|Home | About | Journals | Submit | Contact Us | Français|
Plant development and survival is centered on complex regulatory networks composed of genes, proteins, hormone pathways, metabolites and signaling pathways. The recent advancements in whole genome biology have furthered our understanding of the interactions between these networks. As a result, numerous cell type-specific transcriptome profiles have been generated which elucidated complex gene regulatory networks occurring at the cellular level, many of which were masked during whole organ analysis. Modern technologies have also allowed researchers to generate multiple whole organ metabolite profiles, however, only a limited number have been generated at the level of individual cells. Recent advancements in isolation of individual cell populations have made possible cell type-specific metabolite profiles and allowed enhanced detection and quantification of metabolites that was unavailable when considering the whole organ. The comparison of metabolite and transcriptome profiles from the same cells has been a valuable resource to generate predictions regarding specific metabolite activity and function. In this review, we focus on recent studies that demonstrate the value of cell type-specific transcriptional profiles and their comparison to profiles generated from whole organs. Advancements in the isolation of single cell populations will be highlighted and potential application toward generating detailed metabolic profiles will be discussed.
Plants have long been tractable models to study many biological processes. The recent technological advances in profiling both DNA [especially next-generation sequencing (NGS) and the availability of whole genome sequences] and metabolites (advanced processing software, unbiased peptide identification and higher sensitivity and accuracy in analytical instrumentation) are valuable resources for plant biologists. The vast amounts of data generated by these technologies highlighted the need for a higher resolution in the biological sampling. Many researchers have pointed out the need for omics-based analysis at the level of cell types (see Wang and Bodovitz, 2010 for example). Primarily using the Arabidopsis root, transcript analysis at the cellular level has elucidated many gene networks that were masked at the organ level due to restricted expression (Brady et al., 2011). Multiple cell type-specific transcriptional profiles have been generated (Birnbaum et al., 2003; Brady et al., 2007); however, these profiles represent only one part of a complex biological system. A cell type-specific proteome and metabolome have yet to be generated. A large number of genes and gene families encode the proteins that are required for metabolite production. Despite their importance, metabolites are not well understood and represent a major area of research. Defining the relationship between the transcriptome and metabolome will yield valuable insights into metabolite production, degradation, function and transport. Metabolic profiling may thus be applicable as a highly sensitive and quantifiable phenotyping technology at the cellular level. Functional annotation of genes in Arabidopsis is far from complete, despite being the most extensively studied plant. An omics-based comparison should aid in functional annotation of genes with unknown function and could help refine existing gene networks. Advances in isolating specific cell populations have led to a better understanding of the gene regulatory networks (Iyer-Pascuzzi and Benfey, 2009; Moreno-Risueno et al., 2010; Brady et al., 2011). These techniques can now be used to generate cell type-specific metabolomic data which will further our understanding of both processes.
The sequencing of the Arabidopsis genome has enabled the development of multiple strategies to detect genome-wide transcript expression, referred to as transcriptomics. The oldest and most common method of genome-wide transcript detection involves the hybridization of mRNA populations with a large number of immobilized probes deposited on a surface (i.e. microarrays). Microarrays have been used in several plant species to generate transcriptional profiles for most tissues and many different cell types including in Arabidopsis, maize, rice, barley, soybean, and tomato (Table 1). In Arabidopsis the most frequently used is the ATH1 genome array (Affymetrix), which has been used to generate both whole organ and cell type-specific transcriptome profiles (Redman et al., 2004). One major disadvantage of using microarrays to generate such profiles is that they contain only probes for known genes, which eliminates the possibility of discovering novel expressed genes and limits the more efficient use of this technology to organisms with a sequenced genome, or to sequenced portions of the genome. Another concern is that the first generation microarrays had far less represented genes than the more recent ones, which can make the comparison of results challenging. In addition, the sensitivity of microarrays is reduced by non-specific cross-hybridization, which can alter perceived gene expression (Larkin et al., 2005).
Recent developments in NGS technology, particularly RNAseq, have circumvented some of the limitations of microarrays and considerably expanded our knowledge of the RNA species present in plants (Wang et al., 2010). The general methodology used to perform RNAseq of mRNA includes the following steps: mRNA purification by capturing polyadenylated tails, mRNA fragmentation, reverse-transcription to generate a cDNA library, addition of adaptors to cDNA, amplification by PCR, and sequencing. Results from early RNAseq experiments have already identified many previously unknown expressed transcripts, and have also been used to detect novel non-coding RNA molecules (Lister et al., 2008; Vera et al., 2008). At present, RNAseq raises two main concerns: the variability in the methodologies used to generate cDNA libraries, and the lack of bioinformatic tools available to analyze the results. Despite any current limitations of RNAseq, this technology will undoubtedly lead to further advances in the field of genome-wide analysis.
A distinct transcriptional signature has been generated for many plant organs (root, shoot, leaves etc.), and this whole organ signature represents basic biological processes, which are largely similar among related plant species suggesting conserved functionality. An example of such conserved functionality may be observed by comparing the profile of soybean and Arabidopsis roots and leaves, where the leaves but not the roots of both species are enriched for genes whose products are involved in photosynthesis and CO2 fixation (Vodkin et al., 2004; Schmid et al., 2005). Many whole organ transcriptional profiles have been generated in order to better understand development and to identify networks that respond to environmental stress. The following are two notable examples of such studies.
Plants require nitrogen (N), which is available to crop plants primarily as urea in fertilizers. Spatial localization of N in the soil is very heterogeneous, and therefore roots respond by up-regulating genes involved in ion transport and cellular proliferation to generate new roots to explore N enriched areas. Plants take up N primarily in the forms of ammonium and nitrate, however, the molecular mechanism of uptake and assimilation is still poorly understood. To address this issue, a transcriptional profile was generated in Arabidopsis roots and shoots in response to urea and ammonium nitrate, a common byproduct of urea hydrolysis (Merigout et al., 2008). As expected, the majority of altered genes were associated with transcription and ion transport. Interestingly, this study demonstrated a unique mechanism of N assimilation from urea and ammonium nitrate that is distinct from nitrate. In a second study, a similar transcriptome profile was generated in tomato in response to localized ammonium concentrations in soil (Ruzicka et al., 2010). Surprisingly, in addition to genes responsible for N assimilation and N uptake, genes involved in phosphate uptake were also modulated, suggesting interaction of the N and phosphate uptake pathways. A similar intersection between heat and non-heat stress response pathways has also been observed (Swindell et al., 2007), suggesting that certain stress-responsive genes are shared among multiple networks.
Magnesium (Mg) is another essential macronutrient that plants utilize during development, photosynthesis and nucleic acid folding and catalysis. Despite its importance, very little is known about the genome-wide impact of Mg deprivation, but interestingly the response appears to be species dependent (Hermans and Verbruggen, 2005). To elucidate the molecular response of Mg deprivation in Arabidopsis roots and leaves, transcriptome profiles were generated in response to 4, 8 and 28 hours of Mg deprivation (Hermans et al., 2010). These profiles demonstrated that the roots and leaves differ in numbers and identity of genes responding. One interesting observation was that many of the affected genes were involved in the circadian clock and abscisic acid (ABA) signaling despite having unchanged ABA levels. This suggested that the genes involved in ABA signaling have multiple functions and could indicate a connection between Mg stress response and hormonal signaling. The Mg profile was unique in that unlike other nutrient deprivation profiles, expression of genes involved in Mg uptake was unchanged.
Transcriptomics has provided insights into the regulatory networks controlling organ development across many plant species and has been used to generate many whole organ transcriptional signatures. These signatures have largely been acquired through analyzing mRNA from whole leaves and roots. However, transcript abundance can vary greatly between cell types within an organ and cell specific responses have been observed in response to environmental stimuli (Brandt, 2005). Thus it seemed likely that RNA isolated from single cell types would be more appropriate to identify gene regulatory networks, and as a result, several approaches have been developed to overcome the lack of cell type-specific information acquired from whole organ analysis. Original cell type-specific experiments used RNA from cells grown in vitro which may not accurately reflect the state of a cell (Brandt, 2005); whereas more recent methods have focused on isolating single populations of cells directly from the plant (for additional information, see (Long, 2011)). Metabolite levels are similar to transcript profiles in that they show great variation between individual cell types. As a result, multiple strategies have been developed that use mass spectrometry imaging (MSI) to generate metabolite profiles at single cell resolution in plants, including matrix assisted laser desorption ionization MSI (MALDI-MSI) (Kaspar et al., 2011) and laser desorption/ionization MSI (LDI-MSI) (Cha et al., 2009; Holscher et al., 2009). Four unique methods of isolating single populations of cells for transcriptional profiling are briefly discussed below.
One of the first methods developed to isolate cellular RNA was LCM. This technology involves the use of a microscope to select specific cells from a fixed tissue, based on morphology or histology (Asano et al., 2002). This procedure can be laborious and time consuming because 500-1000 cells have to be individually selected for transcript analysis. However LCM has the advantage of not requiring plant transformation in cell types that can be easily identified, and therefore has been used to profile multiple cell types in the root and shoot of many plants (Asano et al., 2002; Kerk et al., 2003; Nakazono et al., 2003; Woll et al., 2005; Jiao et al., 2009). In one example, LCM was used to generate transcriptional profiles of the apical meristem, quiescent center and root cap of maize primary roots which revealed genes up-regulated in the root cap that are linked to metabolism (Jiang et al., 2007).
As an alternative to LCM that would allow high throughput transcript analysis at the cellular level, a technology was developed to rapidly sort marked cell populations after protoplasting using FACS (Birnbaum et al., 2003). This technique makes use of transgenic lines that express a fluorescent marker in a cell type-specific manner and can easily be scaled to isolate RNA from limited cell populations such as the quiescent center (Nawy et al., 2005). The resulting protoplasts can be used for downstream analysis such as transcriptional profiling. The major disadvantages are that cell walls are removed during the process of protoplasting, excluding the related metabolites and proteins, and the disruption of the cell-cell communication by protoplasting, which might alter the transcriptional profile of the cells, especially under stress. However, it has been demonstrated that except for a small subset of genes, which are activated upon protoplasting, the overall global gene profile is largely unchanged (Birnbaum et al., 2003). The need to generate specialized transgenic lines for each cell type of interest makes this approach only applicable to easily transformable species. To date, protoplasting followed by FACS is the most popular approach used to isolate plant cell types because of the versatile analyses that can follow such isolation: at RNA, protein and metabolite levels. FACS has already been used extensively to generate cell type-specific transcriptional profiles of all cell types of the Arabidopsis root under both stressed and non-stressed conditions and is currently being used to generate proteomic and metabolite profiles for comparison. (Birnbaum et al., 2003; Birnbaum et al., 2005; Nawy et al., 2005; Brady et al., 2007; Dinneny et al., 2008; Gifford et al., 2008).
Both LCM and FACS require harsh treatments to separate the cells and often have low yield of cell types. As an alternative, INTACT was developed and used to profile Arabidopsis root epidermal cells (Deal and Henikoff, 2010; Deal and Henikoff, 2011). The INTACT method requires that the nuclei of specific cell types be tagged using transgenic lines expressing a biotinylated nuclear envelope protein. The tagged nuclei are affinity isolated and can be used for genome-wide expression and chromatin profiling. The major advantages of INTACT are the high yield and purity of nuclei from a desired cell type and lack of requirement of special equipment. The major disadvantage is the recovery of only nuclei, which prevents proteomic or metabolic analysis, and the need to generate specialized transgenic lines for each cell type of interest. In addition, experiments comparing cytoplasmic and nuclear RNA samples have demonstrated significant transcript differences between the two compartments which are predicted to control key cellular processes (Cheng et al., 2005; Barthelson et al., 2007).
To monitor transcript function, a method was developed that uses the expression of a tagged ribosomal protein under the control of a tissue-specific promoter. Immunopurification from transgenic plants expressing this protein selects for mRNA in the ribosome complex and presumably actively being translated, thus generating the so-called translatome (Zanetti et al., 2005). This method of detection can be extended to tissue- and cell-specific analysis by choosing appropriate promoters to drive ribosomal protein expression. One advantage of this technique is the certainty of expression at the time of data collection; however, it does limit analysis to polysomes and exposure to certain stresses has been shown to repress ribosome biogenesis (Dinneny et al., 2008).
High-resolution analysis of cell type-specific gene expression can provide valuable information about the complexity of transcriptional programs (Table 1). Comparison of the transcriptional profile of an organ to the cellular expression profile of that same organ has highlighted the limitations of performing the analysis at the organ level (Brady et al., 2007). High-resolution profiles of specific cell types using both FACS and LCM have revealed additional information about organs that was masked when examining the entire organ. For example, cell type-specific expression analyses using FACS compared to expression in the whole root of Arabidopsis identified several hundred additional genes that were absent from whole root profiles and demonstrated the importance of considering gene expression at the cellular level (Schmid et al., 2005; Iyer-Pascuzzi et al., 2011). This discrepancy between whole organ and cell type expression is not limited to the root. In the leaf, separation of guard cells from mesophyll cells revealed the expression of guard cell specific ABA responsive genes (Leonhardt et al., 2004). In combination with reverse genetics, this cell type-specific expression data allowed researchers to identify the function of a previously unknown guard cell-specific ABA responsive gene. In a third example, cells that were obtained by LCM from the epidermis and vascular tissues of coleoptiles from maize seedlings identified genes differentially expressed in epidermis and vascular tissues. Genes involved in secondary metabolism were epidermis-specific, whereas, genes encoding transporters and metal binding proteins were vascular-specific (Nakazono et al., 2003). These cell type-specific profiles and the examples that follow have identified physiological mechanisms that are cell type-specific within an organ, which were not observed by whole organ transcriptomics confirming the need for high resolution cellular profiles.
Cell type-specific spatiotemporal regulation of transcription is mediated through cell-cell communication and is required for development, cell and tissue identity and stem cell maintenance. The shoot apical meristem (SAM) has well characterized cell-cell interactions required for stem cell maintenance, which allows aerial growth of the plant. SAM gene expression at the cellular level has uncovered individual spatial domains of the stem cell niche and gene networks specific to the SAM (Yadav et al., 2009). This expression map was used as a tool to predict expression patterns of genes with unknown function and identified DNA repair and chromatin modification pathways that suggested maintenance of genome stability is crucial for stem cell function.
The plant root is an attractive model organ. It is transparent and easily imaged, and especially suitable for developmental studies due to its relative simplicity, radial organization and iterative mode of growth. It is also involved in many biological processes. It provides support for the aerial part of the plant, serves as the site of nutrient and water absorption and has a major role in plant stress response. The Arabidopsis root system consists of a single primary root with multiple lateral roots. The root is composed of concentric cylinders of tissue that are generated from a small group of stem cells at the root tip (Figure 1a). These stem cells divide to generate 7 major cell types that are constrained within cell files that extend the length of the root. As new cells are generated, they displace the older cells so that the oldest cells are furthest from the tip. As a method of increasing root area and to expand soil exploration, lateral roots are generated from primary roots by initiating cell division in select pericycle cells. The current model is that pericycle cells dedifferentiate prior to lateral root formation; however, recent cell type-specific transcriptome studies in maize have challenged this model. A comparison of wild-type pericycle cells competent to form lateral roots and a rum1 mutant that lacks lateral roots identified a group of genes that potentially control lateral root initiation (Woll et al., 2005). An additional transcriptional profile of pericycle cells during specification compared to specified non-pericycle cells has been generated and showed little overlap (1 gene) with the profile generated during lateral root initiation (Dembinsky et al., 2007). These data suggest that the molecular networks involved in pericycle specification and lateral root initiation are unique, supporting an idea proposed by Dubrovsky and Ivanov (1984) that lateral root forming pericycle cells remain undifferentiated until the time of initiation, and thus they do not dedifferentiate.
Root hairs extend from the epidermal layer and aid in nutrient and water uptake in addition to serving as the site of interaction between the root and surrounding rhizosphere. Leguminous plants such as soybeans, overcome low N levels in the soil by symbiotic interactions between plant root hairs and rhizobia. This symbiotic interaction requires multiple biological processes, which are largely uncharacterized at a molecular level. Despite several whole root transcriptome studies, less than 20 genes have been shown to be involved in this interaction (Lohar et al., 2006; Hogslund et al., 2009). As a method to further understand the root hair-rhizobia interaction, an assay was developed to isolate root hairs from the root (Libault et al., 2010). These root hair-specific transcriptional studies identified 1973 genes differentially expressed in response to rhizobial infection, which include transcription factors, signaling molecules and genes involved in hormone synthesis. The results of this study suggest a genome-wide response to rhizobial infection and will aid in future exploration and crop improvement.
To fully understand and generate developmental regulatory networks, information about gene expression in both developmental time and space is required. Analysis of gene expression in both space and across developmental time was performed using Arabidopsis (Brady et al., 2007) and rice roots (Takehisa et al., 2011). Brady and colleagues (2007) used Arabidopsis to generate a transcriptional profile of 15 cell types across 13 longitudinal zones (representing developmental time). Analysis of these profiles demonstrated dominant patterns of gene expression at specific developmental time points and uncovered cell type-specific gene expression patterns. Surprisingly many genes showed fluctuations of expression over developmental time such that they were turned on early during development, then turned off, and then turned back on later in development. The spatiotemporal gene expression generated in this study has led to the prediction of previously unknown cellular functions and transcriptional networks (Brady et al., 2011). Takehisa and colleagues (2011) generated a similar profile in rice by analyzing 3 cell types across 8 longitudinal zones. Similar to Arabidopsis, cell type-specific and developmental time point-specific transcriptome signatures were observed. The expression of genes involved in phytohormone biosynthesis and signaling was compared among cell types and across developmental time revealing novel insights into hormone interactions in rice roots. Expression profiles also refined the locations of specific nutrient transporters to specific cell types and developmental zones within the root and identified cell type-specific transporters responsible for movement of nutrients inward towards the vasculature.
Plant growth and development do not strictly adhere to a predefined program and are directly influenced by the environment. As a result, plants have developed multiple mechanisms to deal with external stimuli. Early studies of this phenomenon were focused on the macro level; however, the sequencing of the Arabidopsis genome made it possible to analyze changes at the molecular level. The initial experiments dealt with the phenotypic characterization of mutants having an abnormal response to a specific stress, followed by a laborious mapping process to identify the gene or genes responsible for the phenotype. This approach successfully identified multiple environmental response genes; however, it was limited to analyzing single genes and made it difficult to define gene networks involved in the stress response. Advances in transcriptomics and metabolomics have made possible the generation of genome-wide transcript and metabolite profiles in response to many abiotic stresses and have uncovered multiple stress-specific regulatory networks (for reviews of abiotic stress response networks see (Hirayama and Shinozaki, 2010; Urano et al., 2010)).
The Arabidopsis root provides an informative model system for the discovery of stress response networks because of its simple architecture and cellular organization. Hence, many efforts have focused on understanding how the Arabidopsis root perceives and responds to the soil environment using whole roots (Misson et al., 2005; Kilian et al., 2007; Van Hoewyk et al., 2008; Zeller et al., 2009). These observations have demonstrated that nutrient responsive genes impact gene expression and metabolite production in all parts of the plant as a mechanism to deal with oxidative damage and nutrient deprivation. By examining whole root responses, many important discoveries were made regarding metabolite biosynthesis, stress specific transcriptional responses and demonstrated many general stress response pathways; however, more recent experiments have shown the value of cell type-specific profiling.
Transcriptional profiling of individual cell types within the root demonstrated that many of the stress responses occur at the cellular level, which was obscured when examining the whole root (Dinneny et al., 2008; Gifford et al., 2008). To gain a comprehensive view as to how individual tissues respond to environmental stimuli, Dinneny and colleagues (2008) profiled 6 different cell types across 4 longitudinal zones for changes in gene expression in response to salt exposure and iron deficiency. Of the nearly 4000 cell type-specific genes that were differentially expressed after salt exposure and 1300 genes after iron deficiency, the majority of these were changed in only one cell type, suggesting that a large part of stress regulation occurs at the cell or tissue level. Analysis of developmental zone-specific gene sets for each stress indicated that many genes were enriched in only one zone. Comparison of the cell type-specific and longitudinal data demonstrated that most of the enriched genes are longitudinal zone-specific, suggesting that salt and iron stress regulate processes on the basis of developmental context, in addition to cell type. A comparison of the cell type-specific gene sets showed only 20% overlap between salt and iron responsive genes, indicating that the gene response is stress specific but also likely contains a common stress response. Additional studies have shown similar cell type-specific responses to N, phosphorous, pH and sulfur (S) stress (Gifford et al., 2008; Iyer-Pascuzzi et al., 2011). A more recent comparison between the transcriptional profiles generated from whole roots and individual cell types during stress identified a common stress response in the whole root; however, evidence was lacking for a universal stress response (Iyer-Pascuzzi et al., 2011). Although individual stresses uniquely impact the transcriptional profile, a cell type-specific stress response exists independent from the type of stress. This analysis also revealed that ABA signaling, which Dinneny and colleagues (2008) previously regarded as a general stress response, regulates a different set of genes depending on the stress and cell type. Collectively, these results suggest that the gene response observed in whole roots is comprised largely of multiple individual cellular responses. The exact function of the stress response genes remains largely unknown, but mutant analyses suggest that they are directly linked to gene regulatory networks controlling growth and development.
The exact mechanism of gene induction during stress is also unclear, but likely involves a signal transduction pathway. Calcium is an important second messenger that has been associated with multiple stress responses. For example, salt stress elicits an increase in calcium that activates an anti-porter to promote export of sodium. When the calcium response was examined in different cells, researchers were unable to determine which cells contribute to this response (Knight et al., 1997). Further work that examined calcium dynamics at the cellular level using a combination of YFP-aequorin (calcium reporter protein) gene fusion under control of UAS and cell type-specific GAL4 enhancer trap lines to direct cell type-specific expression, demonstrated a peak of calcium response in the epidermis and endodermis of the root in response to salt stress (Kiegle et al., 2000). The authors also noted that the response varied depending upon the stress; drought stress was epidermis specific while cold stress was uniform in all cell-types. Another common signaling mechanism is redox regulation involving reactive oxygen species (ROS), which are key factors in many cellular activities and signal transduction pathways, making them good candidates as modifiers of the stress response. In agreement with this, ROS are associated with the activation of several genes and proteins including defense related genes, mitogen activated protein kinases (MAPKs) that transmit cellular responses to external stimuli, multiple stress-related genes and genes that control plant development (Desikan et al., 2001; Shin and Schachtman, 2004; Tsukagoshi et al., 2010). Under conditions of potassium, N and phosphorous stress, ROS levels are increased in a cell type-specific manner, that is stress dependent (Shin et al., 2005). Under potassium and N starvation, ROS accumulates primarily in the epidermis; under phosphorous stress, ROS accumulates in the cortex. In most studies, however, it is currently unclear whether ROS is driving cell type-specific gene expression or cell type-specific expression results in ROS accumulation.
The leaf epidermis is composed of pavement cells, border cells, guard cells, trichomes (leaf hairs) and trichome socket cells, each with its own specialized function. Because the epidermis is exposed to the surrounding environment, it is the first line of defense against environmental stress, including herbivores and pathogens. The epidermis layer synthesizes the building blocks for the construction of the cuticle, an extracellular layer largely composed of the polymer cutin and waxes. To identify proteins involved in the biosynthesis of waxes and cutin, Suh et al. (2005) manually dissected epidermal peels from Arabidopsis and determined transcript profiles in both rapidly expanding and nonexpanding cells at different positions along the inflorescence stem (Suh et al., 2005). Known epidermis-specific genes were correctly classified and 15% of the transcripts preferentially expressed in the epidermis were enriched in genes encoding proteins predicted to be membrane associated and involved in lipid metabolism. Several recent studies followed the transcriptome, proteome and metabolome of the epidermis layer in fleshy fruit, particularly in tomato. For example, manually dissected exocarp (i.e. the outer layer including the epidermis) and mesocarp (i.e. flesh, the inner layer) tissues were used for both transcriptome and metabolite profiling and hundreds of genes and metabolites were identified that were enriched in the outer layer (Lemaire-Chamley et al., 2005; Mintz-Oron et al., 2008). In melon, an extraordinary range of complementary analytical technologies were used to profile the metabolome, volatiles and mineral elements in fruit at a number of time points during the final ripening process and tissues collected across the fruit flesh from rind to the seed cavity (Moing et al., 2011). In this extensive study, approximately 2000 metabolite signatures and 15 mineral elements were determined in an assessment of temporal and spatial melon fruit development. Very recently, LCM was combined with NGS to characterize the transcriptomes of the five principal tissues of tomato fruit including the outer and inner epidermal layers, collenchyma, parenchyma, and vascular tissues. More than 20,000 expressed unigenes were identified, including a large number that displayed cell type specific or distinct expression patterns in specific tissues (Matas et al., 2011).
The trichomes are large single cells extending from the epidermis which are capable of synthesis, storage and secretion of multiple secondary metabolites that aid in defense, including, terpenes, acyl sugars, phenylpropanoids and flavonoids. These metabolites are of interest to humans because they are often used as food flavors, perfumes and as pharmaceuticals. There are two major types of trichomes, nonglandular and glandular. Arabidopsis contains only nonglandular trichomes where as other species such as Solanum (tomato) possess both glandular and nonglandular trichomes. Among the glandular trichomes, there are several predominate types dependent upon the plant species. Trichomes are extensively used as a model system to address questions regarding developmental patterning and to study metabolite biosynthesis because of their location on the external surface of the plant. Because they represent single cells, they have been used by many researchers to generate cell type-specific transcriptional and metabolite profiles. The major conclusions from some of these studies are discussed below.
Several studies have used transcriptional profiles to try and elucidate the biochemical pathways involved in secondary metabolite production. Although enzyme transcript levels do not always correlate with metabolite levels, they are suggestive of metabolite production. Using Arabidopsis, the comparison of transcriptional profiles generated from trichomes and whole leaves from two trichomeless mutants yielded 3231 genes that appeared to be specific to trichomes (Jakoby et al., 2008). Among the up-regulated genes were multiple genes involved in biosynthesis pathways that generate secondary metabolites. Trichomes and root hairs were previously known to share a transcriptional network; however the similarity of downstream targets was largely unknown. Comparison of the trichome and root hair expression profiles showed largely overlapping genes, strongly suggesting they share downstream targets. Interestingly, the pathways involved in secondary metabolite synthesis were specific to each cell type. Metabolite profiles in glandular trichomes isolated from four distinct sweet basil lines revealed an array of phenylpropanoids and terpenoid derived compounds that were line-dependent (Xie et al., 2008). Comparison of the enzyme expression in the phenylpropanoid and terpenoid biosynthesis pathways also demonstrated line-dependent expression profiles. However, the enzyme profile was not always indicative of the metabolite profile. A transcriptional profile in Catharanthus roseus was generated from the cells of the epidermis using carborundum abrasion in an attempt to identify components of the monoterpenoid indole alkaloid (MIA) biosynthesis pathway, which are key components of several anticancer drugs (Murata et al., 2008). This profile, termed the epidermome, identified several novel MIA pathway genes in the epidermis that were not present in the public Catharanthus EST database, in addition to genes involved in triterpene and flavonoid biosynthesis. The leaf epidermome demonstrates that a single layer of epidermal tissue produces multiple types of metabolites. However, it does not distinguish which cell type is responsible for production of these metabolites.
Based upon function, Solanum glandular trichomes are classified as either secreting (type I and IV) or storage (type VI). Recent experiments profiling glandular trichomes isolated from Solanum habrochaites leaves demonstrated that all three trichome types contain multiple methylated forms of myricetin, with the tetramethylated form predominating (Schmidt et al., 2011). Myricetin flavonoid is highly methylated and has been reported in a variety of plants but the O-methyltransferases (OMTs) responsible for their biosynthesis were unknown until recently. Using transcriptional profiling, Schmidt and colleagues (2011) identified two genes enriched in the trichomes that encode enzymes capable of methylating myricetin, with the highest level being found in the type I and IV trichomes. Experiments are underway to confirm that these genes function as OMTs. Comparison of transcriptional and metabolite profiles generated from glandular secretory trichomes (type I, IV, VI and VII) in the leaves of five species of Solanum, found that type I and type IV glandular trichomes are very similar at the transcript level (McDowell et al., 2011). The authors also noted that the metabolite levels, but not the metabolite profiles, were different between type I and IV and type VI, and that type VII has limited metabolite synthesis and storage compared to other types. They also noted groups of compounds that were specific to certain trichome types and that were not evenly distributed across the various species. In agreement with previous observations (Harada et al., 2010), several trichome types expressed carbon fixation genes, suggesting trichomes may serve as a site of carbon fixation to be used in secondary metabolism.
The sesquiterpene class of terpenoids are synthesized via modification of a prenyl diphosphate intermediate by terpene synthases (TPSs). A combination of transcriptome and metabolome data has been successful in identifying a sesquiterpene synthase that results in production of β-caryophyllene and α-humulene specifically in the type VI glands of the leaf but not in the stem of tomato (Schilmiller et al., 2010). In a separate study, a comparison of transcriptional profiles generated from Solanum lycopersicum and Solanum habrochaites stem trichomes was performed to identify TPSs, specifically those TPSs involved in production of sesquiterpenes (Bleeker et al., 2011). The authors found 7 synthases expressed in Solanum lycopersicum and 6 in Solanum habrochaites, one of which was induced by jasmonic acid, suggesting its involvement in herbivore-stress response. The majority of the TPSs that were enriched were sesquiterpene synthases and many of these had been demonstrated to produce multiple sesquiterpenes. TPSs for many of the predominant sesquiterpenes were not identified from these data, further supporting the idea that a single TPS is likely responsible for producing more than one sesquiterpene.
It is likely that complex networks coordinate metabolic pathways within the plant during development and in response to external stimuli. Despite the recent advances in metabolomics technologies and the consequent increase in the amount of metabolites examined per study, transcriptome data is still substantially more comprehensive. Thus, numerous studies at the transcriptome level use gene expression data to form hypotheses regarding the activity of metabolic pathways and metabolite production (Yonekura-Sakakibara et al., 2008). At this stage it is unclear to what extent the correlation between transcription and metabolite levels holds true. The problem is even more severe when analyzing whole organs that are composed of many cell types as a correlation might be an average of all layers, but this could be far from the actual situation in specific cell types. In fact, most metabolic studies to date were carried out at the entire organ level and only a few in isolated single cell layers (as in the trichome studies discussed earlier). Cellular resolution should allow enhanced metabolite detection and quantification and will likely uncover cell-cell signaling networks.
To perform research at the cellular level, techniques routinely used to generate cell type-specific transcriptional profiles are now being adapted and modified for generating metabolic profiles in specific cell types. In a recent pioneering study, a combination of FACS and gas chromatograph - time of flight - mass spectrometry (GC-TOF-MS) were employed to directly quantify auxin (IAA) in specific cell types of the Arabidopsis root (Petersson et al., 2009). Data on IAA distribution in 14 different fluorescent marker lines was combined to generate a map of auxin in the root. The authors identified an auxin gradient throughout the root and an auxin maximum at the root apex. The cells of the root apex were separated and showed equal rates of auxin biosynthesis despite unequal auxin distribution, suggesting that directional transport is cell type-dependent. Furthermore, it appeared that cell type-dependent distribution of auxin is highly correlated with the cell type-specific transcriptional profiles previously reported. This study demonstrated the feasibility of FACS-based metabolic profiling at cell type resolution, and that the procedures of protoplasting and sorting do not affect the levels of auxin, suggesting that this technology may prove to be a useful tool in the metabolic cell type-specific mapping of organs.
In recent years, hyphenated Liquid-Chromatography (LC) and high-resolution MS systems [e.g. LC-quadropole-Time-Of-Flight (qTOF)- MS and Fourier Transform (FT)- MS] have been widely exploited for metabolomics studies (Aharoni et al., 2002; De Vos et al., 2007; Malitsky et al., 2008; Giavalisco et al., 2009; Allwood and Goodacre, 2010). The use of such platforms provides several advantages, primarily, the ability to conduct a certain level of structural elucidation of the peaks with unknown chemistry (not unambiguous in most cases). In plant extracts, a recently established mode of analysis using these analytical tools allows the detection and putative identification of hundreds of semi-polar compounds (mostly secondary/specialized metabolites) in a single instrument run. Pilot experiments to examine the possibility of using LC-MS-based metabolomics analyses in FACS sorted cell types demonstrated that such experiments are indeed feasible. Extracts from FACS sorted cells derived from 5 different Arabidopsis marker lines were analyzed using high resolution LC-MS. These experiments raised major technical issues while at the same time providing the first insights into metabolism at the cell type level in the Arabidopsis root. One major technical concern was that a large number of cells needed to be analyzed. In the particular set-up employed, approximately 700,000 cells were used to generate metabolic profiles. Collecting this number of cells per sample is laborious and time consuming using the established FACS sorting method. Improvements in specific cell type isolation efficiency together with an increase in sensitivity of the instrumentation are required in order to make this technology robust. The most abundantly detected components in these experiments appeared to be the sulfur-containing glucosinolates (GLSs). Localization of the active GLSs in the Arabidopsis root is probably complex as they are not biologically active until hydrolysis of the thioglucose bond in response to tissue damage or stress which is catalyzed by the myrosinase enzyme located in a different cell type (Wentzell et al., 2008). Analysis of the transcriptome indicated that, as with other abiotic stresses, there was a considerable amount of cell type-specific response to the lack of sulfur (Iyer-Pascuzzi et al., 2011). Because GLSs require sulfur for their biosynthesis and are distributed in a cell type-specific manner, a comparison between GLS accumulation under sulfur depletion conditions with normal growth conditions, at the resolution of individual cell types could be informative. Indeed, additional experiments to the ones described above in which GLSs were monitored in a targeted manner throughout the 5 root layers under S-limiting conditions displayed interesting trends in terms of their accumulation. Members of a different class of secondary metabolites, the flavonols, produced via the phenylpropanoid pathway, were also detected in these experiments. Figure 2 presents the accumulation pattern of flavonol glycosides in the 5 different cell types of the Arabidopsis roots, predominantly in the cortex. The genes involved in flavonoid biosynthesis are well documented; therefore, cell type-specific expression of these genes was determined using previously generated cell type-specific transcriptional profiles. All of the known genes in the anthocyanin and flavonol pathways were enriched in the root cortex cells (Figure 1b). Flavonoids are further modified via glycosylation, acylation and methylation resulting in diverse chemical and metabolite profiles. Several of the known genes encoding glycosyltransferases involved in this pathway were also enriched in the cortex. This data suggests that the cortex is an important site of flavonoid biosynthesis and subsequent modification.
The value of analyzing individual cell types rather than whole organs has been demonstrated by multiple transcriptome studies. In addition to uncovering novel genes controlling development and stress response, these cell type-specific transcriptome studies demonstrated cell type-specific expression of many proteins involved in secondary metabolism. Therefore, it is predicted that cell type-specific metabolite analysis is also warranted and will yield novel insights into metabolite synthesis and storage. The value of focusing on single cells was recently bolstered by observing equal rates of auxin synthesis in root apex cells yet uneven auxin distribution. Due to the large amounts of metabolites produced by the plant, it is probable that metabolite synthesis and distribution are also not directly proportional. The generation of cell type-specific metabolite profiles will further expand and deepen our understanding about how and where metabolites are produced, localized and move.
Work in the Benfey lab in this area is funded by the NIH and by the NSF AT2010 program. A.A. is the incumbent of the Adolpho and Evelyn Blum Career Development Chair of Cancer Research. The work in the Aharoni laboratory was supported by the European Research Council (ERC) project SAMIT (FP7 program). A.M. was supported by the Weizmann institute faculty of Biochemistry dean fellowship.