Quantitative proteomics as a new piece of the systems biology puzzle Angela Bachia, Tiziana Bonaldib,��� aDIBIT, San Raffaele Scientific Institute, Milan, Italy b Department of Experimental Oncology, European Institute of Oncology, Via Adamello 16 20139 Milan, Italy A R T I C L E I N F O A B S T R A C T The definition of the role of each gene product in its cellular context is of outstanding importance in the post-genomics era. Recent technological innovations have driven research in proteomics from single protein characterization to global approaches, aiming to achieve a comprehensive qualitative and quantitative description of complex molecular mechanisms. In this review, we discuss the methodology of quantitative proteomics as it applies to the analysis of complex biological model systems. A special attention will be given to model systems that are suitable for functional genomic studies, where the potential of quantitative proteomics can be effectively demonstrated. �� 2008 Elsevier B.V. All rights reserved. Keywords: Quantitative proteomics Mass spectrometry Stable-isotope labeling System biology Gene-phenotypization 1. Introduction The definition of the cellular functions of each gene product remains an unfinished goal of the post-genomic era. Despite great efforts in annotation, the physiological role of a large subset of known genes remains elusive. The analysis of phenotypes resulting from the ablation of the function of a gene or a gene product is generally performed in model systems that are amenable to genetic manipulation, including Saccharomyces cerevisiae, Schizosaccharomyces pombe, Caenor- habditis elegans, Drosophila melanogaster, and Arabidopsis thali- ana. In particular, yeast, worm, and fruit fly are excellent organisms to study complex biological functions, mainly due to the possibility to perform loss-of-function and gain-of- function experiments. The basic concept of loss-of-function studies is to screen for the response, at either the cellular or the organism level, to the depletion of one or more gene products [1]. Loss-of-function has been traditionally achieved by classical forward genetics methods. More recently, genetic screens have been facilitated by the RNA interference (RNAi) approach. In some invertebrate models, like fruit fly and worm, RNAi has become the elective tool for high-throughput functional studies thanks to the straightforward protocol of double-strand RNA (dsRNA) transfer, the high penetrance of target depletion, and the relatively low grade of unspecific silencing (the so-called ���off-target��� effect [2,3]). Thus, the application of RNAi has paved the way to high-throughput genetic screens in invertebrates, and more recently also in mammalian cell-culture systems. When focusing on a specific mechanism, phenotypization can be easily achieved by means of specific functional assays designed to detect variations in a particular feature or behavior. When studying a cellular phenomenon from a global perspec- tive, however, the functional characterization of novel genes is more difficult and requires comprehensive screenings [4,5]. Classical phenotypization approaches are based on the evalua- tion of descriptive information such as cell or organism viability and morphology [6]. Alternatively, global gene expression analysis via microarrays has been largely employed to under- stand how gene expression is modulated by different environ- mental conditions. The advantage of this approach is the possibility to measure thousands of gene transcripts in a single experiment, providing an in-depth characterization of a func- tional state in a loss-of-function or gain-of-function setup. J O U R N A L O F P R O T E O M I C S 7 1 ( 2 0 0 8 ) 3 5 7 ��� 3 6 7 ��� Corresponding author. Tel.: +39 02 94375123 fax: +39 02 94375990. E-mail address: tiziana.bonaldi@ifom-ieo-campus.it (T. Bonaldi). 1874-3919/$ ��� see front matter �� 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.jprot.2008.07.001 available at www.sciencedirect.com www.elsevier.com/locate/jprot

The main drawback of transcript-based approaches is the lack of information on the state of the proteome. Profiling proteins instead of mRNAs is more informative, since proteins can be in most cases regarded as the ���effectors��� of cellular functions. Thus, the ���most suitable approach��� for gene- phenotypization and characterization of functional states consists of a global and unbiased analysis of the proteome or of specific proteome subsets. In recent years, several strategies for quantitative proteomics of model organisms have been explored with various degrees of success. Here we briefly introduce quantitative approaches that are currently in use and review in detail those studies that have combined comprehensiveness of quantitative proteomics with the efficient manipulation of certain model organisms. 2. Quantitative proteomics technologies in system biology 2.1. Instrumentation Mass spectrometry has become the primary method for the identification of proteins. Continuous improvements in mass spectrometry technology over the last decade have made it possible for proteomics research to tackle complex biological processes at a global qualitative and quantitative level [7]. Mass spectrometry in protein science can be used to measure the mass of a peptide or a protein (single stage MS). Additionally, it can provide structural information on peptide or protein composition, which are in turn employed to derive aminoacid sequence and to identify post-translational mod- ifications (tandem MS or MS/MS) [8���11]. Since their introduction in the late '80, MALDI (Matrix assisted laser desorption ionization ([12,13]) and ESI (electro- spray ionization [14]) have been the most widely used soft ionization techniques for biomolecular analysis. There is now a large choice of mass spectrometry instrumentation for large- scale proteomics. The most commonly used mass analyzers are ion trap (IT), time of flight (TOF), TOF���TOF, quadrupole/TOF hybrids, IT/orbitrap hybrids and IT/Fourier transform ion- cyclotron resonance mass spectrometers (FTMS) hybrids. These designs have different sensitivity, resolution, mass accuracy and ability to produce high quality MS/MS spectra. These features have been recently and extensively reviewed [15,16] and are beyond the scope of this review. It is never- theless important to note that one of the major achievements of last generation instruments, suchas orbitrap or FT-ICR mass spectrometers, is the ability to measure masses with very high resolution (up to 150000 for the FT-ICR) and very high mass accuracy (up to 1 ppm). This allows approaching the elemental composition of peptides, at leastfor peptidemassescommonly measured in bottom-up proteomics experiments [17]. Improvements in peptide sequencing have also been recently achieved with the introduction of fragmentation techniques alternative to the most popular collision-induced dissociation (CID) that are typically used with the mass analyzer mentioned above. In particular, electron capture dissociation (ECD) has greatly increased the specificity of the analysis and the rate of protein identification when used in combination with CID [18���20]. Another recently introduced fragmentation technique, named electron transfer dissocia- tion (ETD), has been shown to produce extensive peptide sequence information that is often missing in conventional CID, and proved to be especially useful for sequence analysis of post-translationally modifiedand highly basicpeptides [21,22]. Despite these technological innovations, the analysis of a full proteome is still a challenging task, mainly because of the high complexity of protein samples. To overcome this problem, several separation techniques (such as multidimen- sional chromatography, MudPit [23]) or specific enrichment/ depletion techniques (Tandem Affinity Purification [24] equalizer beads [25]) can be applied prior to mass spectro- metric analysis. These approaches increase the proteome coverage and the dynamic range of large-scale proteomics analyses. 2.2. MS-based strategies for quantitative proteomics 2.2.1. Quantitation via stable-isotope labeling Mass spectrometry is not inherently quantitative. The peptide products of proteolytic cleavage of proteins exhibit a wide range of physicochemical properties (size, charge, hydropho- bicity) that influence the intensity of the signal in a mass spectrometric analysis. For accurate quantification, the abun- dance of each individual peptide must be compared under different experimental conditions. An elegant method to achieve this is the labeling of proteomes with stable isotopes. The approach is based on the idea that a stable-isotope labeled peptide is chemically identical to its native counterpart, and behaves identically during chromatographic and/or mass spectrometric analysis, but is distinguishable in a mass spectrometer owing only to a mass difference. The ratio of signal intensities for the labeled and unlabeled peptide pairs provides an accurate measure of relative abundance of peptides/proteins under different biological conditions. In general, strategies for isotope-based quantitative pro- teomics can be distinguished into two groups, depending on whether the isotopic tag is incorporated in vitro during sample preparation or in vivo. We refer to several good reviews for the detailed description of the different strategies and the discussion of their relative advantages and disadvantages [26���29]. Here, we briefly summarize the general concepts of approaches based on the introduction of isotope-based tags, focusing on the most successful and widespread applications in various model systems. 2.2.1.1. In vivo labeling. In vivo labeling approaches involve metabolic incorporation of stable isotopes during protein synthesis in cells grown in special media, supplemented with the isotopes. The first successful experiments were performed in yeast and bacteria, by employing media containing either 14N (light) or 15 N (heavy) isotopes [30,31]. Subsequently, this strategy has been used for quantitative proteomics in several additional eukaryotes, including C. elegans, D. melanogaster, A. thaliana and Rattus norvegicus (see below). A possible pitfall of the metabolic labeling approach via 15 N is the number of labeled nitrogen atoms varying from peptide to peptide with the consequent unpredictability of the exact mass shift generated. An alternative approach for metabolic labeling with stable isotopes that overcomes this problem is the ���stable-isotope 358 J O U R N A L O F P R O T E O M I C S 7 1 ( 2 0 0 8 ) 3 5 7 ��� 3 6 7