Sample Preparation and Enrichment Strategies for Phosphoprotein Analysis by Mass Spectrometry
A generalized phosphoproteomics strategy as used in the author's lab that can be adapted to any particular experiment is shown in Figure 2.7. Variations of this general strategy are in use in many laboratories worldwide. Employing peptide-based fractionation, phosphopeptide enrichment, and high- performance MS, numerous labs are now able to routinely detect and quantify 10,000-20,000 unique phosphorylation sites in single large-scale phosphoproteomics experiment [33, 135, 167]. Despite this effort, these numbers may still be considerably lower than the actual phosphoproteome complement of a mammalian cell . Full and accurate coverage of the phosphoproteome remains a daunting and unrealized analytical challenge, with significant areas for improvement.
Protein phosphorylation is a reversible, highly dynamic PTM regulated by the interplay of opposing enzymatic activities. As such, particular care must be taken to preserve the integrity of the cellular phosphoproteome throughout the process of sample treatment and harvest, preparation of protein extracts, proteolytic digestion, and phosphopeptide capture and analysis. For example, precautions must be taken during cell harvesting to avoid induction of kinase pathways associated with changes in osmolarity, temperature, or nutrient availability. Cultured cell pellets or tissue samples should be snap-frozen in liquid nitrogen, stored frozen at -80 ° C, and processed on ice. Cocktails of
Figure 2.7 Generalized workflow for quantitative phosphoproteomics. The choice of quantitative strategy will depend upon the type of samples being analyzed. See the text for details. After lysis, (a) SILAC-labeled samples are mixed 1:1 and digested with trypsin. (b) Samples to be labeled with isobaric TMT reagents are digested first, then labeled, and mixed 1:1. HILIC chromatography of the protein digests separates phosphopeptides from the bulk of the nonphosphorylated peptides. The less complex phosphopeptide-containing fractions are enriched in 96-well plates using Fe3+-IMAC. The resulting purified phosphopeptide mixtures are separated by nanoLC and analyzed directly by ES interfaced to a high-resolution, high-mass-accuracy mass spectrometer.
broad-spectrum protease and phosphatase inhibitors are routinely included during cell lysis to prevent protein degradation and dephosphorylation.
Protein lysates are routinely prepared under denaturing conditions using detergents or chaotropes to irreversibly halt protein enzymatic activity and to solubilize the largest portion of the cellular proteome possible. Because in general detergents have a deleterious effect on MS performance, proteins in detergent lysates are often separated on SDS-PAGE followed by proteolytic digestion of the gel-embedded proteins. “In-solution” digestion under denaturing conditions may be performed when proteins are extracted with strong chaotropic reagents such as urea and thiourea. A method describing filter-aided sample preparation (FASP) that facilitates buffer exchange to combine the advantages of detergent extraction with in-solution digestion has been frequently employed in MS-based proteomics . Care must be taken when using this approach with the large-scale lysate requirements typical of proteomics experiments that seek to analyze PTMs . Several MS-compatible detergents have been introduced for enhanced protein solubilization and digestion .
Although trypsin is most commonly used in the MS analysis of proteins, tryptic phosphopeptides may not always possess sequence properties suitable for phosphosite localization by MS. This is especially the case for phosphoryl- ated residues, which lie within basic amino acid-rich protein domains. Several studies have shown that using multiple proteases on the same sample can increase proteome [172, 173] and phosphoproteome coverage significantly [130, 132, 174]. For large-scale phosphoproteomics analysis, thousands of phosphopeptide sequences were identified that would have been undetectable when using only trypsin, and fewer than a third of the phosphosites identified were detected in more than one protease data set . However, using multiple protease digests drastically increases instrument time and requires a larger amount of sample. In this case, using five proteases required five times the amount of sample as would be required for a study with just trypsin and increases analysis time fivefold as well. These caveats may make this approach untenable for some studies. As with any proteomics tool, one has to weigh the purpose of the study, the benefits made by the approach, and instrument time restrictions to fully assess the best strategy.
Numerous strategies have been developed to enrich phosphopeptide from the overwhelming number of nonphosphorylated peptides in a complex sample. The most successful of these incorporate chemoaffinity and immu- noaffinity strategies, either on intact phosphoproteins or phosphopeptides derived from proteolytic digests of proteins. Chemical modification and deri- vatization approaches have also been tested to enhance the selectivity of enrichment strategies.
Immunoaffinity approaches using antibodies have been most successfully applied to analyzing tyrosine phosphorylation. Tyrosine phosphorylation constitutes only <2% of the total cellular protein phosphorylation ;
thus, specialized enrichment procedures are necessary to isolate phosphotyrosine-containing peptides from both unmodified peptides and the large pool of serine- and threonine-phosphorylated peptides. Antibody- based affinity purification procedures for phosphoproteomics studies involve selective enrichment of proteins or peptides that contain epitopes comprising single phosphoamino acid residues, or phosphorylated residues positioned within short linear sequence-dependent recognition motifs. Antibodies raised to specifically recognize phosphotyrosine residues in macromolecules have been in use for over 30 years . Several highly specific, fully characterized, and validated antibodies are commercially available, which are well suited for large-scale study of phosphotyrosine signaling pathways in cell cultures and tissue extracts . Due to the low abundance of pTyr, immunoprecipitation with anti-pTyr antibodies typically requires much larger amounts (>10 mg) of protein extract for large-scale, in-depth studies .
Enrichment with anti-pTyr antibodies for analysis by MS has been applied at both the protein and peptide level. Several studies have focused on quantitative elucidation of the temporal profile of phosphotyrosine-mediated signal transduction following various growth factor stimulations. Although immuno- precipitation of tyrosine-phosphorylated peptides was reported as early as 1995 , most of the early work applying immunopurification to tyrosine- phosphorylated proteins was done at the protein level [180-185]. This approach requires lysis under nondenaturing conditions, which facilitates copurification of protein complexes comprising phosphotyrosine-tethered scaffolds and signaling adaptors. If done quantitatively, this approach enables the elucidation of temporal recruitment and assembly dynamics of complete phosphotyrosine signaling modules. A disadvantage of protein-level enrichment is that relatively few of the resulting proteolytic peptides will contain the phosphotyrosine modification. To identify particular sites of phosphorylation, the antiphosphotyrosine immune complex can be digested with trypsin and enriched for phosphopeptides using metal affinity resins (see following text). In contrast, phosphotyrosine enrichment at the peptide level has the advantage that lysis and protein extraction can be performed under harsher conditions, allowing better solubilization of integral membrane signaling proteins and deeper coverage. Additionally, the majority of eluted peptides will be expected to contain the phosphotyrosine modification . Large-scale studies employing peptide-level immunoaffinity enrichment have allowed the characterization of several hundred to thousands of phosphotyrosine sites [178, 187-189].
Many commercial antibodies are available that claim to show specificity for phosphorylated serine and phosphorylated threonine residues. The generally poor selectivity of these reagents as immunopurification tools has limited their use, though motif-specific antibodies have been effectively employed
[190-192]. In general, the global study of serine and threonine phosphorylation is better accomplished using chemoaffinity enrichment techniques.
The two most frequently used strategies for global phosphopeptide enrichment are both chemoaffinity based and include various forms of immobilized metal ion affinity chromatography (IMAC) and metal oxide affinity chromatography (MOAC) where the most common reagent is TiO2. Both have been applied with a great deal of success in a variety of permutations, alone or together using proteolytic peptides from purified proteins or crude cellular extracts, and frequently coupled with up-front chromatographic prefractionation. The binding, wash, and elution steps are most commonly performed offline from the mass spectrometer, in tubes, plates, pipette tips, or microcolumn format.
Immobilization of multivalent cations on an affinity resin support has been widely utilized in protein chemistry, exemplified by the use of Ni2+ for the purification of recombinant proteins engineered to contain a polyhistidine affinity tag. IMAC was introduced for the separation of phosphoproteins in 1986  and later adapted and refined for phosphopeptide enrichment [194, 195]. Enrichment of phosphopeptides by IMAC exploits the affinity of phosphate groups for multivalent transition metal ions, such as Fe3+, Ti4+, Ga3+, and Zr4+. IMAC uses metal chelators such as iminodiacetic acid (IDA) and nitrilotriacetic acid (NTA) linked to solid-phase chromatographic supports to coordinate the metal ion. Fe3+ is the basis for the most common form of IMAC. Available coordination sites of the positively charged metal ions are presented for interaction with the negatively charged phosphate groups on peptides.
One of the major shortcomings of IMAC is its affinity for highly acidic peptides rich in glutamic and aspartic acid residues, which also coordinate well with metal complexes. An early solution to nonspecific binding was chemical derivatization to convert carboxylic acids to their corresponding O-methyl esters prior to IMAC . This approach suffered from incomplete esterification and deamidation of asparagine and glutamine-containing peptides, which increased the sample complexity and complicated the sequence database search space. Rather, systematic investigations into modulation of pH, acid content, and organic modifiers led to the findings that a low-pH (2.0-2.5) loading solution improved selectivity by protonating the acidic amino acids while maintaining phosphate groups unprotonated and available for binding. Additionally, binding in the presence of trifluoroacetic acid (TFA) and acetonitrile (ACN) minimized hydrophobic interactions between peptides and the IMAC resin [196-198]. The enriched peptides are most commonly eluted with basic pH buffers such as ammonium hydroxide, but can also be eluted with phosphoric acid or EDTA. An additional shortcoming of IMAC enrichment is that it is generally intolerant of salts and detergents, which can interfere with phosphopeptide binding .
IMAC has been reported to bind multiply phosphorylated peptides with higher affinity than singly phosphorylated peptides, producing a bias in phosphopeptide enrichment . This appears to be the case for complex, unfractionated peptide digests from whole lysates. This bias all but disappears when the sample complexity is reduced, either through prefractionation and segregation of singly and multiply phosphorylated peptide pools prior to enrichment  or through sequential rounds of IMAC [198, 201]. Recently, Ti4+-IMAC has been reported to be superior in terms of specificity and efficiency to both IMAC using other metals or TiO2 chromatography, particularly when applied to crude enrichment from unfractionated whole cell lysates. [202-204]. This has proven highly advantageous when characterizing cellular phosphoproteomes via performing single enrichments and subsequent MS analyses on numerous samples, as would be the case for label-free quantitative studies [167, 205].
MOAC takes advantage of the ability of some metal oxides, such as TiO2 and ZrO2, to form complexes with phosphate groups. It was reported by multiple groups for the selective enrichment of phosphopeptides in 2004 [206-208]. As with IMAC, nonspecific binding of nonphosphorylated peptides rich in acidic residues is a persistent challenge. To circumvent this, loading under highly acidic conditions and the inclusion of various substituted organic acids have been found to alleviate nonspecific binding. Organic acid additives such as 2,5-DHB  bind metal oxides more weakly than the phosphate groups, but more strongly than carboxyl groups, through a chelating bidentate mode rather than the bridging bidentate mode exhibited for phosphate group binding. Poor solubility of 2,5-DHB has prompted its substitution with more soluble and hydrophilic acids such as glycolic, phthalic, and lactic acids [199, 210, 211]. A disadvantage is that a desalting step is usually required to remove these additives prior to MS. As is the case with IMAC, the nonspecific binding of highly acidic peptides goes away when complex samples are fractionated prior to enrichment. Advantages of MOAC include the superior chemical stability of TiO2 spherical particles. Unlike IMAC, there is no need to charge the resin with metal ions, and the support exhibits exceptional tolerance to buffer excipients such as salts, detergents, and chelating agents .
Is one of the affinity capture technique inherently better than the others for phosphopeptide enrichment? Numerous, often conflicting, reports have claimed the superiority of either IMAC or TiO2 with respect to phosphopep- tide capture performance, as judged by efficiency, selectivity, and recovery. The differences observed between the two resins are often so pronounced as to call into question the experimental design of the studies. Most often a lack of experience in one method or the other combined with undersampling of the enriched pools easily accounts for the very large differences observed. In fact, numerous large-scale phosphoproteomics studies using IMAC [33, 167] and MOAC [135, 212] enrichment demonstrate that, judged by overall performance,
both strategies are highly effective and are capable of similar numbers of phosphosite identifications, particularly when coupled with a multidimensional separation strategy. Unlike what has been reported in earlier studies on limited sample sets, both methods capture similar numbers of singly and multiply phosphorylated peptides.
A related consideration under debate is the perceived orthogonality of the two techniques. The prevailing consensus in the field is that IMAC and TiO2 methods are capable of enriching complementary parts of the phosphopro- teome, with each technique showing particular strengths in its ability to isolate unique classes of phosphopeptides possessing distinct physicochemical properties [201, 213-216]. This would suggest that neither method alone is sufficient for a comprehensive enrichment of the phosphoproteome. Several recent studies and work in our own lab have called this assumption into question. Comparison of TiO2 and Ti4+-IMAC using a very large synthetic phosphopeptide library of known composition, and tryptic lysates of HeLa cells, indicated minimal differences between enrichment of phosphopeptides based on sequence composition, peptide length, or various physicochemical properties . In a direct comparison of rat liver data sets (Figure 2.8a) obtained from different laboratories employing different phosphopeptide isolation strategies centered on either IMAC  or TiO2 , fully 84% of the TiO2 sites were contained within the IMAC data set , and there was no difference in the preference for multiply phosphorylated peptides between the two methods. Likewise, Ruprecht et al. observed that optimized Fe3+-IMAC, Ti4+-IMAC, and TiO2 platforms bound the same phosphopeptide species and concluded that insufficient resin capacity, inefficient elution conditions, and the stochastic nature of data-dependent acquisition MS are the causes of the experimentally observed complementarity between platforms .
Unpublished data from our laboratory are in full agreement with these findings. Using isobaric labeling to quantify differences between IMAC and TiO2 enrichment of hydrophilic interaction liquid chromatography (HILIC) fractions, we find that both resins isolate the same pool of phosphopeptides, irrespective of the number of phosphorylation sites and other physicochemical properties (Figure 2.8b). In fact, only approximately 10% of phosphopeptides were found to be enriched greater than fourfold by one technique or the other, suggesting that in an optimized platform, there are only marginal gains to be made by a combination or sequential IMAC and TiO2 capture. Thus it is clear that both strategies can be highly effective, especially when coupled with a multidimensional separation strategy; however, it is equally clear that both IMAC and MOAC need to be optimized individually and for specific workflows.
All of the methods described earlier utilize the intact phosphate group to facilitate enrichment of phosphopeptides. Historically, several innovative methods have been introduced that use chemical modification as a way to improve the enrichment of phosphopeptides. A widely used derivatization
Figure 2.8 TiO2 and Fe3+-IMAC isolate the same phosphopeptides. (a) Rat liver phosphosites enriched by either HILIC-IMAC  or TiO2  show extensive overlap. Despite different phosphopeptide enrichment strategies conducted in two independent laboratories, 84% of phosphopeptides from the TiO2 experiment are shared with the IMAC data set. Contrary to conventional thinking, no difference in the proportion of singly and multiply phosphorylated peptides recovered was observed between Fe3+-IMAC and TiO2. (b) Tryptic peptides from two adjacent HILIC fractions were divided into six equal parts and enriched three each on either Fe3+ IMAC or TiO2. The six enriched pools were labeled with TMT to produce a 6plex quantitative experiment. Individual reporter ion intensities indicate the relative enrichment of a given peptide from each of the six pools. The scatterplot (left) shows very good correlation between the log2 mean intensities of 2765 unique phosphopeptides enriched by either method. The histogram (right) displaying binned log2 ratios for Fe3+ IMAC versus TiO2 indicates that only approximately 10% of phosphopeptides are enriched >fourfold by one method over the other.
protocol is based on combining p-elimination of the phosphate group under strongly basic conditions, followed by a Michael addition on the resulting dehydroalanine or dehydroamino-2-butyric acid products [218-220]. The resulting side chain permits the attachment of a variety of affinity tags with the added option of being able to incorporate stable isotopes into the tag for
quantification. The method is not suitable for tyrosine-phosphorylated peptides, which do not undergo p-elimination. The method also suffers from chemical side reactions and incomplete conversion. Another chemical deri- vatization strategy uses phosphoramidate chemistry where phosphopeptides can be linked to iodoacetyl groups immobilized on a synthetic polymer solid support or glass beads. Acid hydrolysis of the phosphoramidate bonds allows the phosphate groups to be recovered intact [221, 222]. Though amenable to pTyr as well as pSer- and pThr-containing peptides, the extensive workflow involves several chemical reactions and methyl esterification, with the associated sample losses and increased sample complexity. With the maturation of efficient chemoaffinity enrichment techniques described earlier, chemical derivatization approaches have largely fallen out of favor for phosphoproteome analysis.