Lipidomics in situ: Insights into plant lipid metabolism from high resolution spatial maps of metabolites

Progress in Lipid Research 54 (2014) 32–52

Contents lists available at ScienceDirect

Progress in Lipid Research journal homepage: www.elsevier.com/locate/plipres

Review

Lipidomics in situ: Insights into plant lipid metabolism from high resolution spatial maps of metabolites Patrick J. Horn 1, Kent D. Chapman ⇑ Center for Plant Lipid Research, Department of Biological Sciences, University of North Texas, Denton, TX 76203-5017, USA

a r t i c l e

i n f o

Article history: Received 26 September 2013 Received in revised form 14 January 2014 Accepted 14 January 2014 Available online 27 January 2014 Keywords: Mass spectrometry imaging Seeds Lipids Triacylglycerols Phospholipids MALDI

a b s t r a c t The emergence of ‘omics’ technologies (i.e. genomics, proteomics, metabolomics, etc.) have revealed new avenues for exploring plant metabolism through data-rich experimentation and integration of complementary methodologies. Over the past decade, the lipidomics field has benefited from advances in instrumentation, especially mass spectrometry (MS)-based approaches that are well-suited for detailed lipid analysis. The broad classification of what constitutes a lipid lends itself to a structurally diverse range of molecules that contribute to a variety of biological processes in plants including membrane structure and transport, primary and secondary metabolism, abiotic and biotic stress tolerances, extracellular and intracellular signaling, and energy-rich storage of carbon. Progress in these research areas has been advanced in part through approaches analyzing chemical compositions of lipids in extracts from cells, tissues and/or whole organisms (e.g. shotgun lipidomics), and through visualization approaches primarily through microscopy-based methodologies (e.g. fluorescence, bright field, electron microscopy, etc.). While these techniques on their own provide rich biochemical and biological information, coordinated analyses of the complexity of lipid composition with the localization of these lipids at a high spatial resolution will help to develop a new level of understanding of lipid metabolism within the context of tissue/cellular compartmentation. This review will elaborate on recent advances of one such approach – mass spectrometry imaging (MSI) – that integrates in situ visualization with chemical-based lipidomics. We will illustrate, with an emphasis on oilseed lipid metabolism, how MS imaging can provide new insights and questions related to the spatial compartmentation of lipid metabolism in plants. Further it will be apparent that this MS imaging approach has broad application in plant metabolic research well beyond that of triacylglycerol biosynthesis in oilseeds. Ó 2014 Elsevier Ltd. All rights reserved.

Contents 1.

Overview of mass spectrometry imaging of lipid metabolites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1. Secondary ion mass spectrometry (SIMS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2. Desorption electrospray ionization-mass spectrometry (DESI-MS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3. Matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4. Metabolite image reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5. Considerations and precautions for in situ lipidomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.1. Sample preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.2. Matrix selection and deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.3. Metabolite (in-source) fragmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.4. Ion suppression. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.5. Accurate mass measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

33 33 33 34 35 36 36 36 37 37 37

Abbreviations: MS, mass spectrometry; MSI, mass spectrometry imaging; MALDI, matrix-assisted laser desorption/ionization; ESI, electrospray ionization; NMR, nuclear magnetic resonance; TAG, triacylglycerol; DAG, diacylglycerol; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PA, phosphatidic acid; PI, phosphatidylinositol; DGAT, diacylglycerol acyltransferase; PDAT, phosphatidylcholine: diacylglycerol acyltransferase; FA, fatty acids; P, palmitic (16:0); Ln, linolenic (18:3); L, linoleic (18:2); O, oleic (18:1); S, stearic (18:0); G, gadoleic (20:1); numerical designation of lipids indicates number of carbons in acyl chains, number of double bonds. ⇑ Corresponding author. Address: University of North Texas, 1155 Union Circle #305220, Denton, TX 76203-5017, USA. Tel.: +1 940 565 2969; fax: +1 940 565 4136. E-mail address: [email protected] (K.D. Chapman). 1 Present address: Department of Plant Biology, Michigan State University, East Lansing, MI 48824-1312, USA. 0163-7827/$ - see front matter Ó 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.plipres.2014.01.003

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

2.

3.

4.

1.5.6. Reproducibility and validation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Learning lessons about oilseed metabolism from metabolite location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Does heterogeneity of TAG metabolites suggest differences in localization of enzymes? Yes, no and maybe . . . . . . . . . . . . . . . . . . . . . . . 2.1.1. Cyclopropane fatty acid (CPFA) synthase/desaturase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.2. Delta-12 fatty acid desaturase (FAD2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.3. Delta-15 fatty acid desaturase (FAD3) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.4. Fatty acid elongase 1 (FAE1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.5. Tissue-specific acyltransferase activities? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Total oil distribution by nuclear magnetic resonance (NMR) approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3. Spatial information informs metabolic engineering experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Imaging phospholipids. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Phosphatidylcholine as a TAG metabolite and structural membrane lipid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Phosphatidylethanolamine (PE) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3. Phosphatidic acid (PA) and phosphatidylinositol (PI) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Potential for future development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. Improvements in spatial resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Metabolite identification and quantification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3. Expanding range of metabolites and tissues for imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4. Co-localization of metabolites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5. Alternative imaging approaches and correlated imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.6. Combine with transcriptomics and/or proteomics to co-localize metabolites, enzymes and expressed genes. . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1. Overview of mass spectrometry imaging of lipid metabolites Mass spectrometry imaging (MSI, sometimes also abbreviated IMS) is defined by a group of analytical platforms that map the location and relative abundance of metabolites in situ. Each platform operates with a similar overall process by rastering over a tissue surface and producing a set of ions from endogenous metabolites that can be identified and visualized by measuring their mass-to-charge ratios (Fig. 1). The ability to correlate location and metabolite abundance is the key to MS imaging. It is this attribute and improved instrument accessibility [1] that has resulted in a steady increase in number of publications on MS imaging in the last few years (Fig. 2) both in terms of technical advances as well as addressing specific biological questions. In contrast to conventional chemical extracts where spatial information is minimal or absent, in MSI, prepared tissue samples are analyzed directly retaining crucial spatial information for enhanced biochemical characterization. The three primary MS imaging ionization sources adopted commercially to-date include secondary ion MS (SIMS), desorption electrospray ionization (DESI) MS, and matrix-assisted laser desorption/ionization (MALDI) MS. Each platform varies in its mechanism of ionization and therefore offers different levels of spatial resolution (i.e. smallest distance that offers a distinguishable chemical profile) and available imaging applications. Integration of these ionization sources with high-resolution, accurate mass analyzers has enabled high-confidence identification of the ions generated by MSI. Once the raw data is acquired for each MSI method, a spatial map of an ion or set of ions representing a metabolite, protein, or class of desired molecules, can then be reconstructed and visualized using specialized imaging software. Several recent reviews have been published detailing some of the advances in MS imaging [2–4]. Here we will focus mostly on features, considerations, and applications of MALDI-MSI to plant lipid research.

1.1. Secondary ion mass spectrometry (SIMS) SIMS operates through the emission of secondary metabolite ions from a tissue surface following bombardment by a primary ion beam (1–40 keV) [5,6]. While SIMS was one of first imaging methodologies developed [7–9], recent modifications in SIMS

33

37 37 39 40 41 43 43 43 44 44 45 45 46 46 47 47 47 48 48 49 49 49

instrumentation now allow imaging of biomolecules up to m/z 1000 [10] which has widened its applicability for metabolites but not yet adequate for most proteomics approaches. SIMS sampling resolution of 400 nm to 1–2 lm [11] is superior to comparable imaging platforms and allows potential subcellular imaging and at attomolar concentrations [12]. Unfortunately, SIMS also suffers from extensive fragmentation and low ion yields [5]. Since SIMS requires a strong vacuum, sample preparation is critical to maintaining structural integrity. Sample preparation methods for tissue analysis (similar to those also required for MALDI) including stages of freezing, sectioning, and drying have been adapted from histological protocols. Methods are suitable for cellular or subcellular analysis with minor modifications such as eliminating freeze-drying [6], or through frozen hydration where cells are kept frozen throughout analysis without a drying step, which, incidentally, was found to enhance phospholipid signals [13]. Most of the applications to date for lipid imaging by SIMS [6] have been restricted to animal tissues [14–16] including lipid classes such as fatty acyls, glycerophospholipids, sphingolipids, sterol lipids, and prenol lipids. For plant lipid imaging, the applications are fewer due to the availability of instruments and difficulties in sample preparation but do include a recent report of successfully imaging flavonoids in pea (Pisum sativum) and Arabidopsis thaliana seed sections [17].

1.2. Desorption electrospray ionization-mass spectrometry (DESI-MS) DESI operates through the generation of charged secondary droplets containing metabolite ions by directing pneumatically-assisted charged droplets at a tissue surface under atmospheric pressure [18]. DESI requires little sample preparation, offering a simpler alternative to SIMS and MALDI. Tissues are often imprinted on materials such as porous Teflon [19] as an alternative to tissue sections. Unfortunately, due to the ionization mechanism, DESI in most cases is currently limited to spatial resolutions around 100– 250 lm [20,21]. A recent report, imaging lipids in brain tissues with morphologically distinct features, demonstrates that spatial resolutions of 35 lm are possible through optimization of DESI experimental parameters [22]. Due to the nature of DESI, much of the experimentation in plant tissues has focused on the surface analysis of lipids [23] in particular on leaves and petals [24], or on secondary metabolites in various tissues [19,25–27].

34

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

A

B co Y

Laser

co

X

ea

ea

C

Laser Phospholipids

Triacylglycerols mass spectrometer

Ions

D

E TAG Species: Acyl Chains: Netural Formula: Adduct: Theoretical m/z : Mean Measured m/z: X-Coord Y-Coord

875 850 825 800 775 750 725 700 675 650 625 600

1075 1075 1075 1075 1075 1075 1075 1075 1075 1075 1075 1075

TAG-52:4 TAG-16:0/18:2/18:2 C55H98O6 K+ 893.700 893.701

TAG-54:6 TAG-18:2/18:2/18:2 C57H98O6 K+ 917.700 917.702 Ion Intensity

TAG-56:6 TAG-18:2/18:3/20:1 C59H102O6 K+ 945.731 945.734

0 3.1E+04 2.5E+04 2.0E+05 1.4E+05 1.6E+05 1.1E+05 1.5E+05 2.1E+05 2.3E+05 7.9E+04 6.7E+04

0 4.4E+04 6.6E+04 2.8E+05 1.9E+05 2.3E+05 1.6E+05 2.4E+05 2.9E+05 3.0E+05 9.3E+04 8.6E+04

0 2.8E+04 4.5E+04 2.1E+05 1.6E+05 1.6E+05 1.0E+05 1.3E+05 1.5E+05 2.0E+05 6.0E+04 4.9E+04

Fig. 1. Mass spectrometry imaging (MSI) schematic. (A) Plant tissues to be imaged are prepared and cryosectioned with caution to preserve metabolite quantities and localization (see Section 1.5). The example shown is a Camelina sativa seed representative cross section (30 lm thickness) with labeled embryonic axis (ea) and cotyledons (co). Scale bars = 500 lm. (B) MSI experiments rasterize across an x- and y-coordinate plane at selected spatial resolutions (e.g. 25 lm) generating metabolite ions at each spot selected. (C) At each spot selected, the ionization source (i.e. laser) generates a set of metabolite (and other) ions that are directed into the mass spectrometer where their mass-to-charge ratios are measured. (D) Metabolites can be preliminarily identified using the accurate mass and high resolution capabilities of MS imaging instrumentation. At each spot, the absolute intensities (and relative amounts) can be extracted and annotated for each metabolite detected. (E) Ultimately, the raw metabolite information in part D is analyzed by computational software (see Section 1.4) where MS images of metabolite distribution can be reconstructed and visualized.

1.3. Matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) A third major type MS imaging methodology, MALDI-MS, has been used to generate spatial maps of plant metabolites especially lipids [28,29]. MALDI is a soft ionization method which relies on a sample embedded with a chemical matrix. This chemical matrix

(e.g. commonly an organic acid such as 2,5-dihydroxybenzoic acid, DHB) promotes ionization of the metabolites and reduces the likelihood of these biomolecules being destroyed upon the ionization/ desorption event. The chemical matrix is typically applied to tissue sections which must be carefully prepared as with SIMS (Section 1.1) due to the ionization event occurring under a strong vacuum. A laser beam is rastered over the matrix-embedded tissue

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52 300

250

Publications

200

150

100

50

0 2005

2006

2007

2008

2009

2010

2011

2012

2013

Year of Publication Fig. 2. Publications in mass spectrometry imaging field. Number of peer-reviewed scientific publications containing the term ‘‘imaging mass spectrometry’’ or ‘‘mass spectrometry imaging’’ from the Web of KnowledgeSM database.

surface to generate ions at each spot where the laser shot is applied. The diameter of the applied laser beam frequently limits the spatial resolution of MALDI, most often to >5 lm in current instruments. This microprobe style of imaging relies on positioncorrelated image reconstruction; however, MALDI can also be set up as a microscope-style instrument using a defocused laser source and a position-sensitive detector [30]. The utilization of MALDI as an imaging technology [3] has evolved from many years of using it as a soft ionization source for sampling of proteins [31] and lipids [32] in solutions spotted onto metal stages. The ability to interface MALDI with several downstream MS detectors, its excellent spatial resolution, and lower instrumentation cost, all have resulted in MALDI as an attractive and well-utilized platform for MS imaging. Nevertheless, there are still several areas of consideration for lipid imaging by MALDI-MSI and these are discussed in Section 1.5. Over the past decade, as research institutions have started to establish MS imaging facilities, the number of publications on imaging plant lipids has steadily increased. Some of the first MALDI-MS imaging reports described the mapping of free fatty acids within strawberry seeds (Fragaria spp.) and apple (Malus domestica) [33], surface lipids in floral and leaf tissues A. thaliana [34– 36], phosphatidylcholine (PC) species in rice grain (Oryza sativa) [37], and ginsenosides, a class of steroid glycosides, in roots of Panax ginseng [38]. More recently, MALDI-TOF-MS of Capsicum fruits revealed the presence of capsaicin primarily in placental tissue with minor amounts in the pericarp regions and even lower amounts detected in seed tissues [39]. MS imaging of root nodules of Medicago truncatula–Sinorhizobium meliloti during nitrogen fixation [40] demonstrated the identification of several metabolites including flavonoids, organic acids, and carbohydrates that have been implicated in the transfer of carbon, or signals within the nodule-bacteroid symbiotic relationship. Several studies (described in more detail in Section 2) have been reported for visualizing detailed membrane and storage lipid compositions in ‘‘wildtype’’ [41] and transgenic cotton (Gossypium hirsutum) seeds [42], triacylglycerol (TAG) species in wild-type and transgenic Camelina sativa seeds [43], TAG and PC species in engineered tobacco (Nicotiana tabacum) leaves [44], and storage lipids in avocado (Persea americana) mesocarp tissue [45].

1.4. Metabolite image reconstruction Specialized software is often required for processing the complex raw MS imaging spectral data and producing images of

35

in situ metabolite distributions [46]. Advances within these tailored computer applications have enhanced the utility of MS imaging instrumentation for addressing biological questions. In a typical MS imaging experiment, data acquired at each tissue location (Fig. 1B) contains several pieces of information, most importantly the location acquired at selected x- and y- (and sometimes z-) coordinates and the absolute intensities of each ion (at selected m/z’s) detected (Fig. 1C and D). This absolute intensity data can be normalized and colorized to represent the absolute or relative distribution of a selected m/z or set of m/z’s at each x- and y-position. This in turn produces a representation of the metabolite distribution within the tissue (Fig. 1). Depending on the experimental design, both at the sample (e.g. a single metabolite standard, a chemical extract containing several metabolites, or a tissue section) and acquisition levels (e.g. full scan analysis over an m/z range or combination scheme of full and tandem MS (MS/MS) scans [47]), the complexity of acquired raw data (e.g. files often contain gigabytes of data comprised of several million data points), even for a single tissue location, can be daunting for the most experienced mass spectrometrist. While most of the software packages (both commercial and academic) available now enable users to generate two-dimensional images for external interpretation, it is only a matter of time before algorithms will be developed for multi-layered, computational interpretation of complex metabolite distributions. Metabolite Imager is one application that was developed recently [46] to enable customized imaging and analysis of lipids in tissue sections [41–45] as well as algorithms for visualizing and analyzing many other types of metabolites. Within the Metabolite Imager program, there are several features to address processes involved with both data extraction and image reconstruction. Most often users use a targeted search approach; i.e., they know the metabolite(s) to be imaged and some idea of how it behaves in the MALDI- MS instrument. This includes taking into consideration if the metabolite forms one or more ions (e.g. [M + H]+, [M + K]+, [M + Na]+, etc.), whether the metabolite fragments within the source, whether the metabolite abundance needs to be normalized/corrected due to suppression, and finally whether this metabolite will be considered as a single image measured in absolute intensities or imaged within class of related metabolites and represented as a mole fraction of the class. Metabolite Imager interfaces with widely-used lipidomics/metabolomics databases, such as LipidMAPS (38,000 metabolites), KEGG (3350), MetaCyc (10,000), NIST Chemistry WebBook (69,000), and Plant Metabolic Network (1400) databases, to facilitate identification and analysis. One area that may provide important new insights about unknown metabolites and processes, is the co-localization of metabolites and searching for untargeted metabolites through frequency-based and/or spatially-based algorithms. There are several other applications available either through open source/freeware such as BioMap (Novartis, Basel, Switzerland), DataCube Explorer (AMOLF), and Mirion (JLU) or proprietary software from instrument vendors such as Bruker Daltonics (flexImaging), Shimadzu Biotechnology (Intensity Mapping), AB-SCIEX (TissueView) and Thermo Fisher Scientific (Thermo ImageQuest™ [48]). Complementary applications such as the comparative analysis of multiple tissues through multivariate analysis and data reduction [49], quantification of metabolites [50], and extension of 2D–3D imaging [51] provide additional imaging support. One area that continues to be addressed is the move to a consistent, raw data storage format such as imZML [52] to enhance the compatibility of available software on the different commercial instruments. Ultimately this would result in less application redundancy and would help to address limitations in comparative analysis between different instruments. Although there are several specialized software applications developed and used by different

36

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

groups, it is not unreasonable to think, similar to the array of software packages available for gene expression analysis, that each can be tailored toward addressing specific instrument/conditions and different potential biological questions (e.g., metabolites that are specifically associated with certain cell/tissue types or functional processes). 1.5. Considerations and precautions for in situ lipidomics Similar to the development and advances of most methodologies, there are several areas where caution should be exercised when designing and executing MS imaging experiments. While each MS imaging apparatus (i.e. MALDI vs. DESI vs. SIMS vs. etc.) has some specialized areas of concerns, the focus here is on considerations for imaging lipids using MALDI-MSI; several of the concerns can be applied to other approaches as well. Foremost, sample preparation is an important key to accurate localization of metabolites in tissues. Based on our experience, we recommend minimizing the number of steps for handling tissue samples before analysis. Once the tissue sample is prepared, matrix selection and mode of deposition can alter metabolite responses. Although MALDI is considered a soft ionization method, several metabolites will still fragment upon ionization. A major limitation to quantification and accurate metabolite localization is the suppression of certain metabolite classes by ions derived from other classes. While most metabolites can be identified using accurate mass measurements, this is not always the case. Finally, evaluating the reproducibility of tissue-to-tissue and technical variation is an area that needs further consideration. Each of these areas of concern are presented below in brief to enable future users to consider available approaches and possible complications when visualizing and interpreting in situ metabolite distributions. 1.5.1. Sample preparation Sample preparation continues to be a major consideration for integrity of metabolite compartmentation. Several excellent reviews provide a broad range of experiences in optimizing and evaluating sample preparation methods [53–55]. Currently there are a wide range of published protocols for preparing tissue samples for lipid imaging in both plant [35,37,43] and animal [56–58] systems. One drawback in standardizing sample preparation across diverse tissue systems comes from the variation in imaging instrumentation capabilities. Nevertheless, as the number of applications for MSI has continued to expand, particular in the biomedical community, several of the issues with specimen preparation are well described and now better understood. Some of the key variables that need to be carefully considered include chemical fixation, embedding medium, tissue freezing, cryosectioning, tissue preservation, and matrix deposition. Since the goal of MS imaging is to visualize an accurate representation of the metabolites in situ one has to consider the initial preservation steps that generates a snap-shot of the metabolite state. This step might include chemical fixation and/or freezing. Our initial evidence in cotton seeds showed that lightly fixed (4% paraformaldehyde) embryo sections visually showed less structural damage than unfixed tissues [41]. However, follow-up experiments demonstrated that unfixed seeds could be sectioned as well as fixed seeds if given an appropriate equilibration transition time from freezer to cryostat [42]. One concern in fixed tissues is that lipids such as phosphatidylethanolamine (PE) with an amine group could become cross-linked. However, PE species were detected in sections of fixed embryos, albeit at lower levels than expected based on quantities measured in chemical extracts. Even with suppression of PE by PC (and possibly other generated ions) aside, it appears that some lipids susceptible to cross-linking are still available for ionization after chemical fixation of tissues. Still, there is a

possibility that fixation may influence the detection of some metabolites, so preparation of tissues by freezing alone may be preferable in most cases. For most tissues snap freezing tissue in liquid nitrogen can minimize proteolytic damage and preserve structural integrity [59]. Recent reports in mammalian brain sections demonstrate possible degradation of transient molecules such as acetylcholine through standard sample preparation protocols [60]. A variation of in situ freezing method of the brain samples was necessary to improve the signal-to-noise ratio for small labile molecules. For oilseeds that are in a more dormant state and have a protective seed coat we have found little differences in flash freezing seeds using a combination of cold ( 78 °C) isopentane and dry ice versus a slow freeze of embedded seeds in a 80 °C freezer (despite the concerns of ice crystal formation). Depending on the nature of plant tissue and metabolites attempting to be visualized, the form of fixation and freezing must be chosen carefully [2]. Alternative methods need to be continually evaluated as different plant tissues are tested, including rapid freezing, floating tissues in aluminum foil ‘‘boats’’ on liquid nitrogen, slow freezing, and/or the use of cryoprotectants such as 30% sucrose solution. Regardless of the protocol chosen, tissues should ideally remain frozen (< 20 °C) during and post-sectioning before lyophilization and MSI analysis. Unfortunately, many of the chemicals used for embedding in traditional histology are incompatible with MALDI-MSI, usually because they ionize well, suppress native compounds, and sometimes require clean-up steps that utilize organic solvents that would remove lipids from plant tissues [61,62]. Currently, there have been several published reports using different embedding medium including optical cutting temperature (OCT) polymer [41], gelatin [40,43,63], ice [64], agarose [65] and carboxymethylcellulose [66]. Formalin-fixed, paraffin-embedded sections (FFPE) still represent an alternative method that has been used with some success for analyzing proteins; however, it still has not been shown to work well with lipids [59], possibly due to the use of organic solvents for dehydration and clean-up. It is difficult to evaluate which of these embedding mediums is best since all of these studies used different tissue systems, varied in sample preparation procedures, and utilized different MS instruments. The sectioning thickness imparted by a cryostat is an important consideration for maintaining structural integrity and achieving maximal ionization efficiency. For larger molecules such as proteins, a direct comparison of tissue thickness determined that thinner sections (<10 microns) resulted in the greatest peak intensity of observed proteins [67]. However, for sampling lipids in plant tissues, we have found little differences in overall imaging quality despite variations in sectioning thickness between 25 and 50 microns. This flexibility with analyzing lipids could be partially due to their ease of ionization and matrix compatibility. For seed cryosections, our group thaw-mounts frozen sections on traditional glass microscopy slides. In most cases these tissues show good sample integrity in MALDI analysis although some groups recommend indium-tin oxide (ITO)-coated glass slides or stainless steel for particular applications [68]. When thaw-mounting is not desirable or leads to damaged sections, adhesive films offers alternatives to mounting a tissue section [37,69]. Once tissues are sectioned and adhered to the slide or stage, it is important to consider post-tissue sectioning preservation which might include steps such as lyophilization and storage of samples within a desiccator [70]. 1.5.2. Matrix selection and deposition For imaging approaches that require application of a chemical matrix to promote ionization, the selection of the chemical matrix and the mechanism by which it is deposited are key considerations for achieving high spatial resolution and metabolite detection in situ [2,29]. There are several matrix deposition methods

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

available such as sublimation and variations of spray coating reviewed in [29] that can result in different metabolite responses. The ultimate goal is a homogenous application such that the endogenous metabolite response is not altered due to differences in matrix presence. Since some lipids, and other metabolites, have differential ionization responses with selected matrices, it is important to select the best matrix for visualizing the targeted set of compounds. There are several commercially-available matrix compounds to choose from, usually derivatives of organic acids with low ionization potentials [71]. 2,5-Dihydroxybenzoic acid (DHB) has been the matrix predominantly used for lipid studies, and the one most often used for imaging storage and membrane lipids in oilseeds [72]. 9-Aminoacridine (9-AA) is also a popular matrix suggested for analysis of phospholipids, since uncharged lipids such as cholesterol or triacylglycerols are not ionized appreciably in this matrix and are not detected [73]. 9-AA is most suitable for analysis in negative-ion mode, although it produces a PC byproduct –CH3⁄ that can make distinguishing some PE and PC species more difficult [74]. Other matrices such as 2,4,6-trihydroxyacetophenone (THAP) [75], 2-(2-aminoethyloamino)-5-nitropyridine (AAN) [76], and 6-aza-2-thiothymine (ATT) [75] have been shown to be comaptible for imaging some lipid molecules. Additional imaging matrices continue to be empirically tested including both newly-synthesized chemicals and well-established compounds used for analyzing other metabolite classes [77]. 1.5.3. Metabolite (in-source) fragmentation Although many groups have extensively studied the generation of ions in MALDI-MS, the process is still somewhat uncharacterized due to the nature of several (likely) mechanisms for ion formation [71]. One of the technical challenges for analyzing lipids by MALDI (and other imaging methodologies such as SIMS) is the possible formation of in-source fragment ions, particularly through postsource decay mechanisms when the ions are directed into the MS [71,78]. There are several factors contributing to fragmentation of lipids including choice and deposition of matrix, internal analyte energies and proton affinities, regions of high gas pressure within the MALDI plume, and laser fluences [71,79]. For example, we have observed abundant diacylglycerol (DAG) fragment ions directly from both TAG and PC in our analyses of oilseed tissues [41]. One possible solution for minimizing the fragmentation of TAGs in MALDI is to use nitrocellulose film substrate [80]. While some groups have used in-source fragmentation to their advantage for identification purposes [39] this could lead to complications to interpretations of distribution patterns and ultimate quantification for selected molecules. 1.5.4. Ion suppression Ion suppression of certain lipid classes is one major limitation of MALDI-MS that, in part, prevents the direct quantification of lipid species in tissues and samples. Several groups have shown differences in relative detection of phospholipids and neutral lipids in most conditions [32,41,81]. The presence of certain widely-used embedding materials such as optimal cutting temperature (OCT) also has been shown to suppress lipid detection and should be avoided whenever possible [28,82]. Overcoming metabolite suppression might be possible through the implementation of alternative matrices or possibly matrix-free LDI-MS in some instances [83,84]. Applying correction algorithms using complementary data from additional instruments and/or internal standards can be used to calculate corrected intensity values that are likely more representative of endogenous concentrations [85,86]. We have found that in MS imaging, the relative quantities of acyl lipids within structurallydefined classes such as TAG and PC agree well with compositions quantified through a variety of validation measurements including shotgun lipidomics and MALDI-MS of tissue extracts, as well as

37

GC–MS analysis of tissue-specific extracts [41–43]. Therefore, depending on the degree of quantification desired there are several reliable ways to address ion suppression in MS imaging. 1.5.5. Accurate mass measurements Advances in high-resolution mass analyzers (e.g. Orbitrap, FTICR) capable of accurate mass designations (<5 parts per million, PPM) have enabled the high probability identification of many metabolites by comparing their theoretical and experimentally determined masses [2]. This is especially true for many lipid molecules that tend to fall in the m/z range outside of high background noise produced from the matrices and embedding mediums. However, false positives are still possible even at <1 PPM and additional confirmation of metabolite identification should be utilized whenever possible [87]. Even though tandem MS from tissue sections provides some challenges due to limited number of laser shots per spot and sometimes altered fragmentation patterns relative to ESI-MS/ MS and GC–MS, the positive identification of a metabolite is substantially improved with MS/MS evidence. Additional validation of the relative quantities of certain lipid molecules in MALDI-MS images can be realized by comparing the identification and quantification of targeted metabolites in tissue extracts by ESI-MS/MS [41,42]. 1.5.6. Reproducibility and validation Due to technical challenges with sample preparation and a number of different imaging instruments with specialized capabilities being utilized, it is frequently difficult to assess reproducibility across (and for that matter within) labs for tissue replicates. This has become even more important for plant lipids due to the presence of lipid heterogeneity within several different types of tissues including both vegetative [44] and embryonic (see Section 2) tissues. Our group attempts to address this concern by evaluating and reporting both biological (e.g. representative sections within independent tissues) and technical (e.g. serial sections within same tissue) variation when feasible [41–43]. In order to improve reproducibility and validate results across labs, it will be important in the near future to introduce guidelines for producing high-quality, structured imaging data that can be deposited to the MS imaging community for validation and additional analysis. There are a few of these components in places such as open-source, raw storage data formats (e.g. imZML [52]), standard practices for reporting mass spectrometry data [88,89], publically-available metabolomics databases (see Section 1.4), and a growing community of dedicated MS imaging researchers. There is still much room for improvement including establishing standard operating protocols MS imaging data acquisitions, algorithms for annotating metabolites and evaluating reproducibility, and data deposition and access. 2. Learning lessons about oilseed metabolism from metabolite location Successfully managing the potential pitfalls of MALDI-MSI can lead to an impressive new perspective in cellular metabolism. Recently the distribution patterns of many endogenous lipid metabolites have been mapped by MALDI-MSI in tissue sections of several oilseeds including cotton, L. G. hirsutum [41,42] (Fig. 3), C. sativa [43] (Figs. 4 and 5) and A. thaliana (Fig. 6). Immediately apparent in all seeds examined was the marked heterogeneity in molecular species distribution within the embryos. This unexpected heterogeneity was evident in cross-sections and longitudinal sections of embryos, and when mol percentages were averaged over the entire sections, the relative quantification of TAG species matched well with quantitative analyses of total lipid extracts from seeds. Each of these different oilseed species provides a different framework for investigating lipid metabolism in a spatial

38

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

A

F TAG 52:3 (PLO) m/z 879.742

ea MAX 35 %

B TAG 52:4 (PLL) m/z 877.726

Y co

X TAG 54:6 (LLL) m/z 901.726

G TAG Mol %

C

D TAG 53:3-Cyc (LScP) m/z 891.742

X-Axis (mm)

E Cotyledons

Axis Cotyl......edons

H Distribution Pattern

Embryonic Axis

1

L, Ln, No ID

2

P, O

3

S, A, Po

4

Sc, Dsc

Max Pattern Min

Fig. 3. Heterogeneity of TAG species in cotton embryos. (A–D) Distribution of selected TAG molecules species in cross-sections normalized to relative TAG content. Species are either normalized on a colored scale at 35 mol%, (A–C) or on a grey scale at 3 mol%. (D) Species are denoted with their total acyl carbons and double bonds (e.g. TAG 52:3 with 52C and 3 double bonds), individual major acyl chains with common abbreviated shorthand (sn-non-specific), and m/z for one representative major peak. (E) Reproducibility of major TAG species within independent wild-type cottonseeds (n = 5). Black bars = embryonic axis, grey bars = cotyledons. p-Values, ⁄p < 0.10, ⁄⁄p < 0.05. (F– G) Alternative mode for visualizing TAG distributions through cottonseed. Thin lines represent the individual mol% values at that x, y position for a particular species (purple, TAG-52:4 PLL; red, TAG-54:6 LLL; green, TAG-50:2 PPL; yellow, TAG-52:3 PLO; light blue, TAG-54:5 LLO; dark blue, combined sterculic/dihydrosterculic cyclic TAGs). Thick lines represent a polynomial fit using Microsoft Excel to aid visualization of changes across the section. (H) Summary diagram of distribution patterns based on selected acyl chains. Fatty acid abbreviations: P, palmitic (16:0); Po, palmitoleic (16:1); S, stearic (18:0); O, oleic (18:1); L, linoleic (18:2); Ln, linolenic (18:3); Sc, sterculic (19:1-Cyc); Dsc, dihydrosterculic (19:0-Cyc). Figure modified from Horn et al. (2012) Spatial mapping of lipids at cellular resolution in embryos of cotton. The Plant Cell. 24: 622–636 with permission (www.plantcell.org, Copyright American Society of Plant Biologists) and Horn et al. (2013) Modified oleic cottonseeds show altered content, composition and tissue-specific distribution of triacylglycerol molecular species. Biochimie 2014; 96: 28–36.

context (e.g. different molecular compositions, anatomical structures, different levels of genetic and metabolic manipulation/characterization, etc.). Each, in turn, has revealed interesting biological phenomena that suggest insights to unresolved questions about plant lipid metabolism and unveil new questions and potential approaches for enhancing our understanding of TAG biosynthesis in oilseeds. Several important topics for the production of TAG species in oilseeds from a localization perspective are addressed below including: (1) TAG and PC molecular species distributions can help support/predict tissue-specific distributions for enzymes involved in TAG biosynthesis; (2) NMR approaches for visualizing seed oil content complement and expand imaging by MS; and (3) using imaging information helps refine and enhance metabolic engineering strategies. The pathways for the synthesis of TAGs in oilseeds have received a great deal of attention over the past 20 years as increases in yield or changes in composition could improve value substantially [90]. Several recent reviews describe in detail the biosynthetic components and lingering questions regarding TAG accumulation in oilseeds [91–94]. Below we attempt to relate the distribution of TAG

metabolites in oilseeds to factors or pathways that might be inferred from differences in localization. While we interpret similar biosynthetic motifs to oilseeds that are genetically and anatomically different, the inclusion of additional oilseed species to MS imaging datasets will help elucidate to what extent these pathways overlap in a spatial context for plant embryos in general. The acyl chains synthesized for incorporation into TAG molecules originate from fatty acids (FA) synthesized in the plastids of plant cells [95] (Fig. 7). FA are assembled as C16:0-ACP (acyl carrier protein) or C18:0-ACP by FA synthase (FAS) [96–98]. C18:0-ACP is further desaturated to C18:1-ACP. In oilseeds, a majority of these acyl chains are destined for TAG biosynthesis [91]. ACP-bound FA are presumably first released as free FA by acyl-ACP thioesterases (FATA and FATB) in the plastid, and then activated as acyl-CoA molecules before export to ER as a part of the ‘‘eukaryotic’’ pathway [92,98]. Some evidence suggests PC may be an intermediate in the export of acyl groups to ER [99]. Many of the acyl lipid species imaged by MALDI-MSI are a direct reflection of the FA supplied from the plastid. The acyl-CoA pool provides a dynamic supply of FA for potential incorporation into TAG molecules. Several enzymes are known to be

39

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

B Triacylglycerols

A ea co

56:7 LnLnG

56:6 LnLG

56:5 LLG

56:4 LOG

54:7 LLLn

54:6 LLL

54:5 LLO

52:4 PLL

MAX = 15 mol % TAG

MIN ~ 0% 54:8 LnLnL

C

Phosphatidylcholines

MAX = 30 mol % PC

34:3 PLn

34:2 PL

34:1 PO

36:6 LnLn

MIN ~ 0%

36:5 LnL

36:4 LL

36:3 LO

36:2 OO

38:4 GLn

Fig. 4. Distribution of triacylglycerols (TAGs) and phosphatidylcholines (PCs) in wild-type Camelina sativa seeds. (A) Bright field image of wild-type Camelina cross section. Abbreviations: ea = embryonic axis, co = cotyledons. Scale bar = 500 lm. (B) Distribution of major TAG species in Camelina wild-type seed cross section. Images are normalized to a max of 15 mol% of all TAG species detected. Each species is noted by the number of total acyl carbons, double bonds, and acyl chain species. Fatty acid abbreviations: P, palmitic (16:0); O, oleic (18:1); L, linoleic (18:2); Ln, linolenic (18:3); G, gadoleic (20:1). (C) Distribution of major PC species in Camelina wild-type seed cross section. Images are normalized to a max of 30 mol% of all PC species detected. Each species is noted by the number of total acyl carbons, double bonds, and acyl chain species. Figure modified from Horn et al. (2013). Imaging heterogeneity of membrane and storage lipids in transgenic Camelina sativa seeds with altered fatty acid profiles. The Plant Journal 76: 138– 150 with permission.

involved in acyl chain modifications. The FA synthesized in the plastid, 16:0-, 18:0-, and 18:1-CoA often make up a major portion of the available CoA pool. The membrane-bound complex FAE1 can produce long chain FA (LCFA) such as 20:0- and 20:1-CoA from 18:0and 18:1-CoA, respectively [100,101]. In most plant tissues, ERbound fatty acid desaturases, FAD2 and FAD3, are involved in desaturating 18:1 FA and 18:2 FA to 18:2 and 18:3, respectively, while esterified to PC [102–104]. These modifications are especially important to maintain membrane fluidity [105,106], but also provide polyunsaturated fatty acids for incorporation into TAG [107]. In seeds of some plants that accumulate ‘‘unusual’’ fatty acids, acyl groups are modified with other functional groups (e.g., hydroxy, epoxy, cyclic groups) while esterified to PC. For example, in the Malvaceae family including Steruclia foetida and G. hirsutum (upland cotton), 18:1 FA esterified to PC is available for a cyclization reaction to 19:0-FA (dihydrosterculic acid) and further desaturation to 19:1-FA (sterculic acid) [108–110] by a cyclopropane fatty acid synthase (CPA-FAS) and desaturase (CPA-FAD), respectively (Fig. 7). Whether common or unusual, the acyl groups on PC have several potential fates in the ER in terms of TAG incorporation [92]. They may re-enter the acyl CoA pool via ‘‘acyl editing’’ where they are available for incorporation into glycerolipids via the acyl CoA-dependent acyltransferases of the well-studied Kennedy pathway. PC may be converted back to DAG by PC: DAG cholinephosphotransferase (PDCT; product of the ROD1 gene, [111]). Or, acyl groups on PC may be transferred directly into TAG by an acyl CoA-independent pathway via the phospholipid: DAG acyltransferase (PDAT) [112].

2.1. Does heterogeneity of TAG metabolites suggest differences in localization of enzymes? Yes, no and maybe Visualizing the composition of TAG molecular species in mature cotton embryos (Fig. 3) revealed heterogeneous distribution patterns that were particularly evident between the embryonic axis and cotyledon tissues [41,42]. Most TAG species (Fig. 3A–D) imaged in embryos generally exhibited one of four distribution patterns: (1) cotyledon enrichment (e.g. TAG-54:6, 18:2/18:2/18:2, Fig. 3C and TAG-52:4, 16:0/18:2/18:2, Fig. 3B), (2) embryonic axis enrichment (e.g. TAG-52:3, 16:0/18:2/18:1, Fig. 3A), (3) relative uniformity between both tissues, and (4) exclusive localization in embryonic axis (e.g. TAG-53:3, 18:2/19:1-Cyc/16:0, Fig. 3D). Remarkably, despite this spatial heterogeneity, the overall molecular percentage of most species when summed over an entire seed section agreed well with ESI-MS analysis of total seed lipid extracts demonstrating the value of acquiring the overall compositional and spatial information for TAG molecular species. Patterns for specific TAG species were associated with the enrichment of specific acyl chains (Fig. 3F–H). TAGs containing the polyunsaturated acyl chains linoleic (18:2) and linolenic acid (18:3) were enriched in the cotyledons (Fig. 3H, Pattern 1), in contrast to palmitate (16:0) and oleic acids (18:1) that were enriched in the embryonic axis (Fig. 3H, Pattern 2). Stearic (18:0), arachidic (20:0), and palmitoleic (16:1) acids were localized relatively uniformly throughout the embryo (Fig. 3H, Pattern 3) while cyclic fatty acids, sterculic (19:1-Cyc) and dihydrosterculic (19:0-Cyc), were

40

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

A High Palmitate MAX = 40% PC

B

MAX = 35% TAG

PC-34:2 (PL)

TAG-50:3 (PLnP)

TAG-50:2 (PLP)

PC-32:0 (PP)

TAG-52:2 (POO)

TAG-48:0 (PPP)

TAG-50:1 (POP)

MAX = 15 mol % PC or TAG

MIN ~ 0%

High Oleate MAX = 45 %TAG

C

PC-36:2 (OO)

TAG-54:3 (OOO)

PC-36:3 (LO)

TAG-52:4 (PLnO)

MAX = 18 mol % PC 20 mol % TAG

MIN ~ 0%

High Linoleate

PC-34:2 (PL)

TAG-52:4 (PLL)

MAX = 35 mol % PC 20 mol % TAG

MIN ~ 0%

PC-36:4 (LL)

TAG-54:6 (LLL)

Fig. 5. Distribution of selected TAG and PC species in Camelina with altered seed oil compositions. (A) Selected PC and TAG species from high palmitate seeds (overexpression of CpuFATB1). Scale bar = 500 lm. All Images are normalized within their respective class according to color legend unless otherwise noted. Each MS image represents the distribution of TAG or PC species denoted with the number of total acyl carbons, double bonds, and acyl chain species. Fatty acid abbreviations: P, palmitic (16:0); O, oleic (18:1); L, linoleic (18:2); Ln, linolenic (18:3). (B) Selected PC and TAG species from high oleate seeds (FAD2/FAE1 RNAi). Scale bar = 500 lm. (C) Selected PC and TAG species from high linoleate seeds (FAD3/FAE1 RNAi). Scale bar = 500 lm. Figure modified from Horn et al. (2013). Imaging heterogeneity of membrane and storage lipids in transgenic Camelina sativa seeds with altered fatty acid profiles. The Plant Journal 76: 138–150 with permission.

exclusively localized in the embryonic axis (Fig. 3H, Pattern 4). In all cases, these heterogeneous patterns strongly supported a precursor-product relationship between acyl groups in PC molecular species and TAG molecular species, pointing to regional differences for TAG biosynthetic pathways. The relative localization of individual fatty acids that are either a direct product of de novo fatty acid synthesis (16:0, 18:0, 18:1) or through subsequent modification (16:1, 18:2, 18:3, 20:0, 19:1-Cyc, 19:0-Cyc) suggests that one possible explanation for the tissue-specific heterogeneity in metabolites is through differences in the enzyme activities responsible for producing these fatty acids. 2.1.1. Cyclopropane fatty acid (CPFA) synthase/desaturase A likely explanation for the almost exclusive presence of cyclicFA-containing TAGs in the embryonic axis of mature cotton embryos is a tissue-specific localization of the enzymes responsible for the production of cyclic FAs. Based on previous reports of cyclic fatty acid synthesis [108], dihydrosterculic acid (19:0-Cyc) is

produced first through a methyl group addition via S-adenosyl-LMet to the delta-9,10 position of oleate (18:1) esterified at the sn-1 position of a PC precursor molecule by CPA-FAS enzyme (Fig. 7). Sterculic acid (19:1-Cyc) is then synthesized by desaturation of the 19:0-Cyc acyl moiety through a less characterized CPA-FAD. The enrichment of cyclic fatty acids in cotton seeds and seedling roots [110,113] is consistent with MS imaging results localizing TAGs with cyclic fatty acids to the embryonic axis of mature cotton embryos (Fig. 3D) [41]. The CPFA reactions occur on 18:1 acyl chains of PC, and the putative PC precursors (for TAGs) with cyclic FAs are localized almost exclusively in the embryonic axis of cottonseeds as well. Therefore MS imaging data suggests that CPFA synthase and desaturase should be localized in the embryonic axis and likely absent from cotyledon tissues. Two isoforms of CPFA synthase (GhCPS1 and 2) were identified and reported to contribute to cyclic fatty acid formation in G. hirsutum seeds and roots [110]. PC species containing 18:1 fatty acids, did not appear to be limiting in cotyledons (as a substrate for

41

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

A

B Triacylglycerols

“In Silique”

50:2 PLP

52:2 PLS

52:3 PLO

52:4 PLL

52:5 PLLn

54:2 POG

54:3 PLG

54:4 PLnG

54:5 LLO

54:6 OLLn

56:3 OGO

56:4 LAL

56:5 LGL

56:6 GLLn

56:7 LnGLn

58:3 GOG

58:4 GGL

58:5 GGLn

58:6 LnLB

ea 54:7 LLnL

co

MAX = 15 mol % TAG

500 µm

MIN ~ 0%

C Phosphatidylcholines

34:0 PS

34:1 PO

34:2 PL

34:3 PLn

36:0 SS

36:1 SO

36:2 OO

36:3 OL

MAX = 20 mol % PC

36:4 LL

36:5 LLn

38:3 GL

38:4 GLn

MIN ~ 0%

Fig. 6. Distribution of triacylglycerols (TAGs) and phosphatidylcholines (PCs) in single Arabidopsis thaliana seeds. (A) WT (Col) Arabidopsis seeds were prepared and cryosectioned within mature siliques. The arrow denotes the single seed imaged in parts B and C. ea, embryonic axis, co, cotyledons. (B) Distribution of major TAG species in a single Arabidopsis seed. Images are normalized to a max of 15 mol% of all TAG species detected. Each species is noted by the number of total acyl carbons, double bonds, and acyl chain species. Acyl chain species are designated based on data from The Arabidopsis Book [92]. Fatty acid abbreviations: P, palmitic (16:0); S, stearic (18:0); O, oleic (18:1); L, linoleic (18:2); Ln, linolenic (18:3); G, gadoleic (20:1); B, behenic (22:0). (C) Distribution of major PC species in a single Arabidopsis seed. Images are normalized to a max of 20 mol% of all PC species detected. Histogram compares the molecular percentages calculated from the single seed in parts B and C to HPLC-MS analysis of dry seeds extracted and reformatted from The Arabidopsis Book [93].

CPA-FAS), so it is reasonable to speculate, instead, that expression of these GhCPS isoforms is restricted to embryonic axis tissues. Indeed, Jiao et al. [114] showed that transcript abundances for GhCPS1 and 2 were more than 260 times higher in embryonic axis tissues compared to cotyledons of cotton embryos. So in the case of cyclic fatty acids in cotton embryos, heterogeneous expression of enzymes explains the heterogeneous distribution of PC and TAG metabolites. It is important to note that without the tissue-based heterogeneity observed by MALDI-MS for glycerolipid species with cyclic fatty acids, this question of isoform localization in embryos would not have seemed worth consideration. 2.1.2. Delta-12 fatty acid desaturase (FAD2) The differential distributions of 18:1-containing and 18:2-containing TAGs in the embryonic axis and cotyledons, respectively, of mature cotton embryos (e.g. Fig. 3A–C, E) suggests that there may be spatial differences in FAD2 activity (Fig. 7) [41]. In oilseeds, microsomal FAD2 catalyzes the formation of a methylene

interrupted double bond arrangement (i.e. delta-9,12 positions) by desaturating a 18:1 acyl chain esterified at the sn-2 position of PC [115]. Since these FA differences appeared to be sharply defined at the border between the two functionally different seed tissues, it is possible that this TAG phenotype is a result of altered expression levels for the major seed-specific FAD2 isoform, GhFAD2-1 [116,117]. However, there was no obvious difference between FAD2-1 transcript abundance in the cotyledons relative to the embryonic axis by RT-PCR [42,114]. Oddly, a FAD2-2 isoform showed higher relative expression in the embryonic axis compared with cotyledons [114], opposite to the location of 18:2 enrichment. While it cannot be ruled out that other FAD2 isoforms are differentially expressed in embryo tissues (four isoforms have been reported in cotton [116]), extensive transcriptional profiling studies using RNA-Seq did not suggest it [114]. Taken together, these results suggest that processes other than FAD2 expression might be a significant factor in producing this heterogeneity in 18:2-containing TAG species. For example, the FAD2 enzymatic reactions

42

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

Plasd

ER Acyl 18:1

FAS

FAD2

PC

Acyl 18:2

19:1-Cyc CPA FAD Acyl

19:0-Cyc CPA FAS Acyl

18:1

18:3 PC

PC

PC

PC

Acyl

FAD3

PC

Acyl

FATB

16:0-ACP

LACS

KASII

18:0-ACP

FAE1

Acyl-CoA Pool FATB

18:2 16:0

18:0

18:3

19:1-Cyc

19:0-Cyc

18:1

18:0-CoA 18:1-CoA

20:0 20:1

Acyl-CoA

SAD FATA

Acyl-CoA

18:1-ACP G3P

GPAT

Acyl-CoA

LPA

LPAAT

PA

PLA

Acyl-CoA

PAP/PAH

DAG

PDCT

PC

DGAT

TAG

LPC

PC

Acyl

PDAT

LPC

Acyl-CoA

LPCAT

ER Fig. 7. Simplified scheme for acyl modification and TAG biosynthesis in a hypothetical oilseed. Figure modified from recent reviews [90,91]. The following abbreviations appear in the diagram: TAG, triacylglycerol; PC, phosphatidylcholine; FAS, fatty acid synthase; FATA, acyl-ACP thioesterase A; FATB, acyl-ACP thioesterase B; KASII, ketoacylACP Synthase II; SAD, stearoyl-ACP desaturase; LACS, long-chain acyl-CoA synthetase; DAG, diacylglycerol; DGAT, DAG acyltransferase; FAD2, fatty acid desaturase 2; FAD3, fatty acid desaturase 3; FAE1, fatty acid elongase 1; G3P, glycerol-3-phosphate; GPAT, G3P acyltransferase; LPA, lysophosphatidic acid; LPAAT, LPA acyltransferase; LPC, lysoPC; LPCAT, LPC acyltransferase; PA, phosphatidic acid; PC, phosphatidylcholine; PDAT, PC:DAG acyltransferase; PDCT, PC:DAG phosphocholinetransferase; PLA, phospholipase A; PAP, PA phosphatase; PAH, PA hydrolase; CPA-FAD, cyclopropane fatty acid desaturase; CPA-FAS, cyclopropane fatty acid synthase; ER: endoplasmic reticulum.

(along with most other desaturases) are dependent on oxygen- and an electron-donor for desaturation [118,119]. Due to the anatomical structure of the cotton embryo during development, it is possible there is less available oxygen within the internal embryonic axis tissues compared with outer cotyledonary tissues, consequently leading to reduced FAD2 activity [120]. In any case, the evidence would suggest that the reproducible enrichment of 18:2 containing TAG in cotyledon tissues of cotton embryos is not a result of differential localization of FAD2 expression, and requires further investigation. Several biotechnology strategies have been used to target FAD2 for suppression both in cotton and other oilseed crops to produce high value, high oleic varieties [121,122]. In order to gain additional perspective into the role and localization of FAD2 for generating oleic/linoleic acyl distributions, high oleic cotton seeds were imaged where FAD2 was suppressed through a protein-based strategy. These high oleic cotton seeds were generated through a dominant negative mutation expressing a non-functional allele FAD2 from Brassica napus (Bnfad2) [123,124]. Bnfad2 seeds with the most severe, low seed oil phenotypes showed altered TAG profiles compared with wild type embryos (cv Coker 312) both in overall oleic vs. linoleic ratios in the embryonic axis vs. cotyledons, and in distributions of individual TAG species [42]. In general, the linoleic-rich TAG species (e.g. TAG-54:6 or TAG-18:2/18:2/18:2) enriched in the cotyledons in wild-type embryos were found at much lower levels in Bnfad2 cotyledons. Some oleic-rich TAG species (e.g. TAG-52:2 or TAG-16:0/18:1/18:1) that were minor in wild type cotyledons were much more prevalent in Bnfad2 transgenics. New patterns of TAG distributions were observed in the Bnfad2 embryos as might be expected. One interesting observation was that TAG 16:0/18:1/18:2 was more prevalent in cotyledons than in the embryonic axis, whereas TAG 16:0/18:2/18:2 (more linoleic) showed the opposite distribution (more in axis than in cotyledons). In other words these embryos appeared not to have reduced 18:1 desaturation in the axis relative to cotyledons, which may suggest that oxygen gradients are not an explanation for 18:1/ 18:2 heterogeneity. The relatively uniform expression of the

b-phaseolin-driven Bnfad2 and endogenous GhFAD2-1 in both embryonic axis and cotyledon tissues of mature Bnfad2 seeds, suggested that these altered distributions were not due to transcriptional mechanisms. But since there is likely a mix of active and inactive FAD2 enzymes in these embryos, and Bnfad2 embryos are compromised in total oil content, this interpretation should be made with some caution. Perhaps examining other high oleic lines generated through transcriptional suppression of FAD2 [125] without compromise in overall TAG accumulation, might help further explain how these apparent FAD2-independent metabolite distributions are generated. So while MS imaging confirmed evidence for changes in levels of oleic acid via expression the non-functional Bnfad2 allele, these changes did not appear to be attributed to differences in localization of FAD2 expression. Therefore, it appears that in the case of FAD2 and 18:1/18:2 PC and TAG, heterogeneity of metabolites cannot be explained by heterogeneity of gene expression. While one might suggest that this inconsistency could be a result of variation in images of metabolites obtained by MALDI-MSI, this does not seem to be the case. Extensive validation studies of mol percentages obtained from summing results from MALDI-MS images over multiple seed sections (and multiple species), compared with quantification of mol percentages in lipid extracts from embryos determined by ESI-MS/MS or by MALDI-MS (same instrument as for imaging) were very similar to one another [41–43]. Further, sections obtained from several different embryos showed reproducibly similar patterns of metabolites in situ [42]. This consistency with other quantitative methods, and the reproducibility from embryo to embryo indicates that these glycerolipid metabolites are accurately accounted for in MS imaging. Consequently, further examination is required to understand what other processes are involved in regulating spatial FAD2 activity and 18:2 synthesis for TAG, a question that would not have been revealed without imaging lipid metabolites in situ. Many oilseeds produce FA compositions with a high percentage of polyunsaturated FA (PUFAs) including A. thaliana and C. sativa (40–60%). While Arabidopsis and Camelina species are more similar

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

to each other (Brassicaceae family) than to cotton (Malvaceae family), all three species offer opportunities to evaluate similar components for TAG biosynthesis. In wild-type Camelina (Fig. 4) and Arabidopsis (Col-0) (Fig. 6), TAG composition and metabolites are considerably more complex than in cotton embryos. In Camelina and Arabidopsis, 18:2-rich TAG species (i.e. TAG-52:4, TAG-54:5, TAG-54:6) are generally enriched in the embryonic axis which is in contrast to cotton embryos. In both Arabidopsis and Camelina, the situation may be complicated by ultimate competition for 18:1-FA to produce either 18:2 or very long chain FAs (VLCFAs) such as 20:1 which also appear to be distributed heterogeneously. Because there might be different spatial regulation of FAD2 activities in different oilseeds (cotton, Camelina and Arabidopsis) MS imaging should be helpful as a tool for comparing the localization of metabolites and enzymes in TAG biosynthesis to better understand the consequences of compartmentalization of oilseed lipid metabolism. Generalizations cannot be made across even these three species without more information. 2.1.3. Delta-15 fatty acid desaturase (FAD3) Wild-type Camelina seeds have a different overall fatty acid composition (Fig. 4) relative to cottonseeds primarily due to the activities of FAE1 and FAD3 that produce 20:1 and 18:3 species, respectively, in Camelina (Fig. 7). Based on the PC (Fig. 4B) and TAG (Fig. 4C) composition of WT Camelina seeds, it is not entirely clear as to the spatial contribution of FAD3 activities in these tissues [43]. In WT longitudinal sections, 18:3-rich PC species were preferentially localized in the cotyledon tissues [43]. However, TAG molecular species that contained one or more 18:3 acyl chains were enriched in embryonic axis tissues (unless combined with 20:1). These differences could be a result of PC species destined for TAG biosynthesis versus PC species required as membrane lipids in different tissues. Understanding the significance of these differences in 18:3 distributions may need to take into account differences in the distribution of 20:1-containing species and/or possible differences in acyl-editing. For example, 18:1 is a precursor FA for formation of 18:3 and 20:1 by FAD3 (by way of FAD2) and FAE1, respectively. Since 20:1-containing TAG species are enriched in the cotyledons, a case could be made that the lack of indirect competition from FAE1 activity in the embryonic axis is responsible for an elevated proportional 18:3-containing TAG species in the embryonic axis. Similar heterogeneous TAG distributions were found in single Arabidopsis seeds (Fig. 6) for 18:3-rich species. FAE1 or FAD3 knockout or knockdown mutants might help unravel the relationships between metabolites and enzyme localization. It is of interest to point out that while there are only minor quantities of 18:3 FA in cotton embryos, 18:3 was significantly concentrated within the embryonic axis ([41; Supplemental Fig. 2]), and RNA-seq analysis indicated that GhFAD3 was preferentially expressed in embryonic axis over cotyledonary tissues in cotton embryos [114]. Hence, in cotton embryos at least, there is a definitive co-localization of 18:3 metabolites with FAD3 gene expression. 2.1.4. Fatty acid elongase 1 (FAE1) The presence of FAE1 which elongates 18:1-CoA to 20:1-CoA in Camelina and Arabidopsis seed tissues, results in an interesting TAG distribution profile. In both Camelina (Fig. 4) and Arabidopsis (Fig. 6), 20:1-containing PC and TAG species were enriched in the cotyledon tissues. Based on MS imaging one might infer that FAE1 activity is higher in cotyledons as a result of differences in tissue-specific expression and/or post-transcriptional regulation. However, in previous studies on Arabidopsis embryos both through in situ RNA hybridization and histochemical staining of FAE1 promoter-GUS lines [101], FAE1 was expressed throughout the embryo with the exception of the radicle tip. In addition, MS

43

imaging of transgenic Camelina seeds where 20:1 content was significantly reduced (>6) but still detectable, including high palmitic lines (Fig. 5A, overexpression of FATB reducing 18:1), high oleic lines (Fig. 5B, RNAi-mediated suppression of both FAD2 and FAE1), and high linoleic lines (Fig. 5C, RNAi-mediated suppression of both FAD3 and FAE1), these embryos still showed slightly higher levels of 20:1 TAG containing species in cotyledons compared with the embryonic axis [43]. While similar experiments have not been conducted on Arabidopsis seeds, it would appear that differential expression of FAE1 may not fully explain these differences in lipid metabolite distribution. It is possible that these differences in spatial distribution of 20:1-containing glycerolipid species, resulted from additional factors such as tissue specific differences in acyltransferase activities. 2.1.5. Tissue-specific acyltransferase activities? In all oilseeds imaged thus far, there has been a marked regional heterogeneity in the distribution of PC and TAG molecular species, in particular between the embryonic axis and cotyledon tissues. While there is evidence for major differences in localization of acyl chain modifications between the embryonic axis and cotyledon tissues (above), there also may be spatially distinct mechanisms for shuttling acyl chains through different pathways (Fig. 7) to generate the TAG species [43]. For example, in Camelina tissues the TAG species TAG-56:5 and TAG-54:6 differ only by one FA, 20:1 and 18:2, respectively, and have TAG distribution patterns that almost mirror each other in terms of cotyledon versus embryonic axis enrichment (Fig. 4B). The distinct TAG distributions in WT Camelina and the distribution/availability of the PC-36:4 precursor ((Fig. 4C) mostly 18:2/18:2) could suggest that there are tissuespecific differences in this acylation step either through an acyl CoA-dependent (DGAT, diacylglycerol acyltransferase) or an acyl CoA-independent (PDAT, PC: diacylglycerol acyltransferase) pathway (Fig. 7). In Camelina cotyledons, the prevalence of 20:1 species on TAG molecules, the preference for DGAT selecting 20:1-CoA molecules (over 18:2-CoA, [126]), and the low abundance of PC species esterified with 20:1, all suggest that TAG-56:5 would be synthesized primarily through an acyl CoA-dependent, DGAT-mediated pathway. The biochemical redundancy of PDAT in some oilseeds such as Arabidopsis [127] means there is precedence for a role of acyl CoA-independent pathway in TAG biosynthesis in seeds. On the other hand, there is a relative lack of tri18:2-TAG in cotyledons, despite the availability of PC 18:2/18:2 in Camelina cotyledons, suggesting that PDAT may play a minor role in synthesis of major TAG species in cotyledons at least relative to DGAT. MS imaging of genetically altered Camelina seeds with combinations of PDAT and/or DGAT knocked down/out would help further address their relative contributions in the synthesis of TAG in cotyledons. In contrast to TAG 56:5 in cotyledons, the production of 18:2rich TAG-54:6 in the Camelina embryonic axis might be predicted to be mediated in large part by a PDAT-mediated pathway. This hypothesis is based on several lines of evidence, some of which relate to the differential localization of PC and TAG metabolites. First, TAG-18:2/18:2/18:2 was enriched in the embryonic axis relative to cotyledonary tissues (Fig. 4B). Second, PDAT enzymes prefer 18:2 acyl chains (over 20:1 or other acyl groups) on PC [128,129]. Third, the two most abundant 18:2-containing PC species were localized in the axis as potential substrates for PDAT; PC-36:4 was equally prevalent in cotyledons and axis, and PC-34:2 was enriched in the axis relative to cotyledon tissues (Fig. 4C). An alternative hypothesis that cannot be entirely ruled out is that 18:2 could be incorporated into TAG from the acyl-CoA pool via a DGAT. Both hypotheses rely on PDCT to supply di-18:2 DAG for acylation. The latter hypothesis would rely on extensive acyl editing mechanisms to populate an 18:2-CoA pool for DGAT. These two alternate hypotheses could be further resolved by analyzing the tissue

44

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

specific expression of DGAT, PDAT, PDCT and LPCAT (or PLAs) in embryos, and by developing MALDI-MSI methods for reliable imaging of acyl-CoA metabolites in situ. It is possible that these two alternate hypotheses are not mutually exclusive (that both pathways cooperate to synthesize TAG in the embryonic axis), but in no seed system has the relative contributions of DGAT and PDAT been strictly defined. The ability to image acyl-CoAs in tissues, preferably simultaneously with other major lipid species, would greatly help to assess the acyl-CoA pool available for acylation, and may help to clarify any regional differences in the acyltransferase activities that contribute to TAG biosynthesis. In cotton embryos, where deep-sequencing analysis by RNA-Seq methodology has been performed on cotyledons and embryonic axis tissues separately, some insights are available regarding the relative expression of DGAT and PDAT [114]. Unigenes were identified for DGAT1 and DGAT2 isoforms (2 unigenes each), and the relative transcript abundances for all were lower on a RPKM basis than most PDATs (15 unigenes). For DGATs there was no real difference in transcript amounts between cotyledons and the embryonic axis. For PDATs, transcripts were in both cotyledonary and embryonic axis tissues with two of the more abundant PDAT unigenes showing a 2–3 greater abundance in cotyledonary tissues relative to the embryonic axis. Although it is difficult to assess much more in terms of contributions of these acyltransferases to TAG synthesis in cotton embryo tissues, it is important to note that both acyltransferases occur in both cotyledons and axis, pointing to the redundancy of these pathways and the continued complexity of TAG biosynthesis even in oilseeds with a relatively simple fatty acid composition like cotton (e.g., 90% of the FA are 16:0, 18:1, and 18:2). 2.2. Total oil distribution by nuclear magnetic resonance (NMR) approaches Non-invasive 1H-NMR imaging [130] approaches for visualizing the total lipid distributions (comprising mostly TAGs) in oilseeds such as soybean [131], rapeseed [132], cotton [41], tobacco [133], and Camelina [43] provide additional insight as to the overall compartmentation of storage lipids in situ. NMR approaches do not have the drawback of PC-mediated suppression of TAGs (like MSI), and therefore are extremely valuable and reliable for evaluating the spatial distribution total lipids in seeds. While the spatial organization of total lipids is quite important in developing a more complete understanding of tissue specific differences in TAG accumulation, NMR imaging approaches cannot provide complex molecular species information. However, NMR imaging does provide excellent complementary and supportive information for MALDI-MS imaging of lipids in seed sections. In cotton and Camelina embryos, the majority of the seed oil is distributed throughout mature cotyledons and embryonic axis. But both types of seeds show localized ‘hot-spots’ of higher oil content. While many seeds store the majority of their oil in the cotyledon tissues (with notable exceptions such as the endosperm tissue of Castor seed (Ricinus communis)) [134], the implications of heterogeneous oil deposition are not entirely clear. It is possible the distributions are simply a consequence of oilseed development and a result of balancing the utilization of carbon for diverse cellular processes (including lipid storage) with different ‘‘regional’’ demands. Obviously a better understanding of the spatial significance of these processes, now that imaging methodologies have revealed this heterogeneity, will be important to developing oilseeds with enhanced total oil content [90,104]. In the future, it might be advantageous to pre-screen intact seeds for specific oil distribution patterns before preparing cryosections destined for MALDI-MSI analysis to correlate the complementary information more directly. Indeed these approaches could provide breeders with higher resolution information for oilseed improvement.

2.3. Spatial information informs metabolic engineering experiments In most cases, improvements in engineering oilseeds with desired seed traits has been a result of a clearer understanding of the contribution of each enzyme or pathway to TAG biosynthesis [92]. Often, a few key selected steps are manipulated in a target organism in a seed specific manner. Often when pathways or enzymes are introduced into the host organisms, the changes, whether they are differences in acyl composition or total oil content, are usually underwhelming [104,135]. We believe that the heterogeneous differences of TAG (and PC) molecular species distributions within oilseeds represent another level of intricacy that must be considered when designing changes in TAG content or composition. Transgenic strategies are often tested in the model plant, A. thaliana, which stores about 35% by weight of TAG in its seeds [94]. From a fundamental research standpoint, the first images of spatial distributions of TAGs in Arabidopsis seeds (Fig. 6) provides an important step towards improving our understanding of TAG biosynthesis in this model system. While the spatial resolution shown here, 25 microns, is not able to resolve chemical distributions at the subcellular level, molecular heterogeneity is still evident throughout the embryo tissues. The array of mutants and transgenics available in Arabidopsis combined with MS imaging should help to unravel complexities and consequences of compartmentalization of TAG biosynthesis in seeds. This information may then translate to other oilseeds. It should be emphasized here that the distributions of TAG molecular species in individual Arabidopsis seeds are not only informative, but the overall molecular percentages averaged over a cross section of a single seed is comparable to reported quantitative values from whole seed extracts by high performance liquid chromatography–MS [94]. This consistency between MS imaging data and quantitative data by other methods demonstrates the robust nature of MS imaging and illustrates the value of visualizing lipids in single Arabidopsis seeds, providing detailed information that can be obtained no other way. In fact, it is possible that this methodology would allow for assessing the effectiveness of a given metabolic engineering strategy directly in the silique, even in a segregating line, so that strategies could be evaluated in planta in a much more rapid manner. This concept is made feasible only through this MS imaging approach. Recent results with Camelina have begun to provide clues as to the impact of genetic modifications on seed oil composition and distribution (Fig. 5) [43]. In addition to wild-type Camelina seeds (Fig. 4), TAGs in transgenic seeds were imaged including lines of high palmitate (16:0) seeds, generated by over-expression of the 16:0-ACP-selective thioesterases from Cuphea pulcherrima (CpuFATB1) (Fig. 5A), high oleate (18:1) seeds, generated by RNAi-mediated suppression of FAD2 and FAE1 (Fig. 5B), and high linoleate (18:2) seeds, generated by RNAi-mediated suppression of FAD3 and FAE1 (Fig. 5C). High palmitate (40% of total FA; 6 increase in 16:0 content) lines showed altered distributions of TAG profiles, compared with the wild type background, but revealed an overall more uniform tissue distribution of TAG molecular species. Despite the overall increases in 16:0 content, endogenous mechanisms for TAG biosynthesis limited the amount of 16:0 incorporated into TAG, and this seemed to result from less incorporation of palmitate into the sn-2 position (most likely) of PC and TAG. MS imaging showed that in this case, heterogeneity of TAG biosynthesis might not as important as enzyme specificity, probably by LPAAT. On the other hand, visualizing the distribution of TAG metabolites in higholeic seeds, indicated that regional differences in TAG accumulation may be important to consider in this situation. In high-oleic Camelina seeds, there was higher content of the targeted FA than in high palmitate seeds (70% 18:1). MS imaging revealed an incomplete scrambling of molecular species heterogeneity in

45

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

transgenic seeds, which is what would be required to accumulate >90% TAG 18:1/18:1/18:1 (Fig. 5B). TAG 54:3 (triolein) and the precursor PC species PC 36:2 (18:1/18:1) were higher in cotyledons than in the embryonic axis, indicating differences among tissues in the biosynthesis of targeted TAG species. Similarly, in high-linoleic lines (57% 18:2), there was a spatial difference in the accumulation of TAG 54:6 (and TAG 52:4) pointing to incomplete engineering of pathways in embryonic axis tissues for maximum TAG 54:6 synthesis. Interestingly, different genes were suppressed using different promoters (FAD2 RNAi, soybean oleosin promoter; FAE1- and FAD3, soybean glycinin-1 promoter). Since most of the incomplete suppression occurred in the embryonic axis, this suggests perhaps that a re-evaluation of the promoters selected may be appropriate to drive the accumulation of 18:2 to a more complete efficiency. Here, then are several good examples of the additional information obtained by MS imaging approaches that would otherwise go undetected in conventional lipidomics analysis of total seed lipid extracts in these transgenic lines. While dissection and analysis of embryo parts would provide similar information, imaging MS requires less material (only a single seed), and can allow for more detailed spatial resolution within cotyledons and axis tissues. It also should be emphasized that in several cases it is not only the compositions that are vary within oilseeds but also the total amounts as seen by NMR (see Section 2.2), so combining imaging approaches with chemical analysis provides the best overall view possible of the TAG biosynthetic landscape in seeds.

3. Imaging phospholipids In higher plants, like in most eukaryotes, phosphatidylcholine (PC) is the most abundant membrane phospholipid. In Arabidopsis, saturated FA such as 16:0 are concentrated at the sn-1 position of

PC, while one of several unsaturated FA, such as 18:1, 18:2, or 18:3, can be found at either or both positions [136]. There are several mechanisms for synthesizing PC in plants including via CDP-choline pathway via aminoalcoholphosphotransferase (AAPT), head group exchange via PDCT, and from lysoPC and acyl CoA via LPCAT (Fig. 7; [94]). The relative ease with which PC ionizes using available matrices in MALDI-MSI has made it an ideal candidate for many imaging studies. Other phospholipid classes can be imaged by MALDI-MS as well, and these approaches in plant tissues have, again revealed a stunning heterogeneity in molecular species distributions in different tissues (e.g., see Fig. 8). As more efforts are made in the imaging of phospholipids by MALDI-MSI, it will likely expand the areas of research that can benefit from elucidating the cellular/tissue-specific distribution to include membrane biology and lipid signaling. 3.1. Phosphatidylcholine as a TAG metabolite and structural membrane lipid PC is a key intermediate in the biosynthesis of TAG as well as a major structural lipid in extra plastidial membranes [94]. In oilseeds, the heterogeneous distribution of PC molecular species had acyl compositions and distribution profiles that indicate precursor–product relationships to downstream TAG products (Figs. 3–5; [41–44]). For example in cotton embryos both TAG and related PC molecules with 18:1- or 18:2-rich species were enriched in the embryonic axis or cotyledons, respectively [41]. In wild-type Camelina embryos, TAG and PC molecules with 18:2or 20:1-rich species (Fig. 4B and C) were co-localized and enriched in the embryonic axis or cotyledons [43]. In seeds of engineered Camelina lines, with either acyl modifications upstream (in case of overexpression of FATB in high palmitate lines) or downstream (in cases of RNAi-suppression of FAD2/FAE1 or FAD3/FAE1) of PC

ea co MAX 65 %

MAX = 18 mol % PE

MAX 65 %

MAX = 18 mol % PC

MAX 45 %

MAX = 45 mol % PI

MAX 65 %

MAX = 18 mol % PA

PE

PC

PI

PA 36:2 OO

36:3 OL

36:4 LL

34:2 PL

34:1 PO

MIN ~ 0%

Fig. 8. Membrane and signaling lipids imaged within high oleic Camelina sativa seeds. Distribution of selected phosphatidylethanolamine (PE), phosphatidylcholine (PC), phosphatidylinositol (PI), and phosphatidic acid (PA) species in high oleate seeds (FAD2/FAE1 RNAi). Each set of species is normalized within their respective class and colored according to the provided legend unless otherwise noted. Scale bar = 500 lm. All species were imaged in positive ionization mode, although PI species generally form negative ions. Each species is noted by the number of total acyl carbons, double bonds, and acyl chain species. Acyl chain species are designated based data from ESI-MS. Fatty acid abbreviations: P, palmitic (16:0); O, oleic (18:1); L, linoleic (18:2).

46

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

synthesis, the distributions of PC molecules with consistent acyl compositions to TAG (Fig. 5) were co-localized based on metabolic relationships in a predictable manner. In oilseeds, this is reasonable considering much of the flux for TAG biosynthesis proceeds through PC, and PC has been noted as a limitation for engineering unusual FA [135,137]. In addition to storage lipids, PC serves as an intermediate in the synthesis of other phospholipids as well. Because polyunsaturated fatty acid synthesis occurs on PC, acyl groups in other phospholipids depend substantially on a metabolic relationship with PC. This could be through acyl-CoA or DAG intermediates, both also precursors for TAG biosynthesis, re-enforcing the complex metabolic relationships between membrane and storage lipids. 3.2. Phosphatidylethanolamine (PE) Phosphatidylethanolamine (PE) is an abundant membrane lipid in many plant tissues. The synthesis of PE has been shown to be important for early embryonic development in Arabidopsis [138]. Imaging several of the major PE species (i.e. PE-34:2 or PE-16:0/ 18:2, and PE-36:4 or PE-18:2/18:2) in cotton embryos revealed a heterogeneous distribution of PE molecular species with respect to localization in cotyledonary and embryonic axis tissues [41]. The two major molecular species of PE in cotton embryos (PE34:2 and PE-36:4) shared the same diacylglycerol backbone as the two major PC species (PC-34:2 and PC-36:4). The species PE34:2 and PC-34:2 were both enriched in the cotyledon tissues. Since PC and PE are both synthesized from DAG by AAPT, this similarity in acyl composition and tissue location between the two classes made perfect metabolic sense. However, this was not the case for other PC/PE species. PE-36:4 and PC-36:4 actually showed opposite distribution patterns – with PC 36:4 being enriched in cotyledonary tissues and PE 36:4 being enriched in embryonic axis tissues. This suggests that membrane phospholipid biosynthesis may be more complex than previously appreciated and that DAG precursors may be incorporated in PC and PE differentially in these two seed parts. Transcriptional profiling of embryonic axis and cotyledonary tissues of cotton embryos indicated six unigenes for AAPT were expressed [114]. Most were expressed at relatively low levels, but there were some differences in tissue-specific distribution of several transcripts for AAPT. Transcripts for three PDCT unigenes were reported in embryos, and two appeared to be preferentially abundant in embryonic axis relative to cotyledons. There also were transcripts identified that suggested alternative pathways for PE and PC synthesis via PS decarboxylation and PE methylation, respectively. While there is essentially no biochemical support for these annotated homologues in PE and PC metabolism in cotton embryos, the notable apparent complexity from a spatial perspective in transcript profiles and metabolite distributions, suggest there is much to learn about the regulation of membrane lipid synthesis in plant tissues. Unlike in cotton where the distributions of PC/PE with similar acyl chains were sometimes different from one another, in transgenic high oleic Camelina (RNAi-FAD2/FAE1) the acyl chains for similar PC and PE molecular species match well, except for PC34:1 (PO) and PE-34:1 (PO) (Fig. 8). Even though PE is overall less abundant than PC, the overall molecular percentages agreed with lipidomics analysis of total acyl composition, so the ability to assess localization of both PC and PE species together in tissue sections appears to be reliable as well as informative. Based on the Camelina and cotton distribution patterns of PE vs. PC, MS imaging (and RNA-Seq analysis in cotton) suggests there might be different contributions of pathways involved in synthesizing these membrane lipids. In addition to its role as a structural membrane lipid, PE is a precursor to lipid signaling molecules, N-acylphosphatidylethanolam-

ines (NAPEs) and N-acylethanolamines (NAEs) that are associated with germination and post-germination growth [139]. Extending the imaging of PE metabolites in situ to co-localization with downstream signaling molecules, NAPE and NAE, could provide a spatial context for how these metabolites function as lipid mediators. 3.3. Phosphatidic acid (PA) and phosphatidylinositol (PI) PA plays two important roles in plants; it is an intermediate in storage and membrane lipid biosynthesis, and it is a prominent lipid mediator in stress responses and other signaling mechanisms [140,141]. The transient nature of PA, and its generally low abundance makes it rather difficult to characterize reliably in situ. There are also experimental concerns about whether PA formation is induced from sample preparation, and in MALDI-MSI, PA peaks suffers from the added complexity of overlapping with in-source fragmentation of PC [142]. Nonetheless, there is some information on PA distributions in seeds. When PA was imaged in high-oleic Camelina seeds (Fig. 8) (after taking in consideration possible fragmentation products of PC), the overall distributions of molecular species were consistent with those for PC (except for PC 34:1, PO), PE and TAG (compare Figs. 5 and 8). This may suggest that in these embryos, the major PA molecular species are intermediates in TAG biosynthesis. In cotton embryos, PA molecular species showed a heterogeneous pattern [41]. PA-36:3, an abundant species in cotton embryos, was relatively uniformly distributed throughout the embryo, with some reduction in the embryonic axis. Other PA species, PA-36:4 and PA-34:2 were enriched in the hypocotyl-root axis and in the micropylar regions of the embryo. The distributions of these PA species were different from that of major PC and TAG species suggesting that they may not represent distributions related to TAG biosynthesis, but instead my reflect a different function such as signaling. Phospholipase D (PLD) and PA are known to interact with ABA signaling in processes related to seed dormancy [143,144]. Surprisingly, 51 unigenes for PLD were identified in cotton embryos by RNA-Seq, with multiple transcripts corresponding to the Arabidopsis-alpha, -beta, -gamma, -delta and zeta PLD isoforms [114]. Despite the multitude of transcripts, they were generally equally distributed between cotyledonary and axis tissues, with several somewhat skewed toward cotyledonary tissues. Other pathways to generate PA, such as PLC and DAG kinase were possible based on occurrence of multiple putative transcripts for these enzymes in cotton embryos [114]. The significance of the extensive machinery that is expressed to generate PA in cotton embryos, and its association with observed PA molecular species distributions, if any, will require additional experimental details, perhaps with measurements and images at improved spatial resolution or other modifications of MSI [145]. Perhaps now with the ability to image lipids in single Arabidopsis seeds, and the plethora of PLD mutants available in A. thaliana [146], MALDI-MS imaging approaches may help shed light on the role(s) of PLD isoforms in the production of certain PA species in different locations of embryos. Application of MALDI-MS imaging technology to the localization of signaling molecules such as PA certainly will open up new approaches for refining models of lipid mediator functions in various stress- or hormone-regulated processes in plant tissues. Phosphatidylinositol has multiple roles in plants. Like other phospholipids, it is a major component on extra plastidial membranes, but it also serves as the metabolic precursor for the polyphosphoinositides that function in different aspects of signal transduction, membrane trafficking and cytoskeletal organization [147]. Localization of PI species, then, could be important in establishing new ideas about many different PI-mediated processes. Imaging molecular species of PI in embryos of high-oleic Camelina show that it is possible to visualize these lipid species in MSI. Distributions of several PI species differed from those of other

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

phospholipids with similar acyl compositions, but more work needs to be done to determine the extent of heterogeneity and the functional significance of the spatial distribution of PI in plant tissues. Due to its inherent properties, reports for imaging PI species in situ have mostly been with different types of matrices and operating the MS in negative ion mode [145,148]. A method developed by Korte and Lee [149] allows for simultaneous imaging of lipid ions in both positive and negative mode, and may offer a means to develop more comprehensive pictures of tissue lipids in sections. The recent use of MALDI-MSI to identify PI species unique to different populations of breast cancer cells [150], points to the remarkable value and significance of this approach for determining regional differences in the distribution of PI molecular species. 4. Potential for future development In its current state, MS imaging and particular MALDI-MSI have been demonstrated to be extremely valuable for revealing metabolite distributions at the tissue and cellular level. Going forward, these methodologies still have much room for improvement including but not limited to improved spatial resolution, software development for extracting the metabolite co-localization information, correlating complementary experimental imaging and non-imaging approaches, combining imaging with RNA/protein analysis, reliable and direct quantification, and modifications to expand the metabolite imaging range. The imaging field from a practical point of view is still relatively undeveloped and will benefit from additional studies that work to integrate biochemical image analysis with biological experimentation.

47

duce a typical 20 lm diameter laser beam to 5 lm [152]. Imaging in microscope mode, as opposed to microprobe mode (Section 1.3), offers another alternative to addressing future improvements in spatial resolution but is currently limited to time-of-flight magnetic sector analyzers [30]. One practical consequence of high-spatial resolution acquisition is the potential cost in time and computational resources using current mass analyzers to raster over many spots. For example, imaging a single cell that is 50 lm  50 lm at 500 nm spatial resolution with 1 s per scan (i.e. approximate scanning speed of a high-resolution Thermo LTQ-Orbitrap™ scan based on our experience) will take approximately six hours. Applying this 500 nm high-spatial imaging resolution to a 1000 lm  1000 lm tissue section (e.g. cross-sectional area of a small Camelina seed, Fig. 1E) would generate around 8+ million spots, over a billion data points, take an estimated 90 days to acquire (at current scanning speeds), and generate a compressed raw file of over a terabyte of data, not to mention the possible consequences of running an instrument for this period of time. So even with improvements in imaging acquisition speed, this experimental design would likely require an advanced computing system to reconstruct and analyze these images. It might be in the end that for most experiments a combination of moderate- and high-resolution scans provides the most desirable spatial information [47] while satisfying some of the practical constraints of current methdologies. Achieving reliable subcellular imaging resolution would be a remarkable technical advance in the MS imaging field; however, from a biological perspective the most value would be attained if users can designate which subcellular structural features to analyze and/or correlate this information with microscopic images of selected cells/subcellular organelles.

4.1. Improvements in spatial resolution 4.2. Metabolite identification and quantification Addressing certain biological questions at the cellular and subcellular level by MS imaging, in particular MALDI-MSI, will require improvements in the acquired spatial resolution (<10 lm) while not sacrificing high sensitivity at each spot/pixel. Some MS imaging approaches such as SIMS already surpass this resolution limit with achievable resolutions of <1 lm but comes with other technical limitations (see Section 1.1). Most commercially available MALDI instruments have moderate spatial resolution capabilities (10– 50 lm, e.g. [83]) that are certainly appropriate for imaging tissues and some larger cell types. Indeed resolutions of 25–50 lm were found to be sufficient to address many questions about oilseed lipid metabolism (e.g. the size of a typical cottonseed cotyledon cell is within this size [123]) and other plant tissues without the need for over sampling and/or reductions in laser beam spot size. Similarly, for relatively larger tissues (e.g. 5–10 cm in diameter such as mature avocado fruit [45]) imaging resolutions around 200– 400 lm can be used to reduce acquisition time from days (when using 25–50 lm) to hours. Due to current instrumentation limitations, tissues that are even larger (>10 cm) would likely have to be divided into pieces and reconstructed computationally after data acquisition. Therefore, much information can be gleaned at current resolution limits such that higher imaging resolution is not a necessity for addressing many questions at the tissue and/or cellular level. However, when subcellular resolution is desired, further modification of the ionization source and/or current approaches is required. Some groups have shown that by over sampling, i.e. rastering with a step size smaller than the laser beam diameter, the resolution can be effectively increased [151]. On the other hand, over sampling can result in tailing laser beam profiles that might alter some metabolites distributions or conclusions [2]. One inexpensive modification proposed by the Caprioli research group is incorporation of pinhole filter in the laser beam path to re-

The current generation of mass spectrometers with accurate mass capabilities provide high-confidence identification for larger metabolites, i.e. m/z >400 and in particular appears reliable for glycerolipid identification [41]. Accurate mass measurements allows for elemental composition calculations that can help confirm or address annotation of known or unknown compounds without additional MS/MS experimentation. Some commercial instruments have also started to incorporate ion mobility cells which provide additional separation capabilities (e.g. proteins vs. lipids vs. carbohydrates vs. etc.) based on the metabolites’ physical structure [153,154]. For lipids, ion mobility could be especially advantageous for separating isobaric ions, i.e. lipids with masses too close to be resolved but are structurally unique such as different head groups, number of double bonds, length of acyl chains, and other possible chemical modifications [155]. These features, in addition to structural elucidation through MS/MS fragmentation, provide several tools for annotation. Going forward it will be important to perform systematic studies (i.e. matrix compatibility, ionization efficiency, degree of in-source fragmentation, fragmentation energies and structures produced, etc.) of representative standards to improve annotation of compounds in MSI. In addition, depositing MSI data to publically available metabolomics databases will help enhance annotation. When possible additional chemical validation of MS imaging results through shotgun lipid/metabolomics, LC–MS/MS, GC–MS, TLC, NMR, etc., is always advisable. Metabolite quantification in MS imaging is more complex in its current form than several other well-established MS approaches. Some major current limitations for quantification include differences in ionization efficiency within selected matrices (or modes of sample preparation for matrix-independent imaging), suppression of certain metabolites and metabolite classes (e.g. PC to TAG), in-source fragmentation, and inclusion of internal standards

48

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

in tissues [85,156]. There are a few ways that these limitations are being addressed suggesting that MS imaging will move towards reliable quantification in the near future. Foremost, quantitative MS imaging of ATP in mice brain slices were improved through correlation with quantitative analysis with capillary electrophoresis (CE)-ESI-MS/MS [157]. In some cases, correction factors can be applied to quantify target analytes, such as in a study of dose-dependent assays of a pharmaceutical olanzapine imaged in liver tissues and corrected against quantitative data from tissue extracts analyzed by LC–MS/MS [85]. However, while continuing to validate MS imaging results against other independent methods is certainly an advisable practice, achieving reliable quantitation with MALDIMS imaging itself is also desirable. To achieve this some groups have added isotopically labeled internal standards to quantify analytes by normalizing the analyte ion signal to the internal standard accounting for variations in ionization efficiency [86]. Ideally this could be done with an internal standard mix at different concentrations where calibration curves are inherently mapped alongside endogenous metabolites. These internal standards could be combined with matrix (for MALDI) in solution before deposition to the tissue or applied separately. Naturally, there are concerns about solubility of selected standards and ability to deposit such a solution mix in a quantitative manner. Not to mention the addition(s) of any organic-based solvent containing these standards could also possibly alter the distribution of metabolites in situ. Ideally, this sort of approach would be optimized through a systematic process to test both the deposition of individual standards (and within mixes) within different solvent mediums and modes of deposition. The quantitative deposition and ionization of selected matrices (i.e. number of molecules deposited is quantitatively proportional and predictable to number of molecules ionization and detected) would further strengthen this quantitative approach. Ultimately there are still details about the ionization process [71] that when further elucidated will help design more quantitative experiments. If the ionization, ion stability, and detection parameters are better understood from both a theoretical and applied framework, at least for most metabolite classes, then quantitation should be comparable to other analytical approaches. 4.3. Expanding range of metabolites and tissues for imaging Lipids, especially those bearing acyl chains, are particularly amenable to the ionization processes currently used for MS imaging approaches. Many membrane and storage lipid species have been successfully localized as noted above. However, there are still several lipids that remain relatively elusive including acyl-CoAs, diacylglycerols, polyphosphoinositides, oxylipins, and small lipid molecules such as fatty acids and lipophilic hormones. From a grander standpoint, mapping all metabolites in a tissue simultaneously would be extremely valuable as a complement to metabolomics. Already several metabolite classes beyond acyl lipids have been visualized by MSI such as carbohydrates, organic acids, flavonoids, and glucosides [19,40] and exogenously applied compounds such as pharmaceuticals [158,159]. Nevertheless, in most studies thus far, only a subset of metabolites or selected metabolite classes is detected. It is presumed that matrix compatibility is the number one factor for ionizing different sets of molecules. As additional standard chemical compounds are tested in MS imaging instrumentation, it should be clearer what conditions can best be used to expand the range of metabolites that are amenable to MSI. In practice it might be that MS imaging has to take a similar approach to metabolomics by HPLC–MS where there is one largely-accepted, general matrix (e.g. 2,5-dihydroxybenzoic acid, comparable to the general extraction methods using solvents such as water/methanol in metabolomics) that ionize at least most compounds for analyses. Then, further steps could be taken with other,

more specialized matrices or conditions, for more recalcitrant chemical compounds. In the end, more biochemists taking approaches to visualize metabolism in situ will help to broaden the range of metabolites that can be localized by MSI. MS imaging of mature seeds offer unique technical advantages for visualizing lipid molecules due to their preservation-like state and abundant lipid content. Going forward it will be important to examine the applicability of MS imaging to other diverse plant tissues such as developing seeds, leaves, roots, stems, flowers, etc. for lipid visualization. Indeed there have been several publications imaging these types of representative tissues within various imaging platforms (Sections 1.1–1.3). Our group in collaboration with other researchers has had at least some success imaging lipids in diverse tissues such as tobacco (N. tabacum) leaves [44], and A. thaliana stems, lily (Lilium spp.) anthers and strawberry (Fragaria spp.) stolons (unpublished data). Each of these tissues required careful sample preparation specific to its structural stability (Section 1.5.1) in addition to unique imaging challenges due to the ionization of different ratios of metabolites (and contaminants) in each tissue. For example in wild-type tobacco leaves, imaging TAG species was relatively difficult due to its low abundance (<0.2% dry weight) relative to other produced ions [44]. However, within engineered tissues that boosted the TAG content significantly (>15% TAG by dry weight), there was little problem to successfully image TAG species in leaf cross sections. Imaging the distribution of TAG molecules in these engineered leaves demonstrated a highly uniform deposition of major TAG molecule species within the mesophyll cells and this was supported by confocal laser scanning microscopy that revealed a preponderance of lipid droplets throughout mesophyll cells in the leaves. Without MS imaging, it is difficult to evaluate the detailed metabolic spatial consequences of a chosen engineering strategy. In the case of tobacco leaves designed to produce triacylglycerols at 15% of leaf dry weight, the MALDI-MS imaging indicated a homogenous distribution of TAG molecular species which is important for optimizing TAG yields in vegetative tissues. This is in contrast to the results of several metabolic engineering cases within oilseeds presented in Section 2 further emphasizing the need to evaluate both new tissues as well as genetically engineered tissues. Additionally, going forward it will be important to optimize some of the conditions required to image less abundant molecules and to ensure the structural and chemical integrity of metabolite distributions in a wide range of tissues. The inclusion of other tissues opens up whether it is possible to take in vivo MS imaging measurements on a living plant. While most imaging approaches require the sample to be sectioned and analyzed under vacuum there are other, less invasive approaches such as DESI [20] or LAESI [160] that might make some in vivo imaging measurements possible. However, this could be most applicable to surface metabolites which prevents the probing of metabolism at tissues deeper in the organism of interest. Nevertheless, these surface-oriented applications may yield information about the analysis of several interesting chemical pathways [35]. 4.4. Co-localization of metabolites One important aspect of MS imaging is to connect spatially metabolites that have a precursor/product relationship or are involved in some interconnected metabolic pathway. Metabolites of most major pathways in plants can be analyzed through conventional biochemical analyses [161], and these biochemical pathways can be probed/elucidated through genetic approaches. As the metabolomics field has expanded, chemical profiling of whole cell/tissue metabolism has become feasible. With more MS imaging data (with enhanced range of metabolites, see section 4.3) acquired and made available, data-mining for uncharacterized metabolites may

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

become important to extract how ‘‘location’’ fits into a larger, spatially-relevant metabolic network. Visualizing the co-localization of targeted metabolites such as in PC to TAG precursor/product relationships in oilseeds have helped reveal the location and activities involved in TAG biosynthesis at the tissue level (Section 2). Another example of using the computer algorithms (via Metabolite Imager, Section 1.4) to reveal co-localization of metabolites was to map precursors of the terpenoid, gossypol, in glandular structures of cotton seeds [46]. Algorithms were used to map potential metabolites localized within the same pigmented glands. Two gossypol precursor metabolites, hemigossypol and desoxyhemigossypol, were identified and shown to be coincident in the same glands with gossypol itself, demonstrating the co-localization of precursor-product relationships with more refined spatial resolution. Further advancements in subcellular MS imaging (Section 4.1) analysis of organelles in situ will take the concept of co-localization of metabolites from the cellular level, like with the lysigenous glands of cotton, to the subcellular level, where concepts of metabolic channeling, might be able to be visualized. Additional computational methods aimed at co-localization of unknown metabolites might uncover less studied or even new biochemical pathways in situ. 4.5. Alternative imaging approaches and correlated imaging The broad field of lipid visualization [162] is not limited to MS imaging approaches (several additional variations of which are beyond the scope of this review [2,23,160]), but also includes both traditional approaches such as tissue microdissection with chemical extraction and histological/fluorescent imaging as well as newer approaches such as direct organelle MS [163] and by noninvasive surface imaging methods such as Raman [164,165], NMR [132] and photoacoustic [166,167] imaging. Historically, tissue microdissection has enabled the chemical analysis of selected tissues and might continue to be improved with the incorporation of laser microdissection devices [168]. However, for lipid visualization purposes the resolution is usually limited to whole tissues which can result in lower metabolite quantities due to losses from chemical extraction protocols and tedious sample handling. Confocal imaging of chemically tagged lipids with fluorescent based dyes (e.g. BODIPY) can improve the localization of these lipids at the sub-cellular level but at this stage lacks detailed chemical information and sometimes leads to imaging artifacts. Alternatively, direct organelle MS using electronically-controlled sampling devices can provide limited localization information for sub-cellular organelles with improved molecular identification. Finally, several non-invasive imaging methods are appropriate when tissue samples cannot be destroyed and chemical tagging of the lipids is undesirable. However, these approaches are still limited in the depth of chemical information provided, often only able to distinguish between ‘‘bulk content’’ of structurally different molecules such as lipids, water, and proteins. At this stage MS imaging through MALDI-MS or related approaches appears to provide the best balance between sample preparation difficulty, visualization potential, and detailed chemical information obtained. Still, correlated imaging (reviewed in [169]) between complementary visualization approaches, such as MALDI-MSI (i.e. molecular identification) and confocal Raman microscopy (CRM, i.e. functional groups), provides a promising avenue for enhanced biological imaging. The principle of correlated imaging is to use two (or more) platforms in combination, either as a hybrid instrument or in sequential analyses, to overcome the limitations of either instrument/approach alone. At its simplest example, this concept is already used for most MS imaging experiments by capturing microscopic images of their tissues before chemical analysis. The microscopic images provide information about anatomical features,

49

but with limited compositional information, while the MSI instrument provides the complementary metabolite information. Additionally, several groups make serial tissue sections where a similar tissue section (to one analyzed by MS imaging) is chemically analyzed by approaches such as LC–MS/MS or LDI/SIMS/CRM [170]. We recently applied the concept of correlated imaging by spatially mapping the total oil content distribution in whole cotton embryos by nuclear magnetic resonance (NMR) imaging and representative embryo cryosections by MALDI-MSI [41]. There are still many challenges for correlated imaging that remain to be addressed, such as optimizing each technology alone and in combination, correlating position information from experiments at different resolutions, and resolving differences in the dynamic range of different analytical approaches [169]. But the ability to acquire information from complementary approaches offers a future way to improve the mapping detail of metabolites and other cellular components in situ. 4.6. Combine with transcriptomics and/or proteomics to co-localize metabolites, enzymes and expressed genes One of the current limitations with a metabolomics-only approach to MS imaging is the absence of key pieces of transcriptional (mRNA) and translational (protein) information. Unfortunately most current MS imaging methodologies are not able to simultaneously acquire information on metabolites, protein, and mRNA. The technical limitations center on differences in abundance, matrix compatibility and MS detectors that are optimal for one group of molecules but with limited to no information on others [171–173]. For transcript analysis in a spatial context, the classical technique of in situ hybridization on tissue sections [174,175] or, when feasible, simple tissue dissection and reverse transcriptase-PCR [42] offer two approaches that would complement MS imaging with gene expression data. Interestingly, an older study analyzing both DNA and RNA molecules by infraredMALDI-MS found that prepared RNA transcripts (testing up to 104 nucleotides) could be ionized and detected with minimal structural degradation (in contrast to DNA which was much more unstable) [176]. Other groups have used MALDI instrumentation to sequence RNA digests and detect RNA modifications [177] providing hope these molecules could be imaged at some point in the near future [172]. In situ imaging of RNA is likely much more technically complex and to our knowledge has not been reliably demonstrated at this point. Imaging of the metabolites in situ can helps to predict where the proteins involved in the presumed pathways are localized and how they might be operating (see Section 2). Indeed MS imaging of proteins has its own technical limitations without even considering simultaneous or correlated imaging with other metabolites [178]. In general, proteomics imaging protocols require sample preparation steps that will remove most lipids (and other metabolites) from tissues before analysis [55,62]. Therefore, until improvements are made in sample preparation methods and lipid-protein compatible matrices, it will be necessary to separate the two analyses. One could imagine an approach where serial sections are used to map proteins and metabolites in a given tissue, or perhaps even a single section might be imaged first for lipids followed by matrix/lipid removal and re-imaging for proteins. In any case, combining-omics technologies together to probe metabolism in a spatial context, represents a next frontier for cellular (and subcellular) biochemistry. References [1] McDonnell LA, Heeren R, Andrén PE, Stoeckli M, Corthals GL. Going forward: increasing the accessibility of imaging mass spectrometry. J Proteomics 2012;75:5113–21.

50

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

[2] Lee YJ, Perdian DC, Song Z, Yeung ES, Nikolau BJ. Use of mass spectrometry for imaging metabolites in plants. Plant J 2012;70:81–95. [3] Norris JL, Caprioli RM. Analysis of tissue specimens by matrix-assisted laser desorption/ionization imaging mass spectrometry in biological and clinical research. Chem Rev 2013;113:2309–42. [4] Jungmann JH, Heeren RMA. Emerging technologies in mass spectrometry imaging. J Proteomics 2012;16:5077–92. [5] Boxer SG, Kraft ML, Weber PK. Advances in imaging secondary ion mass spectrometry for biological samples. Annu Rev Biophys 2009;38:53–74. [6] Passarelli MK, Winograd N. Lipid imaging with time-of-flight secondary ion mass spectrometry (ToF-SIMS). Biochim Biophys Acta Mol Cell Biol Lipids 2011;1811:976–90. [7] Castaing R, Slodzian G. Microanalyse par émission ionique secondaire. J Microsc 1962;1:395–410. [8] Benninghoven A, Loebach E. Tandem mass spectrometer for secondary ion studies. Rev Sci Instrum 1971;42:49–52. [9] Chait BT, Standing KG. A time-of-flight mass spectrometer for measurement of secondary ion mass spectra. Int J Mass Spectrom Ion Phys 1981;40:185–93. [10] Johansson B. ToF-SIMS imaging of lipids in cell membranes. Surf Interface Anal 2006;38:1401–12. [11] Brunelle A, Laprévote O. Lipid imaging with cluster time-of-flight secondary ion mass spectrometry. Anal Bioanal Chem 2009;393:31–5. [12] Braun RM, Beyder A, Xu J, Wood MC, Ewing AG, Winograd N. Spatially resolved detection of attomole quantities of organic molecules localized in picoliter vials using time-of-flight secondary ion mass spectrometry. Anal Chem 1999;71:3318–24. [13] Nygren H, Börner K, Malmberg P, Tallarek E, Hagenhoff B. Imaging TOF-SIMS of rat kidney prepared by high-pressure freezing. Microsc Res Tech 2005;68:329–34. [14] Yang H-J, Ishizaki I, Sanada N, Zaima N, Sugiura Y, Yao I, et al. Detection of characteristic distributions of phospholipid head groups and fatty acids on neurite surface by time-of-flight secondary ion mass spectrometry. Med Mol Morphol 2010;43:158–64. [15] Fletcher JS, Lockyer NP, Vaidyanathan S, Vickerman JC. TOF-SIMS 3D biomolecular imaging of Xenopus laevis oocytes using buckminsterfullerene (C60) primary ions. Anal Chem 2007;79:2199–206. [16] Passarelli MK, Winograd N. Characterizing in situ glycerophospholipids with SIMS and MALDI methodologies. Surf Interface Anal 2011;43:269–71. [17] Seyer A, Einhorn J, Brunelle A, Lapre´vote O. Localization of flavonoids in seeds by cluster time-of-flight secondary ion mass spectrometry imaging. Anal Chem 2010;82:2326–33. [18] Takáts Z, Wiseman JM, Gologan B, Cooks RG. Mass spectrometry sampling under ambient conditions with desorption electrospray ionization. Science 2004;306:471–3. [19] Li B, Knudsen C, Hansen NK, Jørgensen K, Kannangara R, Bak S, et al. Visualizing metabolite distribution and enzymatic conversion in plant tissues by desorption electrospray ionization mass spectrometry imaging. Plant J 2013;74:1059–71. [20] Dill AL, Ifa DR, Manicke NE, Ouyang Z, Cooks RG. Mass spectrometric imaging of lipids using desorption electrospray ionization. J Chrom B 2009;877:2883–9. [21] Harris GA, Galhena AS, Ferna´ndez FM. Ambient sampling/ionization mass spectrometry: applications and current trends. Anal Chem 2011;83:4508–38. [22] Campbell D, Ferreira C, Eberlin L, Cooks RG. Improved spatial resolution in the imaging of biological tissue using desorption electrospray ionization. Anal Bioanal Chem 2012;404:389–98. [23] Ellis SR, Brown SH, In Het Panhuis M, Blanksby SJ, Mitchell TW. Surface analysis of lipids by mass spectrometry: more than just imaging. Prog Lipid Res 2013;52:329–53. [24] Müller T, Oradu S, Ifa DR, Cooks RG, Kräutler B. Direct plant tissue analysis and imprint imaging by desorption electrospray ionization mass spectrometry. Anal Chem 2011;83:5754–61. [25] Li B, Hansen SH, Janfelt C. Direct imaging of plant metabolites in leaves and petals by desorption electrospray ionization mass spectrometry. Int J Mass Spectrom 2013;348:15–22. [26] Thunig J, Hansen SH, Janfelt C. Analysis of secondary plant metabolites by indirect desorption electrospray ionization imaging mass spectrometry. Anal Chem 2011;83:3256–9. [27] Ifa DR, Srimany A, Eberlin LS, Naik HR, Bhat V, Cooks RG, et al. Tissue imprint imaging by desorption electrospray ionization mass spectrometry. Anal Methods 2011;3:1910–2. [28] Fernandez JA, Ochoa B, Fresnedo O, Giralt MT, Rodriguez-Puertas R. Matrixassisted laser desorption ionization imaging mass spectrometry in lipidomics. Anal Bioanal Chem 2011;401:29–51. [29] Gode D, Volmer DA. Lipid imaging by mass spectrometry – a review. Analyst 2013;138:1289–315. [30] Kiss A, Jungmann JH, Smith DF, Heeren R. Microscope mode secondary ion mass spectrometry imaging with a Timepix detector. Rev Sci Instrum 2013;84. 013704-013704. [31] Wu WW, Wang G, Baek SJ, Shen R-F. Comparative study of three proteomic quantitative methods, DIGE, cICAT, and iTRAQ, using 2D gel-or LC-MALDI TOF/TOF. J Proteome Res 2006;5:651–8. [32] Fuchs B, Süß R, Schiller J. An update of MALDI-TOF mass spectrometry in lipid research. Prog Lipid Res 2010;49:450–75. [33] Zhang H, Cha S, Yeung ES. Colloidal graphite-assisted laser desorption/ ionization MS and MS(n) of small molecules. 2. Direct profiling and MS imaging of small metabolites from fruits. Anal Chem 2007;79:6575–84.

[34] Cha S, Zhang H, Ilarslan HI, Wurtele ES, Brachova L, Nikolau BJ, et al. Direct profiling and imaging of plant metabolites in intact tissues by using colloidal graphite-assisted laser desorption ionization mass spectrometry. Plant J 2008;55:348–60. [35] Jun JH, Song Z, Liu Z, Nikolau BJ, Yeung ES, Lee YJ. High-spatial and high-mass resolution imaging of surface metabolites of Arabidopsis thaliana by laser desorption-ionization mass spectrometry using colloidal silver. Anal Chem 2010;82:3255–65. [36] Vrkoslav V, Muck A, Cvacka J, Svatos A. MALDI imaging of neutral cuticular lipids in insects and plants. J Am Soc Mass Spectrom 2010;21:220–31. [37] Zaima N, Goto-Inoue N, Hayasaka T, Setou M. Application of imaging mass spectrometry for the analysis of Oryza sativa rice. Rapid Commun Mass Spectrom 2010;24:2723–9. [38] Taira S, Ikeda R, Yokota N, Osaka I, Sakamoto M, Kato M, et al. Mass spectrometric imaging of ginsenosides localization in Panax ginseng root. Am J Chin Med 2010;38:485–93. [39] Taira S, Shimma S, Osaka I, Kaneko D, Ichiyanagi Y, Ikeda R, et al. Mass spectrometry imaging of the capsaicin localization in the capsicum fruits. Int J Biotechnol Wellness Ind 2012;1:61–5. [40] Ye H, Gemperline E, Venkateshwaran M, Chen R, Delaux PM, Howes-Podoll M, et al. MALDI mass spectrometry-assisted molecular imaging of metabolites during nitrogen fixation in the Medicago truncatula–Sinorhizobium meliloti symbiosis. Plant J 2013;75:130–45. [41] Horn PJ, Korte AR, Neogi PB, Love E, Fuchs J, Strupat K, et al. Spatial mapping of lipids at cellular resolution in embryos of cotton. Plant Cell 2012;24:622–36. [42] Horn PJ, Sturtevant D, Chapman KD. Modified oleic cottonseeds show altered content, composition and tissue-specific distribution of triacylglycerol molecular species. Biochimie 2014;96:28–36. [43] Horn PJ, Silva JE, Anderson D, Fuchs J, Borisjuk L, Nazarenus TJ, et al. Imaging heterogeneity of membrane and storage lipids in transgenic Camelina sativa seeds with altered fatty acid profiles. Plant J 2013;76:138–50. [44] Vanhercke T, El Tahchy A, Liu Q, Zhou X-R, Shrestha P, Diva U, et al. Metabolic engineering of biomass for high energy density: oilseed-like triacylglycerol yields from plant leaves. Plant Biotechnol J 2013; doi: 10.1111/pbi.12131. [Epub ahead of print]. [45] Horn PJ, James CN, Gidda S, Kilaru A, Dyer JM, Mullen RT, et al. Identification of a new class of lipid droplet-associated proteins in plants. Plant Physiol 2013;162:1926–36. [46] Horn PJ, Chapman KD. Metabolite Imager: customized spatial analysis of metabolite distributions in mass spectrometry imaging. Metabolomics J 2013. http://dx.doi.org/10.1007/s11306-013-0575-0. [47] Perdian DC, Lee YJ. Imaging MS methodology for more chemical information in less data acquisition time utilizing a hybrid linear ion trap-orbitrap mass spectrometer. Anal Chem 2010;82:9393–400. [48] Verhaert PEM, Pinkse MH, Strupat K, Conaway MP. Imaging of similar mass neuropeptides in neuronal tissue by enhanced resolution MALDI MS with an ion trap-OrbitrapTM hybrid instrument. In: Rubakhin SS, Sweedler JV, editors. Mass spectrometry imaging. Humana Press; 2010. p. 433–49. [49] McDonnell LA, van Remoortere A, de Velde N, van Zeijl RJM, Deelder AM. Imaging mass spectrometry data reduction: automated feature identification and extraction. J Am Soc Mass Spectrom 2010;21:1969–78. [50] Källback P, Shariatgorji M, Nilsson A, Andrén PE. Novel mass spectrometry imaging software assisting labeled normalization and quantitation of drugs and neuropeptides directly in tissue sections. J Proteomics 2012;75:4941–51. [51] Xiong X, Xu W, Eberlin L, Wiseman J, Fang X, Jiang Y, et al. Data processing for 3D mass spectrometry imaging. J Am Soc Mass Spectrom 2012;23:1147–56. [52] Schramm T, Hester A, Klinkert I, Both J-P, Heeren RMA, et al. ImzML – a common data format for the flexible exchange and processing of mass spectrometry imaging data. J Proteomics 2012;75:5106–10. [53] Kaletasß BK, van der Wiel IM, Stauber J, Lennard JD, Güzel C, Kros JM, et al. Sample preparation issues for tissue imaging by imaging MS. Proteomics 2009;9:2622–33. [54] Schwartz SA, Reyzer ML, Caprioli RM. Direct tissue analysis using matrixassisted laser desorption/ionization mass spectrometry: practical aspects of sample preparation. J Mass Spectrom 2003;38:699–708. [55] Goodwin RJ. Sample preparation for mass spectrometry imaging: small mistakes can lead to big consequences. J Proteomics 2012;75:4893–911. [56] Chaurand P, Cornett DS, Angel PM, Caprioli RM. From whole-body sections down to cellular level, multiscale imaging of phospholipids by MALDI mass spectrometry. Mol Cell Proteomics 2011; 10. doi: 10.1074/mcp.O110.004259. [57] Hart P, Francese S, Claude E, Woodroofe MN, Clench M. MALDI-MS imaging of lipids in ex vivo human skin. Anal Bioanal Chem 2011;401:115–25. [58] Angel PM, Spraggins JM, Baldwin HS, Caprioli R. Enhanced sensitivity for high spatial resolution lipid analysis by negative ion mode MALDI imaging mass spectrometry. Anal Chem 2012;84:1557–64. [59] Seeley EH, Caprioli RM. MALDI imaging mass spectrometry of human tissue: method challenges and clinical perspectives. Trends Biotechnol 2011;29:136–43. [60] Sugiura Y, Zaima N, Setou M, Ito S, Yao I. Visualization of acetylcholine distribution in central nervous system tissue sections by tandem imaging mass spectrometry. Anal Bioanal Chem 2012;403:1851–61. [61] Walch A, Rauser S, Deininger S-O, Höfler H. MALDI imaging mass spectrometry for direct tissue analysis: a new frontier for molecular histology. Histochem Cell Biol 2008;130:421–34. [62] Kaspar S, Peukert M, Svatos A, Matros A, Mock H-P. MALDI-imaging mass spectrometry – an emerging technique in plant biology. Proteomics 2011;11:1840–50.

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52 [63] Chen Y, Allegood J, Liu Y, Wang E, Cachon-Gonzalez B, Cox TM, et al. Imaging MALDI mass spectrometry using an oscillating capillary nebulizer matrix coating system and its application to analysis of lipids in brain from a mouse model of Tay Sachs/Sandhoff disease. Anal Chem 2008;80:2780–8. [64] Reyzer ML, Chaurand P, Angel PM, Caprioli RM. Direct molecular analysis of whole-body animal tissue sections by MALDI imaging mass spectrometry. Methods Mol Biol 2010;656:285–301. [65] Kruse R, Sweedler J. Spatial profiling invertebrate ganglia using MALDI MS. J Am Soc Mass Spectrom 2003;14:752–9. [66] Stoeckli M, Staab D, Schweitzer A. Compound and metabolite distribution measured by MALDI mass spectrometric imaging in whole-body tissue sections. Int J Mass Spectrom 2007;260:195–202. [67] Sugiura Y, Shimma S, Setou M. Thin sectioning improves the peak intensity and signal-to-noise ratio in direct tissue mass spectrometry. J Mass Spectrom Soc Jap 2006;54:45–8. [68] Goto-Inoue N, Hayasaka T, Zaima N, Setou M. Imaging mass spectrometry for lipidomics. Biochim Biophys Acta Mol Cell Biol Lipids 2011;1811:961–9. [69] Kawamoto T. Use of a new adhesive film for the preparation of multi-purpose fresh-frozen sections from hard tissues, whole-animals, insects and plants. Arch Histol Cytol 2003;66:123–43. [70] Malm J, Giannaras D, Riehle MO, Gadegaard N, Sjövall P. Fixation and drying protocols for the preparation of cell samples for time-of-flight secondary ion mass spectrometry analysis. Anal Chem 2009;81:7197–205. [71] Zenobi R, Knochenmuss R. Ion formation in MALDI mass spectrometry. Mass Spectrom Rev 1998;17:337–66. [72] Schiller J, Süß R, Fuchs B, Müller M, Petkovic´ M, Zschörnig O, et al. The suitability of different DHB isomers as matrices for the MALDI-TOF MS analysis of phospholipids: which isomer for what purpose? Eur Biophys J 2007;36:517–27. [73] Sun G, Yang K, Zhao Z, Guan S, Han X, Gross RW. Matrix-assisted laser desorption/ionization time-of-flight mass spectrometric analysis of cellular glycerophospholipids enabled by multiplexed solvent dependent analyte matrix interactions. Anal Chem 2008;80:7576–85. [74] Fuchs B, Bischoff A, Süß R, Teuber K, Schürenberg M, Suckau D, et al. Phosphatidylcholines and -ethanolamines can be easily mistaken in phospholipid mixtures: a negative ion MALDI-TOF MS study with 9aminoacridine as matrix and egg yolk as selected example. Anal Bioanal Chem 2009;395:2479–87. [75] Stübiger G, Belgacem O. Analysis of lipids using 2,4,6-trihydroxy acetophenone as a matrix for MALDI mass spectrometry. Anal Chem 2007; 79:3206–13. [76] Lorkiewicz P, Cecilia Yappert M. 2-(2-Aminoethylamino)-5-nitropyridine as a basic matrix for negative-mode matrix-assisted laser desorption/ionization analysis of phospholipids. J Mass Spectrom 2009;44:137–43. [77] Jackson SN, Wang H-YJ, Woods AS, Ugarov M, Egan T, Schultz JA. Direct tissue analysis of phospholipids in rat brain using MALDI-TOFMS and MALDI-ion mobility-TOFMS. J Am Soc Mass Spectrom 2005;16:133–8. [78] Spengler B, Kirsch D, Kaufmann R. Fundamental aspects of postsource decay in matrix-assisted laser desorption mass spectrometry. 1. Residual gas effects. J Phys Chem 1992;96:9678–84. [79] Al-Saad KA, Zabrouskov V, Siems WF, Knowles NR, Hannan RM, Hill Jr HH. Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry of lipids: ionization and prompt fragmentation patterns. Rapid Commun Mass Spectrom 2003;17:87–96. [80] Picariello G, Romano R, Addeo F. Nitrocellulose film substrate minimizes fragmentation in matrix-assisted laser desorption ionization time-of-flight mass spectrometry analysis of triacylglycerols. Anal Chem 2010;82:5783–91. [81] Emerson B, Gidden J, Lay JO, Durham B. A rapid separation technique for overcoming suppression of triacylglycerols by phosphatidylcholine using MALDI-TOF MS. J Lipid Res 2010;51:2428–34. [82] Berry KAZ, Li B, Reynolds SD, Barkley RM, Gijón MA, Hankin JA, et al. MALDI imaging MS of phospholipids in the mouse lung. J Lipid Res 2011;52:1551–60. [83] Hölscher D, Shroff R, Knop K, Gottschaldt M, Crecelius A, Schneider B, et al. Matrix-free UV-laser desorption/ionization (LDI) mass spectrometric imaging at the single-cell level: distribution of secondary metabolites of Arabidopsis thaliana and Hypericum species. Plant J 2009;60:907–18. [84] Nizioł J, Rode W, Zielin´ski Z, Ruman T. Matrix-free laser desorption– ionization with silver nanoparticle-enhanced steel targets. Int J Mass Spectrom 2013;335:22–32. [85] Koeniger SL, Talaty N, Luo Y, Ready D, Voorbach M, Seifer T, et al. A quantitation method for mass spectrometry imaging. Rapid Commun Mass Spectrom 2011;25:503–10. [86] Pirman DA, Reich RF, Kiss A, Heeren RMA, Yost RA. Quantitative MALDI tandem mass spectrometric imaging of cocaine from brain tissue with a deuterated internal standard. Anal Chem 2012;85:1081–9. [87] Kind T, Fiehn O. Metabolomic database annotations via query of elemental compositions: mass accuracy is insufficient even at less than 1 ppm. BMC Bioinf 2006;7:234. [88] Taylor CF, Binz PA, Aebersold R, Affolter M, Barkovich R, Deutsch EW, et al. Guidelines for reporting the use of mass spectrometry in proteomics. Nature 2008;26:860–1. [89] Fiehn O, Robertson D, Griffin J, van der Werf M, Nikolau B, Morrison N, et al. The metabolomics standards initiative (MSI). Metabolomics 2007;3:175–8. [90] Lu C, Napier JA, Clemente TE, Cahoon EB. New frontiers in oilseed biotechnology: meeting the global demand for vegetable oils for food, feed, biofuel, and industrial applications. Curr Opin Biotechnol 2011;22:252–9.

51

[91] Chapman KD, Ohlrogge JB. Compartmentation of triacylglycerol accumulation in plants. J Biol Chem 2012;287:2288–94. [92] Bates PD, Stymne S, Ohlrogge J. Biochemical pathways in seed oil synthesis. Curr Opin Plant Biol 2013;16:358–64. [93] Snapp AR, Lu C. Engineering industrial fatty acids in oilseeds. Front Biol 2012;8:1–10. [94] Li-Beisson Y, Shorrosh B, Beisson F, Andersson MX, Arondel V, Bates PD, et al. Acyl-lipid metabolism. Arabidop Book 2013:e0161. [95] Harwood JL. Recent advances in the biosynthesis of plant fatty acids. Biochim Biophys Acta 1996;1301:7–56. [96] Ohlrogge JB, Kuhn DN, Stumpf PK. Subcellular localization of acyl carrier protein in leaf protoplasts of Spinacia oleracea. Proc Nat Acad Sci U S A 1979;76:1194–8. [97] Schwender J, Shachar-Hill Y, Ohlrogge JB. Mitochondrial metabolism in developing embryos of Brassica napus. J Biol Chem 2006;281:34040–7. [98] Ohlrogge J, Browse J. Lipid biosynthesis. Plant Cell 1995;7:957–70. [99] Tjellström H, Yang Z, Allen DK, Ohlrogge JB. Rapid kinetic labeling of Arabidopsis cell suspension cultures: implications for models of lipid export from plastids. Plant Physiol 2012;158:601–11. [100] Ghanevati M, Jaworski JG. Engineering and mechanistic studies of the arabidopsis FAE1 b-ketoacyl-CoA synthase, FAE1 KCS. Eur J Biochem 2002;269:3531–9. [101] Rossak M, Smith M, Kunst L. Expression of the FAE1 gene and FAE1 promoter activity in developing seeds of Arabidopsis thaliana. Plant Mol Biol 2001;46: 717–25. [102] Okuley J, Lightner J, Feldmann K, Yadav N, Lark E, Browse J. Arabidopsis FAD2 gene encodes the enzyme that is essential for polyunsaturated lipid synthesis. Plant Cell 1994;6:147–58. [103] Lemieux B, Miquel M, Somerville C, Browse J. Mutants of Arabidopsis with alterations in seed lipid fatty acid composition. Theor Appl Genet 1990;80: 234–40. [104] Napier JA, Graham IA. Tailoring plant lipid composition: designer oilseeds come of age. Curr Opin Plant Biol 2010;13:329–36. [105] Welti R, Li W, Li M, Sang Y, Biesiada H, Zhou HE, et al. Profiling membrane lipids in plant stress responses. J Biol Chem 2002;277:31994–2002. [106] Canvin DT. The effect of temperature on the oil content and fatty acid composition of the oils from several oil seed crops. Can J Bot 1965;43: 63–9. [107] Bates PD, Fatihi A, Snapp AR, Carlsson AS, Lu C. Acyl editing and headgroup exchange are the major mechanisms that direct polyunsaturated fatty acid flux into triacylglycerols. Plant Physiol 2012;160:1530–9. [108] Bao X, Katz S, Pollard M, Ohlrogge J. Carbocyclic fatty acids in plants: Biochemical and molecular genetic characterization of cyclopropane fatty acid synthesis of Sterculia foetida. Proc Nat Acad Sci U S A 2002;99:7172–7. [109] Bao X, Thelen J, Bonaventure G, Ohlrogge J. Characterization of cyclopropane fatty-acid synthase from Sterculia foetida. J Biol Chem 2003;278:12846–53. [110] Yu X-H, Rawat R, Shanklin J. Characterization and analysis of the cotton cyclopropane fatty acid synthase family and their contribution to cyclopropane fatty acid synthesis. BMC Plant Biol 2011;11:97. [111] Lu C, Xin Z, Ren Z, Miquel M, Browse J. An enzyme regulating triacylglycerol composition is encoded by the ROD1 gene of Arabidopsis. Proc Nat Acad Sci U S A 2009;106:18837–42. [112] Dahlqvist A, Ståhl U, Lenman M, Banas A, Lee M, Sandager L, et al. Phospholipid:diacylglycerol acyltransferase: an enzyme that catalyzes the acyl-CoA-independent formation of triacylglycerol in yeast and plants. Proc Nat Acad Sci U S A 2000;97:6487–92. [113] Schmid KM, Patterson GW. Distribution of cyclopropenoid fatty acids in malvaceous plant parts. Phytochemistry 1988;27:2831–4. [114] Jiao X, Zhao X, Zhou X-R, Green AG, Fan Y, Wang L, et al. Comparative transcriptomic analysis of developing cotton cotyledons and embryo axis. PloS One 2013;8:e71756. [115] Miquel M. Arabidopsis mutants deficient in polyunsaturated fatty acid synthesis. Biochemical and genetic characterization of a plant oleoylphosphatidylcholine desaturase. J Biol Chem 1992;267:1502–9. [116] Zhang D, Pirtle IL, Park SJ, Nampaisansuk M, Neogi P, Wanjie SW, et al. Identification and expression of a new delta-12 fatty acid desaturase (FAD24) gene in upland cotton and its functional expression in yeast and Arabidopsis thaliana plants. Plant Physiol Biochem 2009;47:462–71. [117] Liu Q, Singh SP, Brubaker CL, Sharp PJ, Green AG, Marhsall DR. Molecular cloning and expression of a cDNA encoding a microsomal w-6 fatty acid desaturase from cotton (Gossypium hirsutum). Funct Plant Biol 1999;26:101–6. [118] Shanklin J, Cahoon EB. Desaturation and related modifications of fatty acids. Ann Rev Plant Biol 1998;49:611–41. [119] Broadwater J, Whittle E, Shanklin J. Desaturation and hydroxylation. Residues 148 and 324 of Arabidopsis FAD2, in addition to substrate chain length, exert a major influence in partitioning of catalytic specificity. J Biol Chem 2002;277:15613–20. [120] Rolletschek H, Borisjuk L, Sánchez-García A, Gotor C, Romero LC, MartínezRivas JM, et al. Temperature-dependent endogenous oxygen concentration regulates microsomal oleate desaturase in developing sunflower seeds. J Exp Bot 2007;58:3171–81. [121] Kinney AJ, Cahoon EB, Hitz WD. Manipulating desaturase activities in transgenic crop plants. Biochem Soc Trans 2002;30:1099. [122] Wilson RF. The role of genomics and biotechnology in achieving global food security for high-oleic vegetable oil. J Oleo Sci 2012;61:357–67.

52

P.J. Horn, K.D. Chapman / Progress in Lipid Research 54 (2014) 32–52

[123] Chapman KD, Neogi PB, Hake KD, Stawska AA, Speed TR, Cotter MQ, et al. Reduced oil accumulation in cottonseeds transformed with a nonfunctional allele of a delta-12 fatty acid desaturase (FAD2). Crop Sci 2008;48:1470–81. [124] Chapman K, Austin-Brown S, Sparace S, Kinney A, Ripp K, Pirtle IL, et al. Transgenic cotton plants with increased seed oleic acid content. J Am Oil Chem Soc 2001;78:941–7. [125] Liu Q, Singh SP, Green AG. High-stearic and high-oleic cottonseed oils produced by hairpin RNA-mediated post-transcriptional gene silencing. Plant Physiol 2002;129:1732–43. [126] Jako C, Kumar A, Wei Y, Zou J, Barton DL, Giblin EM, et al. Seed-specific overexpression of an Arabidopsis cDNA encoding a diacylglycerol acyltransferase enhances seed oil content and seed weight. Plant Physiol 2001;126:861–74. [127] Zhang M, Fan J, Taylor DC, Ohlrogge JB. DGAT1 and PDAT1 acyltransferases have overlapping functions in Arabidopsis triacylglycerol biosynthesis and are essential for normal pollen and seed development. Plant Cell 2009;21:3885–901. [128] Ståhl U, Carlsson AS, Lenman M, Dahlqvist A, Huang B, Banas A, et al. Cloning and Functional characterization of a phospholipid: diacylglycerol acyltransferase from Arabidopsis. Plant Physiol 2004;135:1324–35. [129] Banas´ W, Garcia AS, Banas´ A, Stymne S. Activities of acyl-CoA: diacylglycerol acyltransferase (DGAT) and phospholipid: diacylglycerol acyltransferase (PDAT) in microsomal preparations of developing sunflower and safflower seeds. Planta 2013;6:1–10. [130] Borisjuk L, Rolletschek H, Neuberger T. Surveying the plant’s world by magnetic resonance imaging. Plant J 2012;70:129–46. [131] Neuberger T, Sreenivasulu N, Rokitta M, Rolletschek H, Göbel C, Rutten T, et al. Quantitative imaging of oil storage in developing crop seeds. Plant Biotechnol J 2008;6:31–45. [132] Neuberger T, Rolletschek H, Webb A, Borisjuk L. Non-invasive mapping of lipids in plant tissue using magnetic resonance imaging. In: Armstrong D, editor. Lipidomics. Humana Press; 2009. p. 485–96. [133] Fuchs J, Neuberger T, Rolletschek H, Schiebold S, Nguyen TH, Borisjuk N, et al. A non-invasive platform for imaging and quantifying oil storage in submillimetre tobacco seed. Plant Physiol 2013;161:583–93. [134] Baud S, Lepiniec L. Physiological and developmental regulation of seed oil production. Prog Lipid Res 2010;49:235–49. [135] Bates PD, Browse J. The pathway of triacylglycerol synthesis through phosphatidylcholine in Arabidopsis produces a bottleneck for the accumulation of unusual fatty acids in transgenic seeds. Plant J 2011;68: 387–99. [136] Browse J, Warwick N, Somerville CR, Slack CR. Fluxes through the prokaryotic and eukaryotic pathways of lipid synthesis in the ‘16:3’ plant Arabidopsis thaliana. Biochem J 1986;235:25–31. [137] Bates PD, Durrett TP, Ohlrogge JB, Pollard M. Analysis of acyl fluxes through multiple pathways of triacylglycerol synthesis in developing soybean embryos. Plant Physiol 2009;150:55–72. [138] Mizoi J, Nakamura M, Nishida I. Defects in CTP: phosphrylethanolamine cytidylyltransferase affect embryonic and postembryonic development in Arabidopsis. Plant Cell 2006;18:3370–85. [139] Coulon D, Faure L, Salmon M, Wattelet V, Bessoule J-J. Occurrence, biosynthesis and functions of N-acylphosphatidylethanolamines (NAPE): not just precursors of N-acylethanolamines (NAE). Biochimie 2012;94:75–85. [140] Wang X, Devaiah SP, Zhang W, Welti R. Signaling functions of phosphatidic acid. Prog Lipid Res 2006;45:250–78. [141] Testerink C, Munnik T. Phosphatidic acid: a multifunctional stress signaling lipid in plants. Trends Plant Sci 2005;10:368–75. [142] Jackson SN, Wang H-YJ, Woods AS. In situ structural characterization of phosphatidylcholines in brain tissue using MALDI-MS/MS. J Am Soc Mass Spectrom 2005;16:2052–6. [143] Katagiri T, Ishiyama K, Kato T, Tabata S, Kobayashi M, Shinozaka K. An important role of phosphatidic acid in ABA signaling during germination in Arabidopsis thaliana. Plant J 2005;43:107–17. [144] Villasuso AL, Di Palma MA, Aveldaño M, Pasquaré SJ, Racagni G, Giusto NM, et al. Differences in phosphatidic acid signalling and metabolism between ABA and GA treatments of barley aleurone cells. Plant Physiol Biochem 2013;65:1–8. [145] Cerruti CD, Benabdellah F, Lapre´vote O, Touboul D, Brunelle A. MALDI imaging and structural analysis of rat brain lipid negative ions with 9aminoacridine matrix. Anal Chem 2012;84:2164–71. [146] Wang G, Ryu S, Wang X. Plant phospholipases: an overview. Lipases and phospholipases. Springer; 2012. p. 123–37. [147] Boss WF, Im YJ. Phosphoinositide signaling. Annu Rev Plant Biol 2012;63:409–29. [148] Kim Y, Shanta SR, Zhou L-H, Kim KP. Mass spectrometry based cellular phosphoinositides profiling and phospholipid analysis: a brief review. Exp Mol Med 2010;42:1–11.

[149] Korte AR, Lee YJ. Multiplex mass spectrometric imaging with polarity switching for concurrent acquisition of positive and negative ion images. J Am Soc Mass Spectrom 2013;24:1–7. [150] Kawashima M, Iwamoto N, Kawaguchi-Sakita N, Sugimoto M, Ueno T, Mikami Y, et al. High-resolution imaging mass spectrometry reveals detailed spatial distribution of phosphatidylinositols in human breast cancer. Cancer Sci 2013. http://dx.doi.org/10.1111/cas.12229. [151] Jurchen JC, Rubakhin SS, Sweedler JV. MALDI-MS imaging of features smaller than the size of the laser beam. J Am Soc Mass Spectrom 2005;16:1654–9. [152] Zavalin A, Yang J, Caprioli R. Laser beam filtration for high spatial resolution MALDI imaging mass spectrometry. J Am Soc Mass Spectrom 2013;24:1–4. [153] Kanu AB, Dwivedi P, Tam M, Matz L, Hill HH. Ion mobility-mass spectrometry. J Mass Spectrom 2008;43:1–22. [154] Kliman M, May JC, McLean JA. Lipid analysis and lipidomics by structurally selective ion mobility-mass spectrometry. Biochim Biophys Acta 2011;1811:935–45. [155] Jackson SN, Ugarov M, Post JD, Egan T, Langlais D, Schultz JA, et al. A study of phospholipids by ion mobility TOFMS. J Am Soc Mass Spectrom 2008;19:1655–62. [156] Lietz CB, Gemperline E, Li L. Qualitative and quantitative mass spectrometry imaging of drugs and metabolites. Adv Drug Deliv Rev 2013;65:1074–85. [157] Hattori K, Kajimura M, Hishiki T, Nakanishi T, Kubo A, Nagahata Y, et al. Paradoxical ATP elevation in ischemic penumbra revealed by quantitative imaging mass spectrometry. Antioxid Redox Signal 2010;13:1157–67. [158] Takai N, Tanaka Y, Inazawa K, Saji H. Quantitative analysis of pharmaceutical drug distribution in multiple organs by imaging mass spectrometry. Rapid Commun Mass Spectrom 2012;26:1549–56. [159] Prideaux B, Stoeckli M. Mass spectrometry imaging for drug distribution studies. J Proteomics 2012;75:4999–5013. [160] Nemes P, Vertes A. Ambient mass spectrometry for in vivo local analysis and in situ molecular tissue imaging. TrAC Trends Anal Chem 2012;34:22–34. [161] Sweetlove LJ, Fernie AR. The spatial organization of metabolism within plant cell. Annu Rev Plant Biol 2013;64:723–46. [162] Horn PJ, Chapman KD. Lipidomics in tissues, cells, and subcellular compartments. Plant J 2012;70:69–80. [163] Horn PJ, Ledbetter NR, James CN, Hoffman WD, Case CR, Verbeck GF, et al. Visualization of lipid droplet composition by direct organelle mass spectrometry. J Biol Chem 2011;286:3298–306. [164] van Manen H-J, Kraan YM, Roos D, Otto C. Single-cell Raman and fluorescence microscopy reveal the association of lipid bodies with phagosomes in leukocytes. Proc Natl Acad Sci U S A 2005;102:10159–64. [165] Ozeki Y, Umemura W, Otsuka Y, Satoh S, Hashimoto H, Sumimura K, et al. High-speed molecular spectral imaging of tissue with stimulated Raman scattering. Nat Photonics 2012;6:845–51. [166] Wang B, Karpiouk A, Yeager D, Amirian J, Litovsky S, Smalling R, et al. In vivo intravascular ultrasound-guided photoacoustic imaging of lipid in plaques using an animal model of atherosclerosis. Ultraso Med Biol 2012;38:2098–103. [167] Banerjee B, McKeown KR, Skovan B, Ogram E, Ingram P, Ignatenko N, et al. Ultrasound imaging of the mouse pancreatic duct using lipid microbubbles. SPIE Proceedings 2012. p. 832018. [168] Emmert-Buck MR, Bonner RF, Smith PD, Chuaqui RF, Zhuang Z, Goldstein SR, et al. Laser capture microdissection. Science 1996;274:998–1001. [169] Masyuko R, Lanni EJ, Sweedler JV, Bohn PW. Correlated imaging – a grand challenge in chemical analysis. Analyst 2013;7:1925–39. [170] Li Z, Chu L-Q, Sweedler JV, Bohn PW. Spatial correlation of confocal Raman scattering and secondary ion mass spectrometric molecular images of lignocellulosic materials. Anal Chem 2010;82:2608–11. [171] Dill AL, Eberlin LS, Ifa DR, Cooks RG. Perspectives in imaging using mass spectrometry. Chem Commun 2011;47:2741–6. [172] Thomas B, Akoulitchev AV. Mass spectrometry of RNA. Trends Biochem Sci 2006;31:173–81. [173] Chaurand P. Imaging mass spectrometry of thin tissue sections: a decade of collective efforts. J Proteomics 2012;75:4883–92. [174] de Almeida Engler J, De Groodt R, Van Montagu M, Engler G. In situ hybridization to mRNA of Arabidopsis tissue sections. Methods 2001;23:325–34. [175] Brewer PB, Heisler MG, Hejátko J, Friml J, Benková E. In situ hybridization for mRNA detection in Arabidopsis tissue sections. Nat Protoc 2006;1:1462–7. [176] Nordhoff E, Cramer R, Karas M, Hillenkamp F, Kirpekar F, Kristiansen K, et al. Ion stability of nucleic acids in infrared matrix-assisted laser desorption/ ionization mass spectrometry. Nucleic acids Res 1993;21:3347–57. [177] Douthwaite S, Kirpekar F. Identifying modifications in RNA by MALDI mass spectrometry. In: Jonatha MG, editor. Methods in enzymology. Academic Press; 2007. p. 1–20. [178] Seeley EH, Caprioli RM. Molecular imaging of proteins in tissues by mass spectrometry. Proc Nat Acad Sci U S A 2008;105:18126–31.