OR WAIT null SECS
This installment of "Directions in Discovery" will review current phosphorylation enrichment techniques with a focus on new developments.
The reversible phosphorylation of specific sites on proteins is implicated in the control of multiple cellular functions and processes, including cell growth and differentiation, cell death, gene expression, and signal transduction. Almost a third of all cellular proteins are believed to be phosphorylated at any given point in the cell cycle. Given the importance of protein phosphorylation in cell activities, characterization of the phosphoproteome will be key to understanding the mechanism of cellular processes and in identifying targets for therapeutic intervention. However, phosphoproteome analysis is a daunting analytical challenge. Phosphorylation is dynamic, with many proteins being phosphorylated at different sites and at different times. Therefore, phosphorylation at a given site can be substoichiometric and, thus, in very low abundance. Currently, the dominant technologies for characterizing protein phosphorylation are matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry (MS) and liquid chromatography (LC)–nanoelectrospray tandem MS. Most MS-based protein characterization is performed at the peptide level after cleavage with site-specific proteases. Phophorylation sites are less than ideal candidates for these approaches. Phosphorylated peptides do not ionize well in positive ion mode (the preferred approach in most cases), and are subject to ion suppression in the presence of the great excess of nonphosphorylated peptides. The most effective strategy for phosphoproteome characterization is an initial enrichment to extract phosphoproteins or phosphopeptides from the bulk proteome. There are a variety of enrichment methods, including immunoprecipitation, chemical derivatization, immobilized metal affinity chromatography (IMAC), and enrichment on metal oxide surfaces. This installment of "Directions in Discovery" will review the current approaches, compare their performance, and take a look at some recent enrichment technologies.
Immunoprecipitation of proteins with antibodies specific to phosphorylated residues is an obvious strategy for enrichment of phosphoproteins from complex mixtures (1,2). This approach has been quite successful in the enrichment of proteins phosphorylated at tyrosine residues (3), as high-quality antiphosphotyrosine antibodies that have little crossreactivity with other phosphorylated or nonsposphorylated residues are available. In this application, immunoprecipitation is particularly valuable because phosphorylation at tyrosine is relatively rare, accounting for a small fraction of the phosphorylation sites. Antiphosphotyrosine antibodies generally are not considered suitable for the enrichment of phosphotyrosine-containing peptides because of poor selectivity (4). However, Rush and colleagues (5) successfully enriched phophopeptides from trypsin-digested cell lysates using antiphosphotyrosine antibody immobilized on agarose beads. Zhang and Neubert (4) demonstrated that the selectivity of immunoprecipitation of tyrosine phosphopeptides from digests of cell lysates can be improved significantly by the addition of the detergent n-ocylglucoside to the immunoprecipitation buffer. To date, no antibodies suitable for enrichment of proteins containing phosphoserine or phosphothreonine residues are available.
Chemical modification strategies for the enrichment of phosphoproteins or phosphopeptides rely upon the chemical properties of the phosphate groups. Two approaches have been described. One uses β-elimination of the phosphate group to create a site for introduction of an affinity ligand. The other uses conversion of the phosphate group to a phosphoramidate to facilitate enrichment of phosphopeptides on affinity supports. Chemical modification procedures can be very selective for phosphate enrichment, but their complexity can compromise overall recovery, and side reactions can complicate interpretation of results.
Beta Elimination With Chemical Modification
Under strongly alkaline conditions, phosphoserine and phosphothreonine lose H3PO4 via a β-elimination to yield dehydroalanine and dehydroaminobutyric acid, respectively. The Chait laboratory (6) employed this chemistry as a means of introducing a group for affinity enrichment by a Michael addition to the newly formed double bond (Figure 1). In their initial approach, proteins were first treated with performic acid to oxidize cysteine, cystine, and methionine residues. Then the β-elimination was performed under alkaline conditions, with ethanedithiol added to modify the double bond. Following ethanedithiol addition, the resultant free thiol was used as the attachment point for a biotin affinity tag using an alkylating agent linked to biotin. Tagged proteins were enriched on a monomeric avidin column, then eluted for MS analysis. This approach was limited by the inefficient recovery of tagged peptides from the affinity column, and instability of portions of the affinity tag under MS fragmentation conditions.
The β-elimination strategy was improved later by replacing the biotin affinity tag with an activated thiol affinity resin (7). Following β-elimination and Michael addition with ethanedithiol or dithiothreitol, modified peptides were captured via disulfide exchange on the activated thiol affinity column, then eluted with dithiothreitol.
The improved protocol increased sensitivity from nanomole to subpicomole levels. However, the method suffers from two limitations. First, a small proportion (1–2%) of nonphosphorylated serine residues undergoes β-elimination. This side reaction can be minimized by using dithiothreitol in place of ethanedithiol for the addition reaction, and by reducing the concentration of base used for the elimination reaction. The addition of ethylene diamine tetraacetic acid (EDTA) to chelate trace metals reduces the side reaction 10-fold, but also reduces the reaction efficiency of phosphopeptides. The second limitation of this enrichment strategy is the conversion of O-glycosylated serines to dehydroalanine under the enrichment conditions.
The phosphoramidate chemistry (PAC), developed by the Aebersold laboratory (8,9) enables phosphopeptide enrichment by introducing a reversible covalent linkage to capture phosphopeptides on a solid support. The original PAC approach (8) used a six-step procedure (Figure 2). In the first step, amino groups are blocked by reaction with t-butyldicarbonate. The second step is a carbodiimide-catalyzed condensation of carboxyl groups and phosphate groups with ethanolamine to yield amide and and phosphoramidate groups, respectively. Next, the phosphoramidate bonds are hydrolyzed under acidic conditions to regenerate free phosphate groups. The fourth step is a carbodiimide-catalyzed condensation of phosphates with cystamine, followed by reduction with dithiothreitol to yield a free sulfhydryl group for every phosphate group. In the fifth step, the modified phosphopeptides are captured on a solid support carrying immobilized iodoacetyl groups. Finally, free phosphopeptides are recovered by cleavage from the support with trifluoroacetic acid. This step also removed the t-butyldicarbonate protecting groups, regenerating free amino groups. The carboxyl groups (on the C-terminus and on aspartate and glutamate side chains) remain blocked from the condensation reaction in step 2. This provides an opportunity for stable-isotope labeling of phosphopeptides by the use of isotope-tagged ethanolamine. The PAC approach is applicable to phosphopeptides containing phosphoserine, phosphothreonine, and phosphotyrosine residues. A disadvantage is the low (albeit reproducible) yield of phosphopeptides from the multistep procedure.
A simplified phosphopeptide enrichment method based upon PAC uses a soluble amine-containing dendrimer as a solid support (9). The dendrimer has a tree-like structure with repetitive branches and a terminal surface highly functionalized with amines (Figure 3). The soluble dendrimer provided much higher yields of phosphopeptides compared to an amino-functionalized solid-phase support. This reflected the ability of the large excess of surface amines on the dentrimer to drive the extremely slow PAC reaction. The dendrimer-based enrichment procedure includes three steps. In the first step, peptide carboxylate groups are converted to methyl esters by incubation in anhydrous methanolic hydrochloric acid. This step protects carboxylates during the subsequent steps. Also, methylation provides an opportunity for stable-isotope labeling using isotopically tagged methanol. In the second step, methylated peptides are reacted with the dendrimer in the presence of carbodimide and imidazole. In this step, phosphate groups condense with the dendrimer amines to form phosphoramidate bonds. In the third step, nonphosphopeptides are removed by filtration through a 5-kDa cutoff filter while the dendrimer with immobilized phosphopeptides is retained. Finally, free phosphopeptides are cleaved from the dendrimer by acid hydrolysis of the phosphoramidate bonds.
Immobilized Metal Affinity Chromatography
IMAC is based upon the formation of reversible coordination complexes between amino acid side chains and positively charged metals bound to chelating groups immobilized on a support. The specificity depends upon the chelated metal and the binding conditions. The high affinity of the phosphate groups of phosphoproteins for IMAC supports containing bound Fe3+ or Ga3+ was demonstrated by Andersson and Porath (10). Currently, IMAC is the most popular technique for phosphopeptide enrichment for MS-based phosphoproteome studies (1). Binding is usually optimal under acidic conditions, and elution is achieved using phosphate buffer or ammonia (11). Commercial IMAC materials for phosphopeptide enrichment are available currently in a variety of formats, including microcolumns, spin columns, micropipette tips, magnetic beads, and planar chip formats for MALDI. A limitation of IMAC is the nonspecific binding of nonphosphorylated peptides via the carboxylate groups of aspartic acid, glutamic acid, and C-terminal residues. Increased specificity for phosphopeptide enrichment can be achieved by O-methyl esterification of carboxylate groups (12). A characteristic of IMAC is its tendency to enrich multiphosphorylated peptides (12, 13)
Enrichment on Metal Oxides
Enrichment of phosphoproteins and phosphopeptides on metal oxide materials such as TiO2 has emerged as an alternative to the use of IMAC (14). When compared with IMAC, it has been shown that TiO2 has higher capacity and better selectivity for phosphopeptides (15). Both off-line enrichment on microcolumns and on-line enrichment methods have been described, and commercial TiO2-based enrichment products are available commericially. As with IMAC, coenrichment of acidic nonphosphorylated peptides occurs with TiO2, and O-methyl esterification can be used to minimize this problem. However, this reaction is time-consuming and might not quantitatively block all carboxylate groups. Larsen and colleagues (15) determined that inclusion of 2,5-dihydroxybenzoic acid (DHB) in the loading buffer enhanced the selective retention of phosphopeptides on TiO2 microcolumns. This effect appeared to be due to the competition of DHB with acidic nonphosphorylated peptides for binding sites, perhaps due to the strong coordination of this aromatic acid to TiO2 sites that are preferred by nonphosphorylated peptides. The addition of DHB did not seem to affect binding or elution of phosphopeptides on TiO2.
The efficiency and reproducibility of phosphopeptide enrichment methods is dependent upon multiple variables, many of which are difficult to control in manual enrichment methods. Pinkse and colleagues (16) have developed an automated on-line enrichment method based upon TiO2, which is interfaced directly with reversed-phase LC–electrospray ionization (ESI)–MS-MS. The system consists of a vented three-segment "sandwich" precolumn coupled directly to the nanoLC analytical column. The precolumn consists of a TiO2 enrichment column sandwiched between two C18 reversed-phase segments. Initially, the peptide digest sample is trapped on the first C18 segment at a high flow rate (3 μL/min) using a weak solvent (aqueous acetic acid–formic acid) in the vented configuration. Then peptides are eluted onto the enrichment column under nanoflow conditions (100 nL/min) using a water–acetonitrile gradient. Phophopeptides bind effectively to the enrichment bed under these conditions, and nonphosphopeptides are resolved on the analytical column. Elution of the phosphopeptides then is achieved by injection of 30 μL of 250 mM ammonium hydrogen bicarbonate (pH 9). To improve the yield of phosphopeptides during elution, the eluent is supplemented with sodium phosphate, sodium orthovanadate, and potassium fluoride. This is followed by an injection of 20 μL of 5% formic acid, which serves to regenerate the TiO2 bed for the next analysis. Finally, eluted phosphopeptides are resolved by a second water–acetonitrile gradient. The initial C18 trapping step desalts the peptides and provides optimal TiO2 loading conditions. This permits use of smaller TiO2 bed volumes and lower elution volumes. A limitation of this method is its incompatibility with the use of DHB to reduce nonspecific binding of nonphosphopeptides to the enrichment column (DHB in the on-line method would contaminate the nanoESI source). However, the reduction in flow rate during loading and inclusion of acetic and formic acid in the reversed-phase solvents appeared to minimize nonspecific binding. In a comparison of the on-line method with off-line TiO2 enrichment using DHB in the analysis of a cell lysate, the authors found that, although comparable numbers of phosphopeptides were identified with the two methods, there was only about 30% overlap. This suggested that the two methods accessed different portions of the phosphoproteome.
The application of ZrO2 to enrichment of phosphopeptides has been described by Kweon and Hakansson (17). Zirconium dioxide has amphoteric properties and can behave as a Lewis acid or base, depending upon solution pH. Under acidic conditions, ZrO2 behaves as a Lewis acid with anion-exchange properties. High binding selectivity for phosphopeptides was observed at pH 2. The selective affinity for phosphopeptides diminished with increasing pH, resulting in coenrichment of acidic nonphosphopeptides. Elution of phosphopeptides was optimal around pH 11. Using the same enrichment conditions, ZrO2 was compared to TiO2. Both were found to be more selective for phosphopeptides than IMAC, and ZrO2 was found to be more selective than TiO2 for singly charged phosphopeptides.
The use of a porous anodic alumina (PAA) membrane for phosphopeptide enrichment has been reported by Wang and colleagues (18). This material is self-ordered nanochannel aluminum oxide formed by anodization of alumina in multiprotic acid solutions (for example, phosphoric, sulfuric or oxalic acid). The structure of PAA is a packed array of columnar hexagonal cells with central cylindrical cavities with pore diameters of 4–200 nm and pore densities ranging from 108 to 1011 pores/cm2 . The authors found that PAA membranes prepared by anodization in phosphoric acid were able to specifically adsorb and selectively enrich phosphopeptides. Enrichment was performed by immersing sections of membrane in a peptide solution, then using the PAA membrane as a target for MALDI-TOF MS.
Comparative Performance of Enrichment Methods
Looking at the various methods in use for phophoproteome enrichment, the question arises as to the efficacy of these methods for quantitative phosphoproteome analysis. To answer this, the Aebersold group (19) has compared PAC, IMAC, and TiO2 enrichment methods for analysis of a D. melanogaster cell lysate. For the PAC and IMAC methods, peptides from the digested lysate were subjected to methyl esterification to suppress nonspecific binding of acidic nonphosphopeptides. For the TiO2 method, nonspecific binding was quenched by the addition of phthalic acid or DHB. To avoid the problem of undersampling by LC–MS-MS, the LC–MS patterns of peptides isolated by each method were visualized and compared. This experiment demonstrated that each method reproducibly isolated a subset of phosphopeptides, but the overlap between methods was only about 35%. Thus, none of the methods provides a comprehensive map of the Drosophila phosphoproteome. The poor overlap was shown not to be due to the presence of nonphosphopeptides. Next, the same sample set was subjected to LC–MS-MS analysis. This experiment also demonstrated that only about 34% of the phosphorylation sites were common among the three methods (Figure 4). The results also demonstrated that 95% of the phosphorylation sites identified using TiO2 with DHB as the quenching agent also were found using TiO2 with phthalic acid. Thus, the quenching agent affects phosphopeptide specificity but not selectivity. This study confirmed that IMAC has a bias towards enrichment of multiply phosphorylated peptides, and that the isolated phosphopeptides tend to be larger than the peptides of the unfractionated proteome. The latter effect was attributed to the tendency of phosphopeptides to have more missed cleavages in trypsin digestion, and to the loss of small, hydrophilic phosphopeptides in the reversed-phase cleanup steps.
A quantitative comparison of a variety of commercial and prototype enrichment products using iTRAQ (Applied Biosystems, Foster City, California) stable isotope labeling was performed by Liang and colleagues (20). In this study, a mixture of peptide standards containing nonphosphorylated peptides and phosphopeptides containing phosphoserine, phosphothreonine, and phosphotryosine residues was labeled with each of four iTRAQ reagents. Each iTRAQ-labeled mixture was bound and eluted from one of several enrichment products, including metal chelating groups attached to magnetic beads, metal-oxide beads, and metal-oxide coated magnetic beads. Metal chelating groups with Fe3+ or Ga3+ ions were evaluated. After elution, the eluates were mixed pairwise and subjected to quantitative analysis using MALDI-TOF-TOF MS-MS. To eliminate bias, crossover studies in which the iTRAQ reagents were switched for each pair were performed. Results (Table I) indicated that the greatest capture efficiency was achieved using a Fe3+ metal chelate resin coupled to magnetic beads (Fe3+ -MMC). This material outperformed the Ga3+ -coupled version, Fe3+ - or Ga3+ -coupled IDA-coated magnetic particles, Fe3+ -metal chelate beads, and ZrO2 metal oxide beads. The Fe3+ -MMC performed similarly to TiO2-coated magnetic beads and TiO2 spheres in terms of efficiency and contamination by nonphosphopeptides, even when the metal oxide products were used with DHB as a quenching agent.
Combining IMAC With Other Enrichment Techniques
Obtaining comprehensive coverage of a phosphoproteome can be beyond the grasp of a single enrichment method, and the combination of prefractionaton techniques with an enrichment step, or the sequential coupling of multiple enrichment steps is sometimes (perhaps always) necessary. In the case of phosphotyrosines (which account for 1% or less of phosphoproteomes), an immunoprecipitation step using antiphosphotyrosine antibodies has been followed with IMAC as a secondary enrichment step (21). For global phosphoproteome analysis, both strong cation exchange and strong anion exchange chromatography have been used for prefractionation before IMAC enrichment. Strong cation exchange at low pH typically is used as a general fractionation technique for peptides. Under these conditions, phosphopeptides will be protonated weakly, and should be eluted earlier than the majority of nonphosphoryalted peptides. Gruhler and colleagues (22) have taken advantage of this to prefractionate complex digests, and enrich phosphopeptides in the early eluted strong cation exchange fractions with IMAC-Fe3+ columns. A more intuitive approach for phosphopeptides is the use of strong anion exchange chromatography. This strategy was used by Nuhse and colleagues (13) to fractionate phosphopeptides prior to IMAC enrichment. It was observed that strong anion exchange tends to fractionate phosphopeptides according to their negative charge, with monophosphorylated peptides eluted at low salt concentrations and multiply phophoryated peptides eluted at higher ionic strength.
A method for phosphopeptide enrichment based upon calcium phosphate precipitation was developed by Zhang and colleagues (23). This was based upon the observation that, in the presence of sodium phosphate, the addition of calcium chloride resulted in coprecipitation of phosphopeptides as their calcium salts. Although the method provided phosphopeptide enrichment with high selectivity and sensitivity for a digest of a model phophoprotein (casein), the efficiency of enrichment was less satisfactory when phosphopeptides were in low abundance in a complex digest of multiple proteins. A subsequent enrichment of phosphopeptides on an IMAC-Fe3+ microcolumn following calcium phosphate precipitation improved overall enrichment efficiency. When the method was applied to phosphopeptide analysis of a rice embryo extract, highest recovery of phosphopeptides required calcium phosphate precipitation followed by two serial IMAC-Fe3+ enrichment steps.
Recognizing the bias of TiO2 and IMAC for monophosphorylated and mutiply phosphorylated peptides, respectively, Thingholm and colleagues (24) have developed a sequential enrichment strategy employing both materials. This strategy, termed SIMAC (for sequential elution from IMAC) is outlined in Figure 5. The strategy takes advantage of the elution of monophosphorylated peptides from IMAC under acidic conditions. A peptide digest is bound to IMAC beads, which are then packed into a microcolumn for washing and elution. The unbound peptides in the IMAC flowthrough and wash are subjected to enrichment on TiO2 to capture unbound or poorly retained monophosphorylated peptides. The remainder of the monophosphorylated peptides are eluted from the IMAC bed using 1% trifluoroacetic acid, and these also are captured on TiO2. Finally, multiply phosphorylated peptides are eluted from the IMAC column using dilute ammonium hydroxide. The SIMAC strategy was evaluated and compared to an optimized TiO2 enrichment method using a stem cell lysate digest. The results demonstrated that the SIMAC method yielded more than twice as many phosphorylation sites overall, with a threefold increase in the recovery of multiply phosphorylated peptides, indicating that SIMAC provides a more complete coverage of the phosphoproteome than either technique alone. By separating the multiply phosphorylated peptides, the problem of their ion suppression by nonphosphorylated and monophosphorylated peptides is diminished.
A variety of techniques has been used to enrich phosphoproteins or phosphopeptides from the global proteome. These include antibody-based immunoaffinity precipitation, chemical modification to introduce groups for affinity purification, enrichment on immobilized metal affinity materials, enrichment on a variety of metal oxides, coupling of ion exchange prefractionation with IMAC, or coupling IMAC and metal oxide enrichment. From comparative studies of the performance of these methods, it is clear that the overlap between them is poor, suggesting that each technique isolates a different subset of the phosphoproteome. The coupling of multiple enrichment techniques appears to be the most promising strategy to comprehensively characterize a phosphoproteome. There is also evidence that digests produced using an enzyme with relatively frequent cleavage sites (for example, trypsin) will bias results against shorter and more hydrophilic phosphopeptides that are lost in desalting steps. So a companion strategy in phosphoproteome characterization should be a multipronged attack using more than one protease. Finally, there remains the problem of knowing when the goal is reached: with no benchmark phosphoproteome that is known to be characterized comprehensively, it is difficult to truly validate an enrichment strategy.
"Directions in Discovery" editor Tim Wehr is staff scientist at Bio-Rad Laboratories, Hercules, California. Direct correspondence about this column to Direct correspondence about this column to "Directions in Discovery," LCGC, Woodbridge Corporate Plaza, 485 Route 1 South, Building F, First Floor, Iselin, NJ 08830, e-mail email@example.com
(1) S. Morandell, T. Stasyk, K. Grosstessner-Hain, E. Roitinger, K. Mechtler, G.K. Bonn, and L.A. Huber, Proteomics 6, 4047–4056 (2006).
(2) F. Delom and E. Chevet, Proteome Sci. 4, 15 (2006).
(3) Pandey, A.V. Podtelejnikov, B. Blagoev, X.R. Bustelo, M. Mann, and H.F. Lodish, Proc. Nat. Acad. Sci. USA 97, 179–184 (2000).
(4) G. Zhang and T.A. Neubert, Proteomics 6, 571–578 (2006).
(5) J. Rush, A. Moritz, K.A. Lee, A. Guo, V.L. Gross, E.J. Spek, H. Zhang, X.-M. Zha, R.D. Polakiewicz, and M.J. Comb, Nat. Biotechnol. 23, 94–101 (2005).
(6) Y. Oda, T. Nagasu, and B.T. Chait, Nat. Biotechnol. 19, 379–382 (2001).
(7) D.T. McLachlin and B.T. Chait, Anal. Chem. 75, 6826–6836 (2005).
(8) H. Zhou, J.D. Watts, and R. Aebersold, Nat. Biotechnol., 19, 375–378 (2001).
(9) W.A. Tao, B. Wollscheid, R. O'Brien, J.K. Eng, X. Li, B. Bodenmiller, J.D. Watts, L Hood, and R. Aebersold, Nature Methods 2, 591–598 (2005).
(10) L. Andersson and J. Porath, Anal. Biochem. 154, 250–254 (1986).
(11) M.C. Posewitz and P. Tempst, Anal. Chem. 71, 2883–2892 (1999).
(12) S.B. Ficarro, M.L. McCleland, P.T. Stukenberg, D.J. Burke, M.M. Ross, J. Shabanowitz, D.F. Hunt, and F.M. White, Nat. Biotechnol. 20, 301–305 (2002).
(13) T.S. NFChse, A. Stensballe, O.N. Jensen, and S.C. Peck, Mol. Cell. Proteomics 2, 1234–1243 (2003).
(14) M.W. Pinkse, P.M. Uitto, M.J. Hilhorst, B. Ooms, and A.J. Heck, Anal. Chem. 76, 3935–3943 (2004).
(15) M.R. Larsen, T.E. Thingholm, O.N. Jensen, P.Roepstorff, and T.J.D. JF8rgensen, Mol. Cell. Proteomics 4, 873–886 (2005).
(16) M.R. Pinkse, S. Mohammed, J.W. Gouw, B. van Breukelen, H.R. Vos, and A.J.R. Heck, J. Proteome Res. 7, 687–697 (2008).
(17) H. K. Kweon and K. Håkansson, Anal. Chem. 78, 1743–1749 (2006).
(18) Y. Wang, W. Chen, J. Wu, Y. Guo, and X Xia, J. Am Soc. Mass Spectrom. 18, 1387–1395 (2007).
(19) B. Bodenmiller, L.N. Mueller, M. Mueller, B. Domon, and R. Aebersold, Nature Methods 4, 231–237 (2007).
(20) X. Liang, G. Fonnum, M. Hajivandi, T. Stene, N.H. Kjus, E. Ragnhildstveit, J.W. Amshey, P. Predki, and R.M. Pope, J. Am. Soc.Mass Spectrom. 18, 1932–1944 (2007).
(21) L.M. Brill, A.R. Salomon, S.B. Ficarro, M. Mukherji, M. Stettler-Gill, and E.C. Peters, Anal. Chem. 76, 2763–2772 (2004).
(22) A. Gruhler, J.V. Olsen, S. Mohammed, P. Mortensen, N.J. Færeman, M. Mann, and O. Jensen, Mol. Cell. Proteomics 4, 310–327 (2005).
(23) X. Zhang, J. Ye, O.N. Jensen, and P. Roepstorff, Mol. Cell Proteomics 6, 2032–2042 (2007).
(24) T.E. Thingholm, O.N. Jensen, P.J. Robinson, and M.R. Larsen, Mol. Cell Proteomics DOI M700362-MCP200 (2007).