|Home | About | Journals | Submit | Contact Us | Français|
Protein phosphorylation is a common post-translational modification of enormous biological importance. Analysis of phosphorylation at the global level should shed light on the use of this modification to regulate metabolism, signal transduction, and other processes. We have begun a proteomic analysis of phosphorylation using two-dimensional gel electrophoresis. Chinese hamster ovary (CHO) cells were metabolically labeled using 32P-orthophosphate. The proteins were extracted and run on two-dimensional electrophoresis. Gels were stained using colloidal Coomassie stain, dried, and phosphorimaged. The Coomassie stain allowed the observation of 468 individual protein spots. The phosphorimage showed 181 spots. The phosphoproteome of CHO cells therefore comprises around one third as many proteins as the CHO cell abundance proteome. However, the most intense spots in the phosphoproteome usually do not correlate with intense spots in the abundance proteome. We investigated the effects of labeling time, finding that the number of observable spots increases but the relative intensities also change. We also investigated the effects of adding a phosphatase inhibitor during labeling. Finally, we evaluated a phosphoprotein-specific stain (Pro-Q Diamond) in comparison with radiolabeling methods. There is not perfect correlation between radiolabeled phosphoproteinsand Pro-Q Diamond-stained phosphoproteins.
Since the discovery in 1955 that reversible phosphorylation regulates the activity of glycogen phosphorylase,1 there has been intense interest in studying protein phosphorylation. Phosphorylation and dephosphorylation of proteins regulate a variety of essential biological phenomena such as enzyme activity, signal transduction, transcriptional regulation, cell division, cytoskeletal rearrangement, and cell movement, apoptosis, and differentiation. The phosphorylation state also plays a critical role in cell–cell communication during development, responses to physiological stimuli, homeostasis, and nervous and immune system function. Generally, the phosphorylation state of a protein cannot be predicted from the gene sequence, and thus needs to be determined at the protein level.
Currently there are two popular approaches for examining the phosphoproteome. One is to digest the proteins, separate the phosphopeptides by affinity methods, and then identify these peptides by tandem mass spectrometry (MS/MS).2 The other method is to separate the proteome first with two-dimensional electrophoresis (2DE) and then visualize the phosphoproteins. However, in order to specifically visualize phosphoproteins, cells must incorporate radioactive phosphate through radiolabeling or be treated with a phosphate-specific stain. Though protein separation followed by phosphoprotein visualization is not as facile an identification process as affinity methods, it does have the advantage of providing a view of the phosphoproteome as a subset of the overall proteome. In view of this advantage, we undertook a cursory examination of the phosphoproteome of Chinese hamster overy (CHO) cells, including a comparison of radiolabeling and phosphoprotein-specific fluorescence staining.
Tissue culture plates (60-mm diameter) of CHO cells were grown to 100% confluency. Media was removed, and cells washed with phosphate-free Dulbecco’s modified Eagle’s medium (DMEM) media (Gibco, Carlsbad, CA) three times. The cells were then incubated in 3 mL of phosphate-free DMEM with 100 μCi of K2H32PO4 (Perkin Elmer, Boston, MA) for 4 to 18 h, with or without adding 5 μL of 1 mM microcystin-LF (Sigma Aldrich, St Louis, MO). Following incubation, the 32P-containing media was removed, and the cells were washed three times with phosphate-buffered saline and extracted with 200 μL 1% of sodium dodecyl sulfate (SDS).
Protein extracts from cells were acetone-precipitated by addition of 9 volumes of ice-cold acid acetone (acidified with one drop of 6 N HCl in 10 mL of acetone) followed by vortexing and incubation at −80°C for at least 1 h. After spinning for 10 min at 13,400 × g at 4°C, acetone pellets were air dried and dissolved in rehydration buffer (7 M urea, 2 M thiourea, 15 mM dithiothreitol, 4% Triton X-100, and 1% pharmalytes). The first dimension (isoelectrofocusing) employed Amersham IPG strips (pH 3–10) on the Multiphor II apparatus (Amersham Pharmacia, Piscataway, NJ). Proteins in electrofocused IPG strips were reduced in 32 mM dithiothreitol and alkylated using 216 mM iodoacetamide. The second dimension was run on a 10% SDS-polyacrylamide gel electrophoresis (SDS-PAGE) gel. The 2D gels were stained with colloidal Coomassie (Pierce, Rockford, IL), equilibrated with 5% glycerol, and dried onto Whatman filter paper; radioactivity was visualized on a Storm 860 phosphorimager (Molecular Dynamics, Sunnyvale, CA).
Spots that were excised from 2D gels were trypsinized in gel according to the method of Shevchenko et al. then desalted and concentrated using μC18 ZipTips (Millipore, Billerica, MA). Samples were spotted on a matrix-assisted laser desorption/ionization (MALDI) plate with an equal volume of α-cyano-hydroxycinnamic acid as the matrix. MS employed a QSTAR Pulsar i instrument equipped with an orthogonal MALDI (oMALDI) source (Applied Biosystems, Framingham, MA). MS spectra of the digests were obtained and parent ions selected from each digest for collision-induced fragmentation. The collision-induced fragmentation spectra were then submitted to the Mascot website (Matrix Science, London, UK) for database searching.
Tissue culture plates (60-mm diameter) of CHO cells were grown to 100% confluency. Plates were washed with phosphate-buffered saline, extracted with 1% SDS or incubated with phosphate-free DMEM media overnight, and harvested with 1% SDS. Cell extracts were acetone precipitated and subjected to 2DE apparatus as above. Gels were stained with Pro-Q Diamond phosphoprotein gel stain (Molecular Probes, Eugene, OR) according to the manufacturer’s protocol. These gels were visualized using the DarkReader DR45M Transilluminator (Clair Chemical Research, Doroles, CO). Subsequently, these gels were stained with Sypro Ruby stain (Bio-Rad, Hercules, CA) for total proteins.
Images of gels stained with colloidal Coomassie, Sypro Ruby or Pro-Q Diamond were captured with the AlphaDigiDoc system (AlphaInnotech, San Leandro, CA). These images, as well as phosphorimages, were analyzed using the Melanie 4 software package (GeneBio, Geneva, Switzerland).
Our initial objective was to examine the phosphoproteome as it compares with the abundance proteome (the abundant proteins, as detected by Coomassie blue stain). For this comparison, Melanie 4 detected 468 spots in the Coomassie-stained gel (Fig. 1A1A),), and 181 spots in the phosphorimage (Fig. 1B1B).). Thus, the observable phosphoproteome is about one third the size of the abundance proteome. However, this does not mean that one third of the proteome is phosphorylated. Low-abundance proteins are not visualized by Coomassie stain. In contrast, radiolabeling allows visualization of even fairly low abundance phosphoproteins (though we have not determined our limit of detection). Although the images in Figure 11 are from the same gel, only 51 spots in the phosphorimage match spots on the Coomassie-stained gel (indicated by crosses in Fig. 11).). Around 72% of phoshoproteins (130 out of 181, see Fig. 22)) cannot be matched with counterparts in the Coomassie-stained gel. Thus, most phosphoproteins are low-abundance proteins, falling below the limit of detection of the Coomassie blue stain.
It is tempting to use these data to determine the proportion of the proteome that is phosphorylated. Fifty-one of 468 (11%) proteins observed by Coomassie blue stain were apparently phosphorylated. Using this ratio one might extrapolate to the entire proteome, assuming that the ratio of phosphoproteins as a subset of protein observed is similar for the low-abundance proteins not observed with Coomassie blue. However, this assumption is questionable, for it is conceivable that phosphorylation is more common among low-abundance proteins. Ultimately, we cannot draw conclusions from this type of comparison in terms of the proportion of the entire proteome that is modified by phosphorylation.
The extent of labeling significantly limits the detection of radiolabeled phosphoproteins. We therefore examined the phosphoproteome observed using different labeling times. It should be noted that no discernable changes in the Coomassie-stained pattern were seen regardless of labeling time (data not shown). In a comparison between 4h and 16 h of 32P labeling (Fig. 33),), the longer labeling did allow detection of 22% more 32P-labeled spots. However, the increased labeling time affected the number of spots as well as the relative intensities among spots.
These two views of the phosphoproteome can be meaningfully compared by matching only the more intense spots (i.e., setting a high threshold for spot detection). When such a comparison was made, only 71% of the spots correlated between the two phosphorimages. This indicates that there are changes in the phosphoproteome pattern that are dependent on labeling, though the overall pattern is similar. These changes can also be seen by comparing the relative abundances of the 10 most intense spots from each of these two gel patterns (Fig. 44).). If increasing the labeling time had the straightforward effect of only increasing the signal of all phosphoproteins, then the relative intensities of the spots common to these two views would be conserved. The fact that the relative intensities are not conserved indicates that labeling time does not simply increase the sensitivity of the labeling. We assume that the changes in the phosphoproteome with variation in labeling time are due to the varied turnover rates of different phosphoproteins. Some proteins might not have reached equilibrium in terms of labeling in the shorter labeling period, while others had. These kinetic considerations indicate that caution should be used in interpreting intensities from radiolabeling experiments. Unless we can be sure that we are at equilibrium conditions, we cannot be completely confident of relative stoichiometries of labeling. Additionally, changes may be due to the prolonged exposure to the low phosphate concentration in the media during labeling.
Stoichiometry of phosphorylation is often maximized by the inclusion of phosphatase inhibitors during labeling. Additionally, information might be gained about the substrates of particular phosphatases by comparing labeled proteins with and without a specific inhibitor. We employed the inhibitor microcystin, which inhibits both type 1 and 2A protein phosphatases. As with varying labeling times, a good comparison can be made by matching only the more intense spots (using a high threshold for spot detection). Such a comparison was made between cells labeled in the presence or absence of microcystin added during labeling (Fig. 55).). Though the overall phosphoproteome pattern remains similar, 43% of the spots do not match. Obvious candidates for phosphatase 1/2A dephosphorylation are the spots labeled a–d in Figure 5B5B.. These and more subtle changes can be seen when the 10 most intense spots from each gel are compared (Fig. 66).
A goal of phosphoproteome analysis is the identification of the proteins that are phosphorylated. Protein spots from 2D gels can be excised and the proteins identified using in-gel digestion followed by MS. Eighteen gel pieces that represented matches between spots on the Coomassie-stained gel and the phosphorimage shown in Figure 11 were selected as candidates to be identified and were cut out from the gel. These spots were then digested in-gel and analyzed by oMALDI quadrupole-time of flight MS. Fragmentation spectra (from collision-induced fragmentation) were submitted for identification by comparison to fragmentation patterns predicted from the National Center for Biotechnology Information nonredundant database. Because of the incompleteness of the sequence database for Chinese hamsters, we were unable to identify several of the spots. Spots H, N, and Q were identified as the highly conserved proteins listed in Table 11.. One of these, hRNP K, is known to be a phosphoprotein.4 It is known that the pyruvate dehydrogenase E1 complex is phosphorylated,5 though we could not find in the literature if the beta subunit has previously been determined to be phosphorylated.
Because the majority of the phosphoprotein spots were of such low abundance that they were not observable by Coomassie stain, it is questionable whether or not identification of these spots would be possible, even if working with samples from an organism whose genome has been completely sequenced. Indeed, the correlation of the radioactive spots with Coomassie-stained spots is only tentative, and it is possible that a phosphoprotein might overlap on the gel with an abundant protein that is not phosphorylated but that would be identified from the excised gel piece. Our results suggest that this method is not a preferable method for identifying proteins that are phosphorylated.
Because of the inconvenience of working with radioactive material, we decided to investigate alternative ways of visualizing phosphoproteins after 2DE. We evaluated the commercially available phosphoprotein fluorescence stain Pro-Q Diamond. Following the phosphor-specific staining, a fluorescent stain for total protein was employed. This staining showed that the overall proteome was similar to that seen using Coomassie (Figs. 7A and BB).
We found that there is not perfect correlation between the radiolabeled and Pro-Q Diamond-stained phosphoproteins (Figs. 7C and DD).). The differences could be due to aspects of the radiolabeling, such as favoring phosphoproteins that experience rapid turnover. Alternatively, differences could be due to unequal staining among phosphoproteins or lack of complete specificity of the phosphoprotein stain. A further limitation to the use of this stain in many labs is that it serves only as a fair phosphoprotein stain when laser-based gel scanning is not available as a visualization method. Using transillumination as we did results in a lower-than-optimal signal-to-noise ratio with the stain, and a higher limit of detection. However, we have no reason to suspect that visualization using transillumination should affect the relative staining intensities of the phosphoproteins that are detected.
Because radiolabeling of phosphoproteins is very sensitive and straightforward to quantify, such analysis can give a clear picture of the relative phosphosphorylation of proteins present in a sample. However, care must be taken to determine that the proteins being labeled are in equilibrium with respect to the incorporation of label for this quantitation to be meaningful. The radiolabel approach is accessible to most labs because it does not require sophisticated instruments, though it does involve the inconvenience of working with radioactive materials.
Because of the limitations of 2DE and Coomassie blue staining, many phosphoproteins cannot be visualized, especially low-abundance proteins. Our results show that most phosphoproteins are low abundance, including some of the most highly phosphorylated proteins. Furthermore, identification of phosphoproteins from this approach is difficult because even if a spot is visible to be cut out, it is not guaranteed that the protein identified is the radiolabeled protein seen to be phosphorylated. Though these methods do enable comparative profiling, they do not allow facile identification of the phosphoproteins.