 6 × to 60 ×. The microcapillaries were connected with a micromanipulator and a homemade air-pulse controller system manufactured by the Max Planck Institute for Chemical Ecology workshop according to the plans described by Handler (2000). Injection of the microcapillaries and ejection of the contents of the specialized cells of Hypericum sp. were performed in the open air.
A MALDI micro MX mass spectrometer (Waters, http://www.waters.com) fitted with a nitrogen laser (337 nm, 4 nsec laser pulse duration, 10 Hz and 154 μJ per pulse) was used in reflectron mode and negative polarity for data acquisition using MassLynx version 4.0 software. The chemical identity of the compounds observed was confirmed by comparing MS/MS spectra of standard compounds and the masses obtained using an LTQ ion trap instrument (Thermo Fisher, http://www.thermo.com) with an atmospheric pressure MALDI source equipped with a solid-state Nd:YAG UV laser (MassTech, http://www.apmaldi.com) and running Target 6 (MassTech) and Excalibur version 2.0 (Thermo Fisher) software for data acquisition.
Fixation of plant material for Ultraflex III
Sepal and petal (A. thaliana), laser-microdissected stigma, untreated leaves, theca and pollen (leaves from H. reflexum, others from H. perforatum) were manually separated from the rest of the stamens using micro chisels (Eppendorf, http://www.eppendorf.com) and fixed on carbon conductive adhesive tape (Plano, http://www.plano-em.de), which was in turn fixed on an ITO slide (Bruker Daltonic, http://www.bdal.com). The cryo-sectioned placenta slices (thickness 60 μm) were directly transferred to an indium-tin-oxide (ITO) glass slide. A Staedtler triplus gel-liner (silver, 0.4 mm; Staedtler-Mars, Nürnberg, Germany, http://www.staedler.com) was used to place marks close to the samples to define their position. The granular particles of this gel-roller marker were useful to alleviate the laser positioning during LDI-MSI. Stereomicroscopic images were developed using Image J software (National Institutes of Health, http://rsb.info.nih.gov/ij/).
Imaging on the Ultraflex III® mass spectrometer
An Ultraflex III® mass spectrometer (Bruker Daltonics) was used for the analysis. The instrument was equipped with a Nd:YAG laser. All spectra were measured in negative reflectron mode. To measure the pixels of ca 10 × 10 μm, the minimum laser focus setting (corresponding to a diameter of about 10 μm laser, under-usage of over-sampling) was used. For each raster point, a spectrum was accumulated with 40 laser shots and fixed laser intensity. For image reconstruction, FlexImaging version 2.0 software (Bruker Daltonics) was used. For LDI-MSI of the placenta, 4633 raster positions were measured; 4125 positions for the stamens and 1107 positions for the styli and pollen. All signals within a mass range of m/z 300–600 were recorded. Leaf imaging was performed at 9262 raster positions within a mass range of m/z 300–700. Forty laser shots per pixel were collected from each position, and then the software calculated the average mass spectrum.
We wish to thank Tamara Krügel and the greenhouse team at the Max Planck Institute (Jena, Germany), for raising the Hypericum and Arabidopsis plants. We thank Eva Bremer and Sabine Stahl for providing H. reflexum (Botanical Garden, University Halle-Wittenberg, Halle, Germany), and Louise Hauke (Agrarprodukte Ludwigshof eG, Ranis-Ludwigshof, Germany) and Hans-Jürgen Hannig (Martin Bauer GmbH & Co. KG, Vestenbergsgreuth, Germany) for providing H. perforatum. We are also grateful to Regina Schenk (Humboldt University of Berlin, Germany), Ute Gärber (Julius Kühn-Institut, Bundesforschungsanstalt für Kulturpflanzen, Kleinmachnow, Germany), Wolf-Dieter Blüthner (N.L. Chrestensen GmbH, Erfurt, Germany), Heike Heklau (Botanical Garden University Halle-Wittenberg, Halle, Germany), Hermann Manitz (Herbarium Haussknecht, Jena, Germany) and Johannes Sebastian Nebelmeir (Technical University München-Weihenstephan, Germany) for helpful discussions. We thank Dr Wilmar Schwabe GmbH & Co. KG (Karlsruhe, Germany) for the generous gift of reference compounds, Finzelberg GmbH & Co KG (Andernach, Germany) for larger amounts of herba hyperici perforati (minimum 0.1% dianthrone), Roland Kilper (aura optik GmbH, Jena, Germany) for help in establishing the sampling of cell contents using microcapillaries, Jens Sproß for introduction to the preparation of fused silica micro-capillaries, and Stephan Imhof (Fachbereich Biologie, University of Marburg, Germany) for permission to use a photograph of a flower of H. perforatum. We also thank Sören-Oliver Deininger (Bruker Daltonics) for technical support and discussion. The authors gratefully acknowledge Emily Wheeler for editing help and Daniel Veit for constructing special slides for LMD and construction of the air-pulse controller system.