Identification of phosphosites that alter protein thermal stability

Abstract

Proteomics has enabled the cataloguing of 100,000s of protein phosphorylation sites 1 , however we lack methods to systematically annotate their function. Phosphorylation has numerous biological functions, yet biochemically all involve changes in protein structure and interactions. These biochemical changes can be recapitulated by measuring the difference in stability between the protein and the phosphoprotein. Building on recent work, we present a method to infer phosphosite functionality by reliably measuring such differences at the proteomic scale. MAIN TEXTRecently, Huang et al. 2 developed the Hotspot Thermal Profiling (HTP) method to identify phosphosites that alter protein thermal stability, reporting 719 out of 2,883 (25%) phosphosites with significant effects. The reported melting temperatures (T m ) for phosphopeptides correlated poorly with the T m for their corresponding proteins (R 2 = 0.18) ( Fig. 1a), implying that many phosphosites function by structurally reshaping the proteome. However, the low T m reproducibility between replicates ( Supplementary Fig. 1a) suggests that this conclusion may be due to technical variation (Supplementary Discussion). The HTP workflow consists of phosphopeptide enrichment followed by separate isotopic labeling and mass spectrometric analysis to derive T m values for phosphopeptides and proteins, respectively. Because phosphopeptide samples also contained unmodified peptides, which are expected to have the same T m as the protein, we can use these peptides to assess technical variation between the two samples. Disconcertingly, our re-analysis revealed that 626 out of 3074 (20%) of the co-enriched unmodified peptides had significant stability effects, almost the same percentage as phosphopeptides (22%, 596 out of 2656 in our re-analysis) (Fig. 1b, Dataset S1). Additionally, the T m correlation of these peptides with their protein T m was similarly low ( R 2 = 0.18) to the correlation between phosphopeptides and protein ( Supplementary Fig. 1b). In the absence of a biological explanation, this suggests that the independent labeling and mass spectrometric analysis of peptide and phosphopeptide samples could have introduced substantial technical error precluding the comparison, and perhaps that the reported hits arise from a lack of stringency in the applied statistical analysis (Supplementary Discussion).

Impact of phosphorylation on thermal stability of proteins

“…It has been shown that the ratio between peptide and stationary phase is critical to achieve good enrichment selectivity, 17 (phosphopeptides representing roughly 60% of the total peptides identified, while nowadays more than 90% specificity is routinely achieved-98% in our study) is thus an indication that the way the authors performed phosphopeptide enrichment is suboptimal. Further, the enrichment selectivity of Huang et al was variable from sample to sample, as analyzed in an accompanying work by Smith et al 18 . Importantly, the authors did not validate the linearity of enrichment capacity over a broad range of peptide amount input, and therefore this step might negatively influence their results by introducing quantification biases which will impact the estimation of phosphopeptide melting points-particularly, for an approach with low amount of TiO2 stationary phase (TiO2 spin tips used by the authors, catalog no.…”

A Bayesian semi-parametric model for thermal proteome profiling

“…This reinforces our interpretation that for proteins with non-sigmoid behaviour, we are observing a mixture of behaviours from different functional copies of those proteins. This motivates expansion of the TPP method to deconvolute these behaviours, for example phosphoTPP (Huang et al ., 2019; Potel et al ., 2020; Smith et al ., 2020) and other PTMs. The RNA-binding behaviour could be examined with high-throughput RNA-protein enrichment methods (Queiroz et al ., 2019) and further deconvolution could be obtained by combining TPP with spatial proteomics methods (Mulvey et al ., 2017; Geladaki et al ., 2019).…”

“…Moreover it can be applied to study interactions with metabolites, nucleotides and nucleic acids (Saei et al ., 2018; Dziekan et al ., 2019; Sridharan et al ., 2019; Becher et al ., 2018). Authors have shown that proteins in complex with each other are more likely to have concordant in vivo melting curves (Tan et al ., 2018) and others have demonstrated that phosphorylation can alter thermal stability (Huang et al ., 2019; Potel et al ., 2020; Smith et al ., 2020). Thermal proteome profiling has also been complemented with extensive structural analysis (Feng et al ., 2014; Leuenberger et al ., 2017; Schopper et al ., 2017; Piazza et al ., 2018).…”

Systematic Identification of Protein Phosphorylation-Mediated Interactions

“…Experimental approaches for high-throughput identification of protein phosphosite functions have been largely lacking. More recently, a promising approach has been to use a modified thermal proteome profiling (TPP) approach where the TPP protocol is conducted in conjunction with phosphoproteomics to identify functional phosphosites [18][19][20][21] . This method has proven useful to directly determine whether a phosphosite is likely to be functional or not.…”