JXB Advance Access originally published online on March 10, 2006
Journal of Experimental Botany 2006 57(7):1523-1527; doi:10.1093/jxb/erj126
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
RESEARCH PAPER |
Phosphoproteomics in Arabidopsis: moving from empirical to predictive science
Sainsbury Laboratory, John Innes Centre, Norwich NR4 7UH, UK
*Present address and where correspondence should be sent: Department of Biochemistry, 271H Bond Life Sciences Center, University of Missouri-Columbia, Columbia, MO 65211, USA. E-mail: pecks{at}missouri.edu
Received 21 June 2005; Accepted 23 January 2006
 |
Abstract |
|---|
 Top Abstract IntroductionIdentification of...Insights from phosphoproteomic...Predicting new pathway...Quantitative large-scale...ConclusionsReferences |
|---|
Although protein phosphorylation is integral to the regulation of protein function in diverse biological responses, relatively little is currently known about the rules that govern phosphorylation in plants. This review will discuss how the data acquired by evolving phosphoproteomic methods are beginning to fill the gaps in our knowledge. In addition, ways are suggested in which new quantitative methods in conjunction with extrapolating from genomic data may provide a strategy to predict components of signalling networks that may be co-ordinately regulated.
Key words: IMAC, phosphoproteomics, phosphorylation site prediction
|
Introduction |
|---|
|
TopAbstractIntroductionIdentification of...Insights from phosphoproteomic...Predicting new pathway...Quantitative large-scale...ConclusionsReferences |
|---|
Protein phosphorylation is a highly conserved mechanism for regulating protein function. Phosphorylation may alter protein activity or subcellular localization, target proteins for degradation, or effect dynamic changes in protein complexes. Different kinases may be involved in each of these processes for a single protein, allowing a large degree of combinatorial regulation at the post-translational level. Therefore, in addition to knowing if a protein is phosphorylated, knowing which residue is phosphorylated during a particular response is essential to understanding the mechanistic regulation.
Traditionally, identification of phosphoproteins was performed one-by-one, characterizing a single protein during a single response. Although successful, this strategy has limitations. First, it is slow and cumbersome, generally requiring some level of protein purification to identify the phosphoprotein, followed by the production of antibodies to confirm its identity. Second, sequential identification of phosphoproteins is slow to reveal sufficient components to allow generalizations about which pathways are being affected. Finally, comparisons to determine how different stimuli impinge on signalling pathways will be biased towards the analysis of known components or known phosphorylation sites on these components. Thus, the identification of new components is not generally an outcome of these comparisons.
Phosphoproteomics seeks to overcome all these limitations via large-scale analysis of protein phosphorylation in entire cells or tissues. With recent advancements in the enrichment of phosphopeptides from complex mixtures of peptides (Ficarro et al., 2002; Nühse et al., 2003; Beausoleil et al., 2004; Gruhler et al., 2005), using mass spectrometry to sequence hundreds if not thousands of phosphorylation sites from a single experiment is now becoming a reality. The abundance of phosphorylation site sequence information is providing new insights into mechanistic regulation of plant proteins. Moreover, interrogation of these data provide the basis for generalizing rules governing protein phosphorylation in plants. In addition, new reagents and methods are now available that will allow large-scale quantitative comparisons of phosphoproteomes.
|
Identification of phosphoproteins and phosphorylation sites |
|---|
|
TopAbstractIntroductionIdentification of...Insights from phosphoproteomic...Predicting new pathway...Quantitative large-scale...ConclusionsReferences |
|---|
Despite the importance of phosphorylation, the detection and analysis of phosphoproteins historically has been a difficult process. A number of options exist for using 2-D gels to identify candidate phosphoproteins (Lewis et al., 2000; Peck et al., 2001). In addition to the fact that 2-D gels are limited in the number of proteins they can resolve, however, phosphoproteins are particularly troublesome because of low stoichiometry; the phosphorylated form of a protein is usually only a small fraction of the total population. Together with the generally low abundance of signalling proteins, phosphoproteins can be extremely difficult to detect and sequence from 2-D gels; and even if a candidate protein is identified by mass spectrometry, validation is required to ‘prove’ that the candidate is a phosphoprotein (for the reasons discussed below). Therefore, 2-D gels do not fulfil the goals of high-throughput phosphoproteomics.
The reason that proteins from 2-D gels need to be confirmed as phosphoproteins is that standard sequencing by mass spectrometry rarely yields the phosphopeptide, let alone the phosphorylation site. A peptide needs to be ionized to be detected by mass spectrometers, and ionization of phosphopeptides is usually suppressed in the presence of non-phosphopeptides. Essentially, the suppression effect makes phosphopeptides ‘invisible’ in complex mixtures. Circumventing this problem is the root of modern phosphoproteomics.
The most common method to enrich for phosphopeptides is immobilized metal affinity chromatography (IMAC). Under acidic binding conditions, the strong positive charge of the transition metal, usually Fe3+ or Ga3+, selects the negatively charged phosphate group from the mixture (Posewitz and Tempst, 1999; Stensballe et al., 2001). Using IMAC, the possibility exists to shift the experiment from examining individual proteins to examining complex peptide mixtures arising from proteolytic digests of total protein, typically using trypsin.
A potential downfall of IMAC is that it may also bind peptides containing acidic residues. Methylation of acidic residues was found to improve the specific binding of phosphopeptides from yeast (Ficarro et al., 2002) and from Jurkat T cells (Brill et al., 2004). This approach has been successfully used with plant samples to identify eight phosphopeptides from Arabidopsis thylakoid membranes, including three new phosphopeptides (Hansson and Vener, 2003).
An alternative IMAC procedure was produced in this laboratory that does not require secondary modification chemistry, but still yields highly pure (75–90%) phosphopeptides from Arabidopsis plasma membranes (Nühse et al., 2003). In this case, deconvoluting the sample by including strong anion exchange prior to IMAC greatly increased the number of phosphopeptides identified (300 phosphorylation sites from about 200 proteins). Although the methylation of acidic residues was not necessary in these experiments, it remains to be seen if this is a unique property of plasma membrane proteins.
A further recent option is the use of strong cation exchange (SCX) columns under very specific pH conditions that greatly enriches phosphopeptides. This approach resulted in the identification of phosphorylation sites from more than 900 proteins using 8 mg HeLa cell nuclear protein (Beausoleil et al., 2004). This method has the advantage that it is relatively easy and very robust. However, the purity of phosphopeptides was far lower than either of the methods described above. Another group combined a modified SCX fractionation prior to IMAC to identify phosphorylation sites from more than 500 yeast proteins (Gruhler et al., 2005). Interestingly, very little overlap was observed between phosphorylation sites identified in this study and those from the study using methyl-esterification of acidic residues (Ficarro et al., 2002), indicating that different methods may preferentially select different phosphopeptides.
In summary, a number of methods exist for selectively enriching phosphopeptides from complex peptide mixtures. As is the case for most proteomic approaches, a ‘best’ method may not exist. Instead, biases or advantages may be discovered by using one method for certain types of studies. The important fact remains, however, that these methods allow the simultaneous sequencing of hundreds to thousands of phosphopeptides. More importantly, these experiments often identify the precise residue that is phosphorylated.
|
Insights from phosphoproteomic data |
|---|
|
TopAbstractIntroductionIdentification of...Insights from phosphoproteomic...Predicting new pathway...Quantitative large-scale...ConclusionsReferences |
|---|
Very little is known about protein phosphorylation in plants. Current phosphorylation site prediction programs are trained on mammalian data and have limited value for accelerating our knowledge. Because kinase specificity motifs generally consist of only a few residues, searching with these low complexity parameters (e.g. Arg-X-Ser/Thr, where X can be any residue) results in 20–30 putative sites predicted for any protein, most of which are sequence noise. The probability of a false-positive prediction is enhanced because many of the well-characterized mammalian kinases do not appear to exist in plants. Conversely, these programs failed to predict more than 40% of the sites found in a large, in vivo phosphorylation site data set from this laboratory (Nühse et al., 2004). Therefore, reliance must be placed on empirical data from in vivo phosphorylation sites for building an understanding of phosphoregulation of proteins in plants.
An initial dataset of more than 300 phosphorylation sites from plasma membrane proteins (Nühse et al., 2004) was analysed. It was noticed that more than 80% of the phosphorylation sites were outside the known Pfam domains. Even within Pfam domains subdivided into highly conserved regions, as is the case with kinases, the phosphorylation sites were found to be outside these conserved regions. If the conserved residues within the Pfam domain confer its fold/structure to perform a particular function, the lack of phosphorylation sites in these regions makes sense. Clearly, a residue needs to be solvent exposed in order to be a target for a kinase; and many of the structural residues are likely to be buried within the fold. Moreover, signalling pathways could not achieve any specific regulation if they targeted residues that are critical to the basic function of a class of proteins. By continuing this logic, it is inferred that the predominance of phosphorylation in variable domains exists to modulate protein activity under a particular set of circumstances. The most straightforward test of this hypothesis was to compare phosphorylation sites found in different members of a gene family. In a published study, sites found in a cellulose synthase, CesA3 (Nühse et al., 2004), were compared. Not only were the phosphorylated residues outside the Pfam domains, they were in regions highly divergent among the ten gene family members in Arabidopsis. This comparison indicated that, in addition to specific transcriptional regulation of gene family members, these related proteins are likely to undergo specific post-translational regulation as well.
If these phosphorylation sites are critical to the function of the gene family member, one would expect the residues to be conserved in other plant species. Indeed, the target residues were highly conserved among the apparent orthologues of CesA3 (Nühse et al., 2004); and the same holds true for the majority of the other gene family members investigated. While these proteins must be sequenced in other species to confirm that they are also phosphorylated, these results indicate that post-translational regulation defined in Arabidopsis is likely to be applicable to a broad range of plant species, including crop plants.
This observation is particularly important because large-scale phosphorylation site sequencing requires a fully sequenced genome. Unlike other proteomics methods, sequencing phosphopeptides requires that the protein is identified from just one peptide. With unsequenced genomes, problems arise both from missed identifications (i.e. a sequence is not in the database) as well as from false identifications being forced as the search program tries to make matches between spectra and sequences. From gel-based identification of proteins, mis-identification of a single peptide is not a great concern because one generally obtains five to six good quality MS/MS spectra that will cross-support the identification. With LC-MS/MS-based phosphoproteomics, however, each spectrum must be manually interrogated to be confident of the correct assignment, even when using a fully sequenced genome such as Arabidopsis. Until sufficient genome data are available for efficient phosphoproteome sequencing, the information from Arabidopsis should be transferable to allow mechanistic studies in other plants.
In summary, phosphorylation in plants tends to be outside the Pfam domains and more conserved between orthologues than between paralogues. Phosphorylation also tends to occur in unstructured regions and towards the N- or C-termini. Most of these observations are likely to be related. For instance, solvent exposed loops tend to be unstructured and outside the Pfam domains, or at least are hypervariable regions within the Pfam domain. Similarly, the N- and C-termini of proteins tend to have less structure, making these regions both good targets for phosphorylation and/or participating in inter- or intramolecular interactions. The fact that these unstructured regions tend to be conserved in orthologues throughout evolution indicates that these flexible regions are not random but, in fact, serve specific functions. As discussed below, this evolutionary maintenance of signalling components can be exploited by utilizing genome information to predict new components that are co-ordinately regulated during a response.
|
Predicting new pathway components |
|---|
|
TopAbstractIntroductionIdentification of...Insights from phosphoproteomic...Predicting new pathway...Quantitative large-scale...ConclusionsReferences |
|---|
The conservation of phosphorylation sites among orthologues has further implications. If the specific regulation of a protein has been conserved throughout evolution, the upstream kinase must be conserved as well. Each kinase will only phosphorylate a specific pattern of residues, defining its so-called kinase motif. Once this motif has been identified, it can be used to search a database to predict other proteins that contain the motif and, therefore, that may also be regulated during the response. Determining exactly what the motif is, however, generally requires a great deal of mutagenesis to determine which residues are critical. Moreover, these experiments de facto require that the kinase has already been identified, and very few kinase-substrate pairs have been identified in plants.
There is an argument that genome sequence data can be exploited to predict other proteins co-ordinately regulated by the same kinase during a response without prior knowledge of the kinase itself. For the kinase to recognize its substrate, the kinase motif must be, at least partially if not entirely, inherent within the primary sequence of the target protein. Therefore, the kinase motif must be evolutionarily conserved as well. However, as discussed when describing the problems of phosphorylation site prediction programs, kinase motifs generally consist of only a few residues unpredictably spaced from the phosphorylated residue; and this vague information creates too many false-positive predictions to be useful.
To circumvent this limitation, one should consider the sequence surrounding a phosphorylation site as representing the specificity determinants of an entire ‘signalling module’ rather than of only a kinase (Fig. 1). The reason for this view is that a regulated pathway consists of more than a kinase and substrate. After a protein is phosphorylated, the signal must eventually be down-regulated. Down-regulation may occur via a phosphatase or via protein degradation (e.g. ubiquitination targeting a protein to the proteasome). In either case, the mechanism would require some level of specific recognition based on the region surrounding the phosphorylated residue. A similar line of logic applies to phosphobinding proteins. Phosphorylation often leads to the formation of new protein complexes via interactions between the phosphorylated region and proteins containing specific phosphobinding domains. Again, specificity must arise from some of the residues surrounding the phosphorylation site. If these assumptions are true, then evolutionary divergence of residues surrounding the phosphorylation site must be constrained amongst orthologues in order to maintain the signalling module. Any non-essential change to a surrounding residue could be tolerated, but any change in a critical residue would result in the loss of regulation.
|
Aligning orthologous sequences surrounding the phosphorylation site should, therefore, define the motif of a signalling module. This more accurate motif can be exploited to train a bioinformatics search to identify additional proteins that may be regulated by the same signalling module and, therefore, may be part of the same biological response. Because the motif still is likely to be relatively short, false-positives may arise during the search. But if the logic employed is correct, false-positives can be screened out by repeating the alignment of orthologues for each of the candidate proteins. If a candidate is indeed regulated by the same signalling module, the motif should be evolutionarily conserved as well. Indeed, this logic was tested with MAP kinase substrates identified from an earlier phosphoproteome screen and it was possible to identify multiple new substrates of MAP kinases (E Andreasson, D Studholme, and S Peck, unpublished results). Thus, in conjunction with ongoing work in this laboratory to create a phosphorylation site database for Arabidopsis, this approach will accelerate the identification of signalling networks in silico. This approach will also help complete phosphoproteomics projects. As with any proteomics approach, large-scale sequencing generally does not saturate all the information available and rapidly reaches a point of diminishing returns. Thus, the bioinformatics search using conserved motifs can be used to generate testable hypotheses, which will be particularly useful for potentially low abundant signalling components.
|
Quantitative large-scale phosphoproteomics |
|---|
|
TopAbstractIntroductionIdentification of...Insights from phosphoproteomic...Predicting new pathway...Quantitative large-scale...ConclusionsReferences |
|---|
To maximize the utility of the strategies outlined above, a quantitative method is required to detect changes in the phosphoproteome. Quantitative comparisons of phosphopeptides have become feasible only relatively recently. The primary limitation is that the intensity of spectra cannot be compared between independent experiments. Therefore, quantitative comparisons require concurrent fractionation and ionization of peptides. Currently, three methods are available: stable isotopic labelling of amino acids in cell culture (SILAC), isotopic labelling via methyl esterification of acidic residues, and isobaric tagging of primary amines.
SILAC was first described as an inexpensive means of performing quantitative proteomic experiments in mammalian cell cultures (Ong et al., 2002). Cell cultures were grown in different media, one containing unlabelled amino acids and one containing an isotopically labelled amino acid such as arginine. Labelling via arginines and lysines is preferred because trypsin digests after these residues, meaning that only one label will be incorporated in each peptide. After treatment of one of the cultures, proteins isolated from both cultures can be mixed. Except for the mass difference from the isotopic labelling, the peptides are identical, meaning that they will behave in the same manner during fractionation and ionization; so the intensity of the peaks in the spectra will be directly comparable. This strategy was used in conjunction with SCX and IMAC enrichment of phosphopeptides to examine changes in the phosphoproteome of yeast in response to mating pheromone (Gruhler et al., 2005). The primary limitation of this method is that the cells must be amenable to isotopic labelling. Thus, the method is likely to be very good for experiments in cell culture, but of limited value to investigate the phosphoproteome of intact plants.
A second method is based on the methyl esterification of acidic residues used to decrease non-specific binding to IMAC resins. In this method, one set of peptides is methylated with deuterated methanol and the other with non-deuterated methanol. Conceptually, the remainder of the experiment is similar to that described for SILAC, with the primary difference being a non-standard isotopic difference between sets of peptides because the number of acidic residues varies from peptide to peptide. This method was used to identify proteins changing in phosphorylation status during mammalian sperm capacitation (Ficarro et al., 2003). As opposed to using SILAC, isotopic labelling via methylation is performed after protein isolation and is, therefore, compatible with experiments using any tissue.
The most recent solution involves isobaric labelling of all primary amines (Ross et al., 2004) and is commercially available under the name of iTRAQ (Applied Biosystems). This method allows comparison of up to four samples in a single experiment; and, as with methyl esterification, labelling occurs after isolating proteins, meaning that iTRAQ is compatible with comparisons involving any plant tissue. In conjunction with multidimensional chromatography and IMAC, iTRAQ was found to be a very reproducible method for comparing changes in the phosphoproteome (T Nühse, A Bottril, S Peck, unpublished results). Again, direct comparisons of methods will be necessary to conclude if one is better than another in certain circumstances; but the ability to design quadruplex experiments is certainly attractive as it allows comparisons of a control and treatment using a wild-type and mutant plant.
|
Conclusions |
|---|
|
TopAbstractIntroductionIdentification of...Insights from phosphoproteomic...Predicting new pathway...Quantitative large-scale...ConclusionsReferences |
|---|
These methods are only the beginnings of gaining a better perspective of protein phosphorylation in plants. However, technology is advancing to the point where the expansion of knowledge should occur quite rapidly over the next few years; and with the maturation of quantitative experiments, an exponential increase in the identification of regulated phosphorylation sites should soon be seen. Plant science may have a somewhat unique advantage when it comes to interrogating these new data. The amount of genomic sequence from species separated by hundreds of millions of years provides immediate insights into the factors governing biochemical specificity. In general, mammals are likely to be too closely related to make effective use of the type of bioinformatics screening process discussed in this review. Thus, plant science may be in a position to ask some extremely interesting questions. How does a signalling pathway evolve? Can new components evolve into an existing pathway following gene duplication? The diversity of plant form and function may be an excellent laboratory for exploring these crossroads of evolution and biochemistry.
|
Acknowledgements |
|---|
This work was supported by the Gatsby Charitable Foundation.
|
References |
|---|
|
TopAbstractIntroductionIdentification of...Insights from phosphoproteomic...Predicting new pathway...Quantitative large-scale...ConclusionsReferences |
|---|
, , , , , , , , . . . , –.Large-scale characterization of HeLa cell nuclear phosphoproteins. Proceedings of the National Academy of Sciences, USA (2004) 101:12130–12135.
, , , , , . . . , –.Robust phosphoproteomic profiling of tyrosine phosphorylation sites from human T cells using immobilized metal affinity chromatography and tandem mass spectrometry. Analytical Chemistry (2004) 76:2763–2772.[Medline]
, , , et al. . . , –.Phosphoproteome analysis of capacitated human sperm. Journal of Biological Chemistry (2003) 278:11579–11589.
, , , , , , , . . . , –.Phosphoproteome analysis by mass spectrometry and its application to Saccharomyces cerevisiae. Nature Biotechnology (2002) 20:301–305.[CrossRef][ISI][Medline]
, , , , , , . . . , –.Quantitative phosphoproteomics applied to the yeast pheromone signaling pathway. Molecular and Cellular Proteomics (2005) 4:310–327.
, . . , –.Identification of three previously unknown in vivo protein phosphorylation sites in thylakoid membranes of Arabidopsis thaliana. Molecular and Cellular Proteomics (2003) 2:550–559.
, , , , , , , , . . . , –.Identification of novel MAP kinase pathway signaling targets by functional proteomics and mass spectrometry. Molecular Cell (2000) 6:1343–1354.[CrossRef][ISI][Medline]
, , , . . . , –.Large-scale analysis of in vivo phosphorylated membrane proteins by immobilized metal ion affinity chromatography and mass spectrometry. Molecular and Cellular Proteomics (2003) 2:1234–1243.
, , , . . . , –.Phosphoproteomics of the Arabidopsis plasma membrane and a new phosphorylation site database. The Plant Cell (2004) 16:2394–2405.
, , , , , , . . . , –.Stable isotope labeling by amino acids in cell culture, SILAC, as a simple and accurate approach to expression proteomics. Molecular and Cellular Proteomics (2002) 1:376–386.
, , , , , . . . , –.Directed proteomics identifies a plant-specific protein rapidly phosphorylated in response to bacterial and fungal elicitors. The Plant Cell (2001) 13:1467–1475.
, . . . , –.Immobilized gallium(III) affinity chromatography of phosphopeptides. Analytical Chemistry (1999) 71:2883–2892.[Medline]
, , , et al. . . , –.Multiplexed protein quantitation in Saccharomyces cerevisiae using amine-reactive isobaric tagging reagents. Molecular and Cellular Proteomics (2004) 3:1154–1169.
, , . . . , –.Characterization of phosphoproteins from electrophoretic gels by nanoscale Fe(III) affinity chromatography with off-line mass spectrometry analysis. Proteomics (2001) 1:207–222.[CrossRef][ISI][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  R. Shin, S. Alvarez, A. Y. Burch, J. M. Jez, and D. P. Schachtman Phosphoproteomic identification of targets of the Arabidopsis sucrose nonfermenting-like kinase SnRK2.8 reveals a connection to metabolic processes PNAS, April 10, 2007; 104(15): 6460 - 6465. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||