Integrative systems biology: an attempt to describe a simple weed

Abstract

Genome-scale studies hold great promise for revealing novel plant biology. Because of the complexity of these techniques, numerous considerations need to be made before embarking on a study. Here we focus on the Arabidopsis model system because of the wealth of available genome-scale data. Many approaches are available that provide genome-scale information regarding the state of a given organism (e.g. genomics, epigenomics, transcriptomics, proteomics, metabolomics interactomics, ionomics, phenomics, etc.). Integration of all of these types of data will be necessary for a comprehensive description of Arabidopsis. In this review we propose that ‘triangulation’ among transcriptomics, proteomics and metabolomics is a meaningful approach for beginning this integrative analysis and uncovering a systems level perspective of Arabidopsis biology.

Introduction

The completion of the Arabidopsis genome sequence facilitated extraordinary progress toward understanding plant biology. In particular, complete genomic sequence data drove the development of genome-wide transcriptional approaches, such as microarrays. Genome-scale studies (hereafter -omics) include, but are not limited to, analysis of RNAs, proteins, and metabolites. Significant progress has been made annotating and determining the function of many Arabidopsis genes. However, a plant is not just the sum of its genes, but a complex system where gene product interactions result in emergent properties. Therefore, with the ultimate intention of studying the biology of the whole organism, it is important to frame the next long-term goals for plant scientists.

We propose that a key long-term goal is the integration of different genome-scale approaches. The first steps in this direction have already occurred although the tools for the integration, visualization, and modeling of -omics data are still at a relatively early stage [(for reviews, see 1,2)]. In this review, we focus on gene expression, protein and metabolite profiling data, briefly introducing each of the individual approaches, and then highlighting recent efforts to integrate these -omics (Figure 1).

Figure 1.

The biology of the whole organism: integration of different -omics. A simplified schematic representation of -omics. Transcriptomics, proteomics and metabolomics are measured comprehensively by genome-scale methods. The integration of these -omics (as shown by the intersecting Venn-diagram) provides insight into systems-level understanding of Arabidopsis.

Omics technology development: transcriptomics, proteomics and metabolomics

Transcriptomics

The completion of the Arabidopsis genome sequence has facilitated whole genome transcriptomic (gene expression) studies. The development of microarray technology enabled the simultaneous examination of thousands of genes; thus providing a comprehensive view of gene activity. Microarray gene expression data now cover organs, tissues, cell-types and developmental events, as well as responses to a variety of environmental perturbations [3–5,6^••,7,8]. For profiling transcripts in model organisms with well-annotated genomes such as Arabidopsis, microarrays have played an invaluable role in our understanding of plant gene expression. Despite the power of microarrays, they are limited to providing relative abundance information about identified genes and gene models. Therefore, deep sequencing of transcripts (RNA-seq) provides an alternative to microarray technology [9,10]. Additionally, because RNA-seq does not depend on genome annotation, RNA-seq has emerged as the method of choice for transcriptional profiling in nonmodel organisms. RNA-seq approaches aim to detect diverse RNA molecules, including mRNA, noncoding RNA and small RNAs [11,12]. The unparalleled ability of RNA-seq to provide sequence information at single basepair-resolution enables the identification of novel genes, alternative-splicing, single nucleotide polymorphisms, and transcript abundance upon DNA methylation-state modification [13,14^•,15]. Recently RNA-seq in specific cell types and developmental regions of the Arabidopsis root detected over 60 novel miRNAs [16^••]. As a new technology, there are unique challenges that come with analyzing RNA-seq data including developing methods, algorithms and pipelines (e.g. library preparation procedures, RNA quantification, isoform detection and quantification, etc.). Despite these challenges, and because of the improved throughput and lower cost, RNA-seq has already shed light on the complexity and regulation of the plant transcriptome.

Proteomics

The Arabidopsis genome sequence enabled the prediction of genes and the proteins they encode. Comprehensive proteomic analysis seeks to determine the localization, quantity and post-translational modifications of all proteins in an organism. This information is complementary to transcriptomic analyses as it provides the functional readout of gene expression profiles. However, proteomic analysis has been more challenging than transcriptomic analysis for a number of reasons described below.

As generally practiced, proteomics first detects peptides and then assigns them to a gene model [(see 17)]. When dealing with complex samples, protein representation can be biased, with an overrepresentation of large or abundant proteins compared to small proteins [18^••]. Additionally, information about protein accumulation and posttranslational modifications are required to fully understand a plant as a whole. One could envision that the in vivo protein concentrations could be measured by targeting specific cell-types and tissues. The Arabidopsis root offers an ideal system for proteomic analysis at cellular resolution comparable to what has been achieved for transcriptional analysis. Technological advances should improve detection and identification issues, enabling complete proteome analysis in the future.

There are multiple approaches for proteomic profiling [(for a review, see 20)]. The traditional approaches are gel-based such as SDS-PAGE, which is useful for protein ‘fingerprinting’ of complex extracts for protein quantities and post-translational modifications [19,21]. Advances in mass spectrometry (MS) measurements have enabled protein quantification from complex samples. Shotgun proteomics, which combines liquid chromatography and MS, has emerged as a promising method for comprehensive proteomics [22,23]. MS-based proteomic studies in plants have focused on identification of proteins in various organelles (e.g. plastids and mitochondria) in an effort to reduce protein complexity [20,24–27]. Quantitative proteomics encompassing whole plants requires analysis of complex samples containing many unique protein species. A recent proteome survey in Arabidopsis identified almost 50% of predicted protein-coding genes by tandem MS [18^••]. This group found over 50 novel or alternative gene models highlighting the utility of this approach to identify known and novel genes through high throughput proteomic methods.

A protein-interaction map, using a yeast-2-hybrid system, was generated for Arabidopsis providing interaction information for ~8000 protein coding genes [28]. This map has already been used to gain insight into the response of plants to pathogen attack [29^•]. These studies demonstrate the utility of these large-scale projects. As detection techniques improve, more proteins and protein complexes will be profiled enhancing our knowledge of protein localization, abundance and interactions.

Metabolomics

The objective of metabolomics is to identify and quantify all metabolites in plants. Metabolomics is challenging in part because of the vast range of compounds found in plants; a single accession of Arabidopsis contains more than 5000 metabolites [30]. Because a major effort is needed for unequivocal identification of metabolites and no single approach can detect all compounds, combinations of different and complementary extraction and detection techniques are necessary to increase the coverage of a metabolome [31–33]. Keeping in mind that we would like to detect the complete repertoire of metabolites in a cell and understand how different metabolic pathways are coordinated across the entire organism, state-of-the-art metabolomic techniques are necessary. Metabolic profiles of different organs, tissues and even cell-types will provide greater insight into plant complexity. Arabidopsis, and especially the Arabidopsis root, provides an excellent system to start such metabolomic studies.

To cover the vast range of the metabolome, one of the most promising approaches is the combination of gas chromatography (GC — for primary metabolites), liquid chromatography (LC — for secondary metabolites) or capillary electrophoresis (CE — for ionic metabolites) with MS or metabolic fingerprinting using nuclear magnetic resonance (NMR) [34–38]. Metabolic fingerprinting techniques using NMR or Fourier transform-infrared spectroscopy (FT-IR) can also be used to investigate the dynamics of a metabolic network. If the primary goal is to have a better efficiency in sample separation, two-dimensional GC can then be combined with fast acquisition of rate mass spectroscopy (GC × GC–TOF-MS) [39].

The combination of these techniques has generated a vast amount of metabolomics data, which must be properly annotated and tracked to yield fruitful results. Although integration of data from multiple research groups has been difficult, efforts to construct common data repositories [40–42] and data analysis software are ongoing [1,43].

Data integration from molecular information: the gene–protein–metabolite relationship

Transcriptomics–proteomics

As gene expression profiling and proteomic methods improve, data can be combined to achieve a better understanding of Arabidopsis as a system. One important aspect of future investigation will involve the identification of variably spliced transcripts and the discrete proteins they encode. Before next generation sequencing analysis, estimates of alternatively spliced genes based on EST analysis ranged from 22 to 30% [44–47]. More recent information using RNA-seq predicts that 42% of Arabidopsis genes with introns are alternatively spliced [14^•]. Questions remain regarding the functional significance of these isoforms (e.g. what percentage are expressed as proteins?). These questions will soon be addressable by examining the full proteome and comparing it to next-generation gene expression data.

Several groups have used gene expression and proteomic data together to gain information unattainable from either approach on its own. Integrative transcriptomic and proteomic studies have been performed for organelles, cell-types and organs in Arabidopsis, including chloroplasts [24,48], pollen [49], guard cells [50], and trichomes [51]. The most comprehensive proteomic study in Arabidopsis to date integrated the proteome findings with gene expression data to reveal potential biomarkers for roots, flowers, leaves, seeds, siliques and cell culture [18^••]. Colocalization of transcripts and proteins reduces the likelihood that either occurred by contamination or chance. Comparing gene expression levels and protein abundance is challenging due to differing mRNA and protein stability [52]. Nevertheless, a few studies including some in plants have shown a small yet significant positive correlation between mRNA and protein abundances [18^••,24]. Studies of individual cell-types are now possible which will provide more refined information regarding the colocalization of transcripts and proteins. Proteomic data can also be used to inform genome annotation and characterize post-translational modification as has been demonstrated in a number of recent studies [18^••,53,54]. These studies revealed both missing and improperly annotated genes, highlighting the advantage of using proteomics for gene annotation.

Transcriptomics–metabolomics

Identification of specific compounds from experimental data (i.e. MS and NMR spectra) is a noteworthy, challenging task. Correlating mass peaks with transcripts could be a powerful strategy for identifying metabolites in complex extracts. Gene-to-metabolite associations have now been characterized for stress responses, plant defense and hormone-induced responses [55,56,57^•,58,59]. Early integration of transcriptomics and metabolomics studies looked at the global and dynamic response during sulfur and nitrogen depletion at the system-level [60–63]. Detailed analyses resulting from this type of integration have identified several genes including those involved in glucosinolate biosynthesis, anthocyanin biosynthesis, chain elongation enzymes and glucosinolate transport [64–66]. Recently, large-scale dynamic transcriptomic and metabolomic studies have been undertaken to gain a comprehensive understanding of how biological systems respond to other stresses such asm elevated CO₂ and salinity [67]. Additionally, integration of transcriptomic and metabolomic data from multiple related species and/or genotypes has been useful for identifying genes and processes underlying complex traits [68].

Integrated transcriptomic and metabolomic analyses have been successfully combined with reductionist approaches to investigate regulatory mechanisms involved in gene expression and metabolites. Specifically, the altered expression of the transcription factors that regulate anthocyanin biosynthesis [64] allowed the identification of genes involved in later steps of this metabolic process [69,70]. Additionally, ectopic expression of a transcription factor that regulates the cold response also showed metabolomic changes [56]. Moreover, mutations in the abscisic acid (ABA) biosynthesis pathway together with the integrated analysis of the transcriptome and metabolome demonstrated that ABA can reconfigure metabolite levels as a response to dehydration stress [59].

The hypothesis that a correlation exists between gene expression and metabolite accumulation patterns has proved valid when trying to identify the function of genes [71]. Therefore, to facilitate the integration of gene expression data in the context of a functional/metabolic pathway, software packages such as MapMan, PageMan and MetGenMAp were developed [72–74]. These programs have been successfully used to identify genes and metabolic pathways involved in response to nitrogen deficiency, diurnal cycles, and more recently, seed dormancy and germination [63,75–77]. This ‘guilt-by-association’ approach was used to predict the function of genes coregulated under given conditions and identify genes involved in metabolite biosynthesis and transcriptional regulators of many different pathways [78,79]. Taken together, these results suggest that co-occurrence of transcripts and metabolites is a powerful approach for deciphering gene function.

Conclusion and future prospectives

As the studies that are highlighted here demonstrate, the integration of multiple genomic-scale studies can reveal novel biology. A comprehensive systems-level understanding of Arabidopsis will require -omics methods to be integrated and combined [80,81]. In the near future, each of these -omics approaches will be used in an integrative fashion to inform and validate the findings of other genome-scale projects. Proteomics has been used to predict metabolic activity in the roots and shoots of Arabidopsis demonstrating the interconnectivity of these -omics efforts [82]. These studies explore the relationship between genomic information and the products directly and indirectly encoded by the genome, which will lead to novel testable hypotheses regarding the connection between genotype and phenotype.