Interrogating Protein Phosphorylation and PTMs with Mass Spec


Post-translational modifications (PTMs) modulate the physical and chemical attributes of proteins and act as an information currency. Understanding how that information is stored and what it means should be a boon for research into systems biology and mechanism of action, and it may even provide a window into a patient’s state of health or disease, as well as the best treatments for a condition. That understanding will most certainly rely on identification and exploration of biomarker signatures composed at least in part of phosphorylated proteins.


Mass spectrometry (MS) has revolutionized how proteomics, including phosphoproteomics, is being performed.

It’s now possible, for example, to scan samples for tens of thousands of phospho-specific sites on thousands of proteins in a single sample. What these resulting signature profiles tell us is being actively investigated by biological, medical and pharmaceutical researchers alike.

Researchers have historically used various ways to characterize phosphorylated proteins. Antibodies against context-free phosphorylated serine, threonine and tyrosine residues are widely available, and an increasing number of antibodies that will recognize specific phosphorylated sequences also can be found. These can be used in ELISA-like assays, for example, as well as in forward- and reverse-phase protein arrays, flow cytometry, Western blotting and immunohistochemistry.

Yet there are a number of limitations on such immunologically based methods. They are generally not suited for large-scale multiplexing. Specific and reliable antibodies to a particular phosphosite do not always exist. Even when those antibodies are available and can be used as a secondary antibody to provide specificity in a sandwich assay, “most phosphosites don’t exist in isolation—there are normally other serines, threonines, tyrosines that can be phosphorylated, plus a bunch of other amino acids that can be modified in different ways,” points out Ian Pike, chief operating officer of Proteome Sciences. “There is clearly some crosstalk between these sites which can make some antibodies bind more strongly and some less strongly.”

MS can identify post-translationally modified sites with exquisite specificity, without the need for affinity reagents. Protocols integrating techniques such as iTRAQ isobaric labeling into a discovery-mode (global) proteomics workflow can quantity 20,000 to 40,000 phosphosites in a single experiment from multiple samples, assuming that sufficient sample is available [1]. A typical workflow for large-scale phosphoproteome profiling consists of four principal steps: The protein is extracted and enzymatically digested; peptides are isobarically tagged; the sample is enriched for phosphopeptides, most commonly using either immobilized metal affinity chromatography (IMAC) or metal oxide affinity chromatography (MOAC) with TiO2; and the samples are then subject to liquid chromatography-tandem mass spectrometry (LC-MS/MS) analyses.

By coupling nanoLC to a high-resolution, fast-scanning tandem MS, as in Thermo Scientific’s Q-Exactive platforms, researchers have profiled up to 10,000 phosphopeptides from as little as 100 µg of protein starting material [2].

Such profiling enables the systematic identification of what is up- or down-regulated in normal and aberrant signaling pathways and networks. It can lead to a deeper understanding of the mechanisms responsible for pathogenesis and uncover biomarkers of disease or potential therapeutic targets, points out Wei-Jun Qian, staff scientist at the Pacific Northwest National Laboratory.

In a recent publication, a Portuguese group reported treating neuronal cortical cell cultures with the amyloidogenic Aβ peptide, derived from Alzheimer’s Amyloid Precursor Protein (APP). They found significant differences in the phosphorylation of 141 proteins, many of which are associated with Alzheimer’s Disease (AD)-related processes such as signal transduction, endocytosis, cytoskeletal organization and intracellular transport [3].

“I think it’s important to generate the whole phosphoproteome, not to focus too much on specific phosphopeptides as biomarkers—to look for signatures of pathway activities, because every patient is different,” says , professor of translational oncoproteomics in the Department of Medical Oncology of the VU University Medical Center in Amsterdam. This reveals the whole picture—how many pathways are involved in a malignancy, for example—“to really make an educated guess about whether a single inhibitor will be successful, or whether one should target multiple pathways in a combination strategy.”

Targeted phosphoproteomics
In a global, shotgun, biomarker-discovery experiment, “you want to understand what the playing field looks like,” explains Jacob Jaffe, director of the LINCS Proteomic Characterization Center for Signaling and Epigenetics at the Broad Institute. “And then make some hypotheses about what might be different between disease and (apparently) healthy conditions.” The next step—as in any clinical study—would be to increase the number of samples, while using a targeted (typically multiple-reaction monitoring, MRM) approach to narrow that playing field to only those analytes that you hypothesize may differ, to increase the statistical power.

In MRM approaches, the MS is instructed to look only for specific analytes and to ignore everything else.

“The MRM assay could be used to screen a large number—several hundred—of potential biomarkers for further validation,” says Chris Crutchfield, assistant professor of pathology and laboratory medicine at the University of Cincinnati. Stable isotopic reference standards often are spiked into the samples.

But Crutchfield cautions that the incredible specificity of MS could be a double-edged sword. An immunoassay to a particular phosphospecific site might recognize the protein that contains that site, regardless of its other PTMs. On the other hand, MS would see each variant—those acetylated vs. de-acetylated at all sites, for example—as a different species, “and the signal is going to get diluted out by all those very tiny, specifically nuanced peptide components of the protein,” according to Crutchfield. Probability-based algorithms can be used to collect and sort related peptides, enabling the generation of biomarker signatures.

What’s a biomarker, anyway?
To date, phosphoproteomics has generated some information that may be useful in the research lab to look for drug targets or understand biology, but relatively few phosphosites or signatures that will help physicians diagnose or prognose patients have been identified. It’s hard to pin down the definition of a biomarker, and Jaffe prefers instead to qualify the term as “clinical disease biomarker” or “mechanistic biomarker,” for example. “Pharma companies love to use the word ‘biomarker’ when it refers to something they can look at in cells that reports on the activity of their latest and greatest drug.”

One reason is that to date most work has been conducted using cell lines or animal models rather than human-tissue samples—partially because of the amount and numbers of sample needed for well-powered results. In addition, phosphorylations are relatively labile, and care must be taken immediately to preserve their state. A recent study looking for biomarkers of ischemia found that within two minutes of a sample being removed from a xenograft model, a number of phosphorylation sites were either significantly hyper- or hypophosphorylated (relative to samples immediately snap-frozen)—including sites on a heat-shock protein known to be involved in ischemia [4].

Another reason is that although National Institutes of Health (NIH) guidelines recommend clinical biomarkers be found in accessible bodily fluids, these are notoriously complex samples, chock-full of highly abundant proteins that could swamp out the signal from a rare phosphosite. Pike and his colleagues used isobaric tandem mass tags (TMTs, which are “exactly the same concept” as iTRAQ) to multiplex four postmortem brain-tissue samples with six cerebrospinal fluid (CSF) samples to look for biomarkers of Alzheimer’s disease, in an effort “to bridge that tissue understanding into bodily fluids,” he says [5].” The less complex tissue samples essentially force the MS to sequence common samples from the CSF, “then hopefully we can pick up these lower-abundance but mechanistically important biomarkers.”

Phosphorylation is a very dynamic process that affects nearly every cellular pathway and network, piggybacking on protein expression to add a new level of regulation and control. Even where protein levels remain steady, those proteins may be modulated in many different ways, all of which could connote something different about the underlying biological state. “Phosphosignaling is one of the important information currencies in biological systems,” says Jaffe. “If you believe that, then you have to believe that this information can be leveraged to read out disease vs. healthy states.” And mass spectrometry has emerged as a highly accurate and comprehensive way to do such leveraging.

 [1] Chan, CY, et al., “The current state of the art of quantitative phosphoproteomics and its applications to diabetes research,” Expert Rev Proteomics, 13(4):421-33, 2016. [PMID: 26960075]

[2] Jimenez, CR, Verheul, HM, “Mass spectrometry-based proteomics: from cancer biology to protein biomarkers, drug targets, and clinical applications,” Am Soc Clin Oncol Educ Book, e504-10, 2014. [PMID: 24857147] 

[3] Henriques, AG, et al., “Altered protein phosphorylation as a resource for potential AD biomarkers,” Sci Rep, 6:30319, July 28, 2016. [PMID: 27466139]

[4] Zahari, MS, et al., “Phosphoproteomic profiling of tumor tissues identifies HSP27 Ser82 phosphorylation as a robust marker of early ischemia,” Sci Rep, 5:13660, September 2, 2015. [PMID: 26329039] 

[5] Russell, CL, et al., “Comprehensive Quantitative Profiling of Tau and Phosphorylated Tau Peptides in Cerebrospinal Fluid by Mass Spectrometry Provides New Biomarker Candidates,” J Alzheimers Dis, 55(1):303-313, November 1, 2016. [PMID: 27636850]