Proteomics pipeline for phosphoenrichment and its application on a human

Francesco Finamore a,1, Nadia Ucciferri a,1, Giovanni Signore b,c, Antonella Cecchettini a,d, Elisa Ceccherini a, Marianna Vitiello a,e, Laura Poliseno a,e, Silvia Rocchiccioli a,*

A B S T R A C T

Cell signalling is tightly regulated by post-translational modification of proteins. Among them, phosphorylation is one of the most interesting and important. Identifying phosphorylation sites on proteins is challenging and requires strategies for pre-separation and enrichment of the phosphorylated species. We applied four different methods for phospho-enrichment involving TiO2 and IMAC matrix to human melanoma cell lysates of starved A375 induced for 1 h with 1% FBS. Comparison of protocol efficiency was evaluated through peptide concentration, sulphur and phosphorus content and peptide analysis by LC-MS in the collected fractions.
Our results underlined that each single method is not sufficient for a comprehensive phosphoproteome analysis. In fact, each methodology permits to identify only a fraction of the phosphoproteome contained in a whole cell lysate. The selection of the most efficient protocols and a combination of two phospho-enrichment methods allowed the assessment of this workflow able to pinpoint the main actors in the phospho-proteome cascade of A375 human melanoma cells treated with Vemurafenib.

Keywords:
Phosphoproteomics
ICP-MS
LC-MS/MS
BRAF inhibitor

1. Introduction

Proteins play a crucial role in many biological processes and their phosphorylation is a key reversible post-translational modification (PTM) that regulates protein function, subcellular localization, complex formation, degradation of proteins and therefore cell signalling networks. Indeed, phosphorylation represents an important tightly controlled PTM highly conserved from prokaryotes to humans. Accordingly, analysis of phosphoproteins provides clues on which proteins and pathways might be activated and involved in particular conditions or diseases [1].
It is well-known that 30% of all proteins in cells are phosphorylated at some point during their expression [2,3]. Thus, the development of qualitative and quantitative analytical strategies able to identify and quantify a broad range of phosphorylated peptides and proteins would be an invaluable tool to unveil subtle details of the biological processes. From the huge impact of phosphoproteomics in deep understanding of molecular mechanism, bursts the searching for a reproducible, specific, possibly uniform, methodology for a comprehensive phospho-proteome analysis of biological samples.
Unfortunately, phosphoprotein analysis is challenging, as phosphorylation is a transient labile modification. As a consequence, phosphorylated proteins are often present at significantly lower concentration compared with their non-phosphorylated counterparts [4], while their biological relevance is often independent on their concentration. A further complication in the development of mass spectrometry-based analytical methods to assay phosphoproteins arises from their poor ionization efficiency and ion suppression effect. Sample complexity, low relative abundance respect to naïve proteins, the low ionization efficiency of phosphorylated peptide sequences hampers the detection of phosphorylated proteins by mass spectrometry.
Phospho-enrichment methods were developed to overcome these issues; these approaches try to remove non-phosphorylated proteins or peptides, therefore easing detection of the less abundant phosphorylated species. While this field is widely investigated in the literature [5–7], a standard and well accepted method has not been yet established, as each phosphoprotein/peptide isolation method has its own bias.
Many strategies are described in literature [8,9] with contrasting results, thus leading to a general lack of consensus on the most suitable methodology, since efficiency of the reported approaches could vary significantly also depending on the biological system examined.
A systematic comparison of the proposed approaches in a cell model of melanoma would provide a useful tool to select the most appropriate enrichment protocol.
We focused on two widely employed protocols based on metal affinity, for phospho-peptide enrichment: IMAC (Immobilized metal affinity chromatography) [10] is composed by a matrix of resins with associated metal ions, in our case Fe(III) chelate; MOAC (Metal oxide affinity chromatography) [11,12] is characterized by a matrix composed of TiO2, with no need for resin anchoring as for IMAC. These two different methods lead to different outcomes in phospho-enrichment [13]. While for IMAC a quite standardized approach referring to buffers is available, the information presented in the literature for TiO2 methodology remains contradictory, especially regarding the loading buffer and the need for a quenching agent to limit non-specific binding [14]. Hydroxy acids or aliphatic hydroxycarboxylic acids should compete for TiO2 binding with increased affinity compared with carboxylic groups of acidic non-phosphopeptides, without interfering with the phosphate group, thanks to the higher binding energy of the latter. 2, 5-Dihydroxybenzoic acid (DHB) is a good quenching agent but unfortunately column clogging, precipitation around the orifice, and significant suppression of phosphopeptide ionization prevent its applicability for LC-ESI-MS [15]; lactic acid is the best alternative [15,16], and also glycolic acid has shown promising results [14], but in some cases optimized loading buffers without any hydroxy acid seem to be sufficient [14,17].
The effectiveness of the different strategies was valuated comparing quantitative (phosphopeptide enrichment and protein concentration) and qualitative (protein identification overlaps between methods) parameters in starved/induced A375 cells.
Finally, the most promising enrichment combination was applied to compare the biological processes associated to the phosphoprotein content derived from melanoma A375 cell lines in presence and absence of vemurafenib, a well-known BRAF inhibitor (BRAFi) [18].

2. Experimental section

2.1. Reagents

Reagents were purchased from Sigma Aldrich (St. Louis, USA) unless otherwise specified.

2.2. Cell treatment

To set up the protocols, A375 human melanoma cells were seeded at 106 density in a 10 cm plates and grown in DMEM High Glucose 10% FBS overnight. Then, cells were washed 3 times with PBS and serum starvation was performed adding DMEM High Glucose 1% FBS for 24 h. Finally, stimulation was performed with DMEM High Glucose 10% FBS for 1 h, then cells were harvested and centrifuged.
For the assessment of the method and to evaluate phospho-enriched pathways, A375 cells were treated with DMSO (control sample) or with 2 μM vemurafenib for 48 h, then cells were harvested and centrifuged. All pellets were washed twice with PBS to remove FBS traces.
Cells were lysed in a buffer containing 25 mM Tris-HCl/10% ACN (Romil, UK), ultrasonicated 5 times 20 s and 40 s rest and denatured heating at 99 C for 5 min (SoniPrep 150, MSE). After denaturation samples were centrifuged at 12000g (Biofuge Fresco, Heraeus Instruments) and the surnatants quantified using BCA assay (Pierce, Thermo Scientific, Waltham, USA). Denaturation was chosen instead of addition of protease inhibitors to avoid decrease of enrichment specificity [14].

2.3. Sample pre-processing

500 μg of proteins were used for each phospho-enrichment protocol. Reduction was performed using dithiotreitol 5 mM at 80 C for 20 min and alkylation with iodoacetamide 10 mM at 37 C for 30 min. Proteins were digested incubating overnight with 1:100 trypsin (Roche, Germany) substrate at 37 C. Solutions were acidified with trifluoroacetic acid (TFA) and centrifuged at 12000 g for 10 min. Supernatant was then loaded on a C18 cartridge to purify peptide solution, and evaporated in a SpeedVac. An aliquot of peptide solution was directly analysed in LC- MS.

2.4. Phospho-peptides enrichment

Protocol involving titanium used TopTip (Glygen Corp., Columbia, USA) while IMAC resin was the PHOS-Select™ gel (Sigma Aldrich, St Louis, USA) and 80 μL of slurry were loaded in a mobicol column (MobiTec, Gottingen, Germany). Protein extract was split in batches of € 500 μg of peptide solution, which were diluted in the correspondent loading buffer.
Each protocol contemplates: 1) resin equilibrating step, 2) loading phase with longitudinal vortexing for 2 h, 3) eluate reloading for 30 min vortexing, 4) first wash, 5) second wash, 6) First elution step vortexing 30 min, 7) Second elution step. After passage 3 the aspecific fraction was obtained (ASP); after passage 4 and 5, Wash 1 (W1) and Wash 2 (W2) solution, respectively: the collection of eluates from passage 6 and 7 represented the phospho-peptide enriched specific fraction (SPE). Buffer volume was 200 μL, and centrifuge speed was at 4000g for 3 min for TopTip, while 10000g for 1 min for mobicol (Table 1). Each eluted solution was acidified and evaporated in the SpeedVac. Each chromatographic run was repeated twice and eluted fractions were respectively pooled before further analysis.
Peptides were purified in the C18 cartridge and evaporated again. Each fraction was diluted to 100 μL with 2% ACN/1% formic acid and measured twice at 280 nm with Nanodrop (Thermo Scientific, Waltham, USA). Finally, 50 μL of each solution were diluted to 1 mL of Ultrapure Water used for ICP-MS analysis.
Only Glycolic acid and IMAC protocols reported above were used for assessment in A375 treated with DMSO and Vemurafenib. Only the elution fraction was retained and subsequently analysed by nanoLC-MS/ MS as reported below.

2.5. ICP-MS analysis in starved/induced A375

Quantification of phosphorous and sulphur content in the samples was performed on an Agilent Technologies 7700 Series ICP-MS. Trace- level nitric acid (100 μL) was added to each sample (1 mL), and the resulting solution was directly analysed in the ICP-MS instrument. Phosphorous and sulphur content in the sample was blank-subtracted using a 3% solution of trace-level nitric acid in trace-level water as reference. All measurements were performed in duplicate.

2.6. nanoLC-MS/MS analysis

5 μL (or 2 μL for high concentrated solutions) of each solution were injected twice for LC-MS analysis. Peptides were chromatographically separated using a nano-HPLC system (Eksigent, ABSciex, USA). Samples were pre-concentrated in a pre-column cartridge (PepMap-100 C18 5 μm 100 A, 0.1 20 mm, Thermo Scientific, Waltham, USA) and then separated in a C18 PepMap-100 column (3 μm, 75 μm 250 mm, Thermo Scientific, Waltham, USA) at a flow rate of 300 nL/min. Runs were performed with eluent A (Ultrapure water, 0.1% formic acid) under 60 min linear gradient from 5 to 40% of eluent B (ACN, 0.1% formic acid) followed by 10 min of a purge step and 20 min re- equilibration step.

2.7. Phospho-protein identification

LC-MS/MS wiff data were uploaded into ProteinPilot™ Software (ABSciex, USA) to perform protein identification. Peak lists were queried against human protein UniProtKB database (relase May 2016). An error tolerance of 0.05 and 0.1 Da were selected for precursor and fragment ions, respectively. Cysteine carbamidomethylation was selected as fixed modification while methionine oxidation and serine, threonine and tyrosine phosphorylation were used as variable modifications. Only precursor ions with a charge state of 2, 3 and 4 were considered for peptide identification. A minimum number of 2 peptides were chosen to consider a protein as identified. In order to reduce the number of false positive hits, a false discovery rate (FDR) lower than 5% was used.

2.8. Gene ontology (GO) and protein interaction network analysis of A375 phospho-proteome

Gene ontology (GO) analysis was performed using the Cytoscape plug-in ClueGO. GO parameters used for functional grouping of phospho-proteins identified from A375 with and without vemurafenib were as follows: a p-value 0.01 integrated with a Bonferroni step down correction, a GO tree interval between 3 and 8, a minimum number of genes per cluster of 3 with a 4% of genes, a kappa score of 0.4 and an initial group size of 3 terms with a percentage of overlapping terms per group of 50%. GO term fusion of similar associated genes was enabled. Significant GO biological process terms were displayed as a heat map indicating in colour-code the percentage of associated proteins per GO term.

3. Results

3.1. LC-MS/MS and ICP-MS approaches for phospho-protein profile of A375 cells

Four methods of phospho-peptide enrichment were compared using A375 human melanoma cells that were first serum-starved in medium containing 1% FBS and then stimulated for 1h in complete medium (10%FBS). The best combination was then applied to A375 cells treated for 48 h with 2 μM Vemurafenib [18], a common BRAFi, or DMSO (vehicle). The experimental workflow is thoroughly described in Fig. 1. We used a standardized procedure for IMAC. TiO2 resin was employed with three protocols differing in the composition of the acidic mixture used in order to wash and remove the aspecifically bound content, i.e. (i) without hydroxyl acid (ii) with 80 mg/mL glycolic acid (hereafter TiO2-glycolic) and (iii) with 30% lactic acid (hereafter TiO2-lactic). Protein extracts were split in two batches that were pooled after chromatographic separation to avoid replicate variability and column overload during the enrichment phase. In every fraction peptide concentration was evaluated by absorbance at 280 nm, by sulphur and phosphorus contents and by peptide and phospho-peptide ID.
The count of phospho-peptides and total peptides retrieved in each fraction (Table 2A) was used to assess the efficacy and specificity that each method has for phospho-groups. The proteins and phospho- proteins recovered in each fraction were identified using SwissProt 2018 database for human as taxon (Table 2B).
Theoretically, enrichment is effective when all phospho-peptides bind to the resin and elute only in the specific fraction. According to this consideration, enrichment by IMAC provided the best results in term of phospho-protein identification while TiO2-glycolic resulted the method with the lowest efficiency in phosphopeptide enrichment.
To overcome the intrinsic limitations of phospho-peptides detection by mass spectrometry, the relative concentration of total phospho- peptides in each fraction was determined by two established approaches: absorbance at 280 nm (Fig. 2A) and relative concentration of sulphur content in the samples obtained by ICP-MS (Fig. 2B), a commonly accepted parameter that well correlates with protein concentration. Results from these two techniques show that ICP-MS slightly overestimated sulphur content in Wash 2 and SPE fraction, even if the relative trend was maintained. Quantification of phosphorous content by ICP-MS evidenced a significant enrichment in phosphorylated species in SPE fraction for all methods. Different enrichment profiles were shown and IMAC and TiO2-glycolic provided the best performance compared to TiO2-lactic and without hydroxy acid solutions.
A qualitative analysis was also performed to evidence the overlap and complementarity between the four methods. Total identified phospho-proteins across methods are listed in Appendix A, Table S1. Common phospho-proteins were 18% (83 out of 471), while unique phospho-proteins were 8% in protocol without hydroxy acid solutions; 13% for TiO2-glycolic; 9% for the TiO2-lactic and 21% for IMAC (Fig. 3). The combination of IMAC and TiO2-glycolic presents the best protein coverage. The IMAC alone identifies 316 phospho-proteins, TiO2-glycolic alone identifies 178 phospho-proteins (Table 2B). Of these 110 are common between the two methods (Table S1). The total number of unique phospho-proteins identified by IMAC þ TiO2-glycolic are 384 of 471 total ID (representing the 81% of protein coverage).

3.2. Phospho proteome enrichment in A375 cells treated with vemurafenib and GO terms analysis

In order to determine the functional impact of BRAFV600E on human melanoma protein profile, we compared the phospho-protein content of human A375 cells in presence or absence of the selective BRAFV600E inhibitor vemurafenib. Considering that the combination of IMAC and TiO2-glycolic led to the best protein coverage (81% protein coverage) among the considered paired methods, we used this combination to characterize the phosphoprotein profile associated to melanoma cells treated or not with the drug.
By analysing with nanoLC-MS/MS the eluted fractions, we obtained the identification of 60 unique phosphoproteins in untreated A375 cells and 159 phosphoproteins in A375 cells treated with vemurafenib (Appendix B, Table S2).
Results from GO analysis showed that vemurafenib does not have a significant impact on biological process terms related to epigenetic processes like histone and DNA methylation (DNMT1), gene silencing (HIST1H1C, PUM1, TP53BP1), regulation of gene expression (ATRX, CTNNB1, NELFA) and chromatin organization (ATRX, HIST1H1B–C-D- E).
On the other hand, GO terms associated to mRNA processing like splicing (PRPF38A, RBM17, RBM25, SON, SRSF), transcription (GTF2F1, NELFA, NELFB) and transport (AHCTF1, BICD2, KTN1) were found to be particularly enriched in presence of vemurafenib compared to control cells (Fig. 4). Most of the proteins associated to the mRNA processing term were identified only in cells treated with vemurafenib (black nodes), while only 4 proteins were detected in control cells (DMSO treatment) (dark grey nodes) and only one protein was found to be in common between the two test conditions (light grey node) (Fig. 5).

4. Discussion

We compared four enrichment protocols applied to human melanoma cell lysates and analysed the efficiency and phospho-specificity for each method as well as the quantitative and qualitative differences among the protocols.
Evaluation of each method in phospho-enrichment capacity was performed by mass spectrometry considering three parameters: 1) the number of unique phospho-proteins identified in each fraction for each method; 2) the number of phospho-peptides identified in each fraction for each method and 3) the percentage of phospho-peptide ID/total peptide ID in each fraction for each method. Moreover, semi- quantitative analysis obtained by absorbance at 280 nm allowed to compare protein content while phosphorous concentration through ICP- MS confirmed the protocol specificity for phospho-binding.
Total protein ID showed a decreasing trend from the aspecific to the wash fractions in all methods. This suggests that wash steps successfully remove aspecific binding allowing good enrichment in the specific fraction (Table 2B). The specificity of phospho-peptide binding is better highlighted by the analysis of phospho-peptides/total ID peptides ratio (Table 2A).
Leakage of phosphorylated peptides in the aspecific fraction in the IMAC/TiO2 resin can be due to the competition with a large concentration of carboxylic groups present on carboxylated peptides. Those residues have slight affinity and can occupy binding sites thus reducing availability for phosphate groups. Actually, leaked phospho-peptides in cells (i.e. lost during the two wash steps, defined on the basis of percentage of identified phospho-peptides as (W1 þ W2)/Specific) accounted for 6%, 18%, 2.3% and 2.6% of the specifically bound peptides when none hydroxy acid solutions, TiO2-glycolic, TiO2-lactic, and IMAC methods were used, respectively. It is worth noting that TiO2- glycolic method always led to suboptimal results, with substantial amounts of phospho-peptides lost in the washing steps. A closer look evidences that phospho-peptide leakage observed with TiO2-glycolic method actually increases during the second wash. It is tempting to deduce that a competitive effect of glycolic acid on the resin might be present in our experimental conditions.
Our results clearly indicate that there is a correspondence between number of detected phospho-peptides and number of phospho-proteins thus indicating a good distribution of peptide to protein relative abundance. The data evidence that IMAC is consistently the more effective protocol, leading to the most abundant identification of phospho- proteins in A375 cell model. However, TiO2-lactic protocol is generally in line with IMAC, with 316 and 236 for cell lysate (SPE in Tab 2B) using IMAC and TiO2-lactic, respectively.
To highlight the importance of the enrichment step in phospho- proteome analysis, we want to stress that all the identified phospho- peptides were undetectable before enrichment (data not shown). Concerning the classes of identified proteins, it is worth noting that Zinc Finger proteins were only identified when IMAC approach was employed. These are histidine rich proteins able to bind Zinc. Actually, histidine-rich peptides are considered competitors in IMAC Fe-binding; this can explain their abundance in IMAC Specific fraction. Our findings accounting for 700 phospho-peptide ID are in line with other high throughput methodologies, where 1000 phospho-peptides were found using IMAC by loading methyl ester peptides derivatives from 500 μg of lysate in a single analysis [19]. Note that the derivatization step reported in this work requires a more complex methodology allowing a complete phospho mapping. We also performed an alternative quantification based on ICP-MS analysis of phosphorous and sulphur to evaluate phospho-enrichment. Since its high sensitivity (down to low ppt) and its versatility for metal and non-metal compounds, we use ICP-MS as validation approach to confirm the results obtained by nanoLC-MS. This technique has been already used to this aim [20,21] and provides semi-quantitative results that ease a direct comparison of protein and phospho content in the different methods.
First, we evaluated relative sulphur abundance (a parameter correlated with cysteine and hence protein content) to estimate peptide concentrations comparing it to the absorbance reading at 280 nm. Unfortunately, the correlation between sulphur content and UV absorbance measurements did not show a significant trend. This fact could be tentatively ascribed to the presence of sulphur-containing species arising from mammalian cell lysis or to a greater uncertainty in the measure due to the values close to the detection limit of the instrument, although further studies are necessary to clarify this behaviour.
Phosphorous content was higher (up to 20 folds) in SPE fraction compared with aspecific, W1, and W2, where was low or almost undetectable for all the methods (Fig. 2B). Enrichment efficiency is emphasized by the opposite trends of sulphur (dashed lines) and phosphorous concentration that clearly indicate the presence of relevant amounts of phosphorylated proteins in SPE fraction [22]. The observed discrepancy between phospho-peptide abundance or total phospho-peptide ID evaluated by mass spectrometry (Table 2A) and results obtained by ICP-MS in case of TiO2-glycolic and TiO2-lactic can be reconciled in view of the different approaches. While ICP-MS analysis detects any source of sulphur or phosphorus, a significant filter is operated by LC-MS, neglecting peptides that are not correctly identified by processing or post-processing parameters or that contain post-translational modifications.
The overlap of identified phospho-protein among the four methods is summarized in Fig. 3. Interestingly, only 18% of all phospho-proteins identified could be detected by all methods. Even the two based enrichment protocols TiO2-glycolic and TiO2-lactic -that rely on chemically similar eluting agents-resulted in only around 40–50% of common phospho-proteins. Each method identified a quite significant portion of unique phospho-proteins, meaning that each method adds some degree of information to the whole proteome.
Method complementarity was already observed in phospho- enrichment analysis. Bodenmiller and colleagues [23] described a significant complementarity in 3 methods based on different binding material, detecting partially overlapping segments of the phospho-proteome.
Conversely, Ruprecht and colleagues [24] attributed complementarity of different materials for phospho-peptide enrichment to binding capacity, biased or incomplete elution, shortcomings in the physical formats (batch, tip), and limited analytical capacity of the mass spectrometer. They also demonstrated that the overlap of phospho-peptide identifications increased substantially when the analytical depth increased, applying a double serial enrichment to a peptide mix and analysing it in a 48 h run. This protocol is anyway too complex and time consuming to be useful in the characterization of phospho-proteome for high-throughput studies.
In order to define the functional profile of human phospho-proteins analysed in this study, we used the data distribution of the identified phosphorylated proteins enriched by IMAC and TiO2-glycolic methods, that resulted in the maximum yield of protein detection.
The comparison between phospho-proteins derived from human melanoma A375 cells treated with and without vemurafenib allowed us to enrich a meaningful fraction of proteins involved in transcription process and RNA metabolic processes, including processing, splicing and transport. A large burden of studies focused their attention on the characterization of phospho-protein content of BRAF mutated cells and in its dynamics upon the use of BRAF kinase activity inhibitors [25]. Parker et al. demonstrated the effectiveness in shutting down MAPK signalling through the inhibition of BRAF by vemurafenib in thyroid cancer cells [26] and in melanoma cells [27]. Despite the latter cellular models are different from that one used in this study, regulation of transcriptional activity and RNA processing were found to be the most representative biological processes influenced by the effect of vemurafenib. Interestingly, it is worth to mention that one of the mechanism of resistance through which BRAF escape from the effect of its inhibitors is the generation of splicing variants of BRAFV600E that are shorter compared to their counterpart [28]. Moreover, it was shown that melanoma cells that are resistant to vemurafenib recover their sensitivity to it through the inhibition of pre-mRNA splicing [29]. Taken together our results suggest that splicing modulation is an early event that occur when cells are still sensitive to the effect of vemurafenib, and thus before the onset of acquired resistance.

5. Conclusions

Since biological mechanisms are driven by activation of molecules mainly through phosphorylation, methods to enrich phosphorylated proteins are fundamental in proteomics study related to pathologies.
We performed a comparative analysis on the efficiency of four phospho-peptide enrichment protocols and compared the identified phospho-proteins from each method in terms of phospho-protein/ peptide ID number by mass spectrometry and protein concentration estimated by absorbance and S and P concentrations. We found that each method allowed the enrichment of different portions of the phospho- proteome and none alone is sufficient for a complete and comprehensive phospho-proteome analysis. However, the combination of IMAC and TiO2-glycolic led to the best protein coverage among the considered paired methods with 81% protein coverage.
These two enrichment approaches applied to human melanoma cell model treated with Vemurafenib allowed us to highlight preferential biological processes mostly related to an active transcriptional activity induced by the constitutional active BRAFV600E. Moreover, the redundant function of BRAFV600E in melanoma cells pave the way towards the identification of panels of phospho-protein markers in order to evaluate the efficiency of new potential inhibitory drugs, beside Vemurafenib. Due to the difficulty to map the whole phospho-proteome, it is important to find the right protocol -or protocol combination- leading to the portion of phospho-proteins that best represents the model under investigation.

References

[1] F. Ardito, M. Giuliani, D. Perrone, G. Troiano, L. Lo Muzio, The crucial role of protein phosphorylation in cell signalingand its use as targeted therapy, Int. J. Mol. Medicine 40 (2017) 271–280, https://doi.org/10.3892/ijmm.2017.3036.
[2] D.T. McLachlin, B.T. Chait, Analysis of phosphorylated PLX4032 proteins and peptides by mass spectrometry, Curr. Opin. Chem. Biol. 5 (2001) 591–602, https://doi.org/ 10.1016/s1367-5931(00)00250-7.
[3] M.K. Tarrant, P.A. Cole, The chemical biology of protein phosphorylation, Annu. Rev. Biochem. 78 (2009) 797–825. https://doi:10.1146/annurev.biochem.78.0709 07.103047.
[4] K. Blackburn, M.B. Goshe, Challenges and strategies for targeted phosphorylation site identification and quantification using mass spectrometry analysis, Briefings Funct. Genomics Proteomics 8 (2008) 90–103, https://doi.org/10.1093/bfgp/ eln051.
[5] S. Lemeer, A.J. Heck, The phosphoproteomics data explosion, Curr. Opin. Chem. Biol. 13 (2009) 414–420, https://doi.org/10.1016/j.cbpa.2009.06.022.
[6] P. Picotti, Phosphoproteomics takes it easy, Nat. Biotechnol. 33 (2015) 929–930, https://doi.org/10.1038/nbt0416-441b.
[7] H. Zhou, J.D. Watts, R. Aebersold, A systematic approach to the analysis of protein phosphorylation, Nat. Biotechnol. 19 (2001) 375–378, https://doi.org/10.1038/ 86777.
[8] J. Fílla, D. Honys, Enrichment techniques employed in phosphoproteomics, Amino Acids 43 (2012) 1025–1047, https://doi.org/10.1007/s00726-011-1111-z.
[9] C. Yang, X. Zhong, L. Li, Recent advances in enrichment and separation strategies for mass spectrometry-based phosphoproteomics, Electrophoresis 35 (2014) 3418–3429, https://doi.org/10.1002/elps.201400017.
[10] H. Block, B. Maertens, A. Spriestersbach, N. Brinker, J. Kubicek, R. Fabis, J. Labahn, F. Schafer, Immobilized-metal affinity chromatography (IMAC): a review, Methods Enzymol. 463 (2009) 439–473, https://doi.org/10.1016/S0076- 6879(09)63027-5.
[11] M.R. Larsen, et al., Highly selective enrichment of phosphorylated peptides from peptide mixtures using Titanium Dioxide microcolumns, Mol. Cell. Proteomics 4 (2005) 873–886, https://doi.org/10.1074/mcp.T500007-MCP200.
[12] T.E. Thingholm, et al., Highly selective enrichment of phosphorylated peptides using titanium dioxide, Nat. Protoc. 1 (2006), https://doi.org/10.1038/ nprot.2006.185, 1929–35.
[13] X. Yue, A. Schunter, A.B. Hummon, Comparing multistep immobilized metal affinity chromatography and multistep TiO2 methods for phosphopeptide enrichment, Anal. Chem. 87 (2015) 8837–8844, https://doi.org/10.1021/acs. analchem.5b01833.
[14] U.K. Aryal, A.R.S. Ross, Enrichment and analysis of phosphopeptides under different experimental conditions using titanium dioxide affinity chromatography and mass spectrometry, Rapid Commun. Mass Spectrom. 24 (2010) 219–231, https://doi.org/10.1002/rcm.4377.
[15] N. Sugiyama, et al., Phosphopeptide enrichment by aliphatic hydroxy acid- modified metal oxide chromatography for nano-LC-MS/MS in proteomics applications, Mol. Cell. Proteomics 6 (2007) 1103–1109, https://doi.org/10.1074/ mcp.T600060-MCP200.
[16] J. Wu, et al., Global profiling of phosphopeptides by titania affinity enrichment, J. Proteome Res. 6 (2007) 4684–4689, https://doi.org/10.1021/pr070481m.
[17] Y.H. Ahn, et al., Coupling of TiO(2)-mediated enrichment and on-bead guanidinoethanethiol labeling for effective phosphopeptide analysis by matrix- assisted laser desorption/ionization mass spectrometry, Rapid Commun. Mass Spectrom. 21 (2007) 3987–3994, https://doi.org/10.1002/rcm.3278.
[18] M. Vitiello, et al., Context-dependent miR-204 and miR-211 affect the biological properties of amelanotic and melanotic melanoma cells, Oncotarget 8 (2017) 25395–25417, https://doi.org/10.18632/oncotarget.15915.
[19] S.B. Ficarro, et al., Phosphoproteome analysis by mass spectrometry and its application to Saccharomyces cerevisiae, Nat. Biotechnol. 20 (2002) 301–305, https://doi.org/10.1038/nbt0302-301.
[20] X. Yan, L. Yang, Q. Wang, Detection and quantification of proteins and cells by use of elemental mass spectrometry: progress and challenges, Anal. Bioanal. Chem. 405 (2013) 5663–5670, https://doi.org/10.1007/s00216-013-6886-1.
[21] E. Maes, et al., The use of elemental mass spectrometry in phosphoproteomic applications, Mass Spectrom. Rev. 35 (2016) 350–360, https://doi.org/10.1002/ mas.21440.
[22] R. Krüger, et al., Plant protein phosphorylation monitored by capillary liquid chromatography–element mass spectrometry, Biochem. Biophys. Res. Commun. 355 (2007) 89–96, https://doi.org/10.1016/j.bbrc.2007.01.108.
[23] B. Bodenmiller, et al., Reproducible isolation of distinct, overlapping segments of the phosphoproteome, Nat. Methods 4 (2007) 231–237, https://doi.org/10.1038/ nmeth1005.
[24] B. Ruprecht, et al., Comprehensive and reproducible phosphopeptide enrichment using iron immobilized metal ion affinity chromatography (Fe-IMAC) columns, Mol. Cell. Proteomics 14 (2015) 205–215, https://doi.org/10.1074/mcp. M114.043109.
[25] S.A. Stuart, et al., A Phosphoproteomic comparison of BRAFV600E and MKK1/2 inhibitors in melanoma cells, Mol. Cell. Proteomics 14 (2015) 1599–1615, https:// doi.org/10.1074/mcp.M114.047233.
[26] R. Parker, R. Clifton-Bligh, M.P. Molloy, Phosphoproteomics of MAPK inhibition in BRAF-mutated cells and a role for the lethal synergism of dual BRAF and CK2 inhibition, Mol. Canc. Therapeut. 13 (2014) 1894–1906, https://doi.org/10.1158/ 1535-7163.MCT-13-0938.
[27] R. Parker, et al., Phosphoproteomic analysis of cell-based resistance to BRAF inhibitor therapy in melanoma, Front. Oncol. 5 (2015) 95, https://doi.org/ 10.3389/fonc.2015.00095.
[28] P.I. Poulikakos, et al., RAF inhibitor resistance is mediated by dimerization of aberrantly spliced BRAF(V600E), Nature 480 (2011) 387–390, https://doi.org/ 10.1038/nature10662.
[29] M. Salton, et al., Inhibition of vemurafenib-resistant melanoma by interference with pre-mRNA splicing, Nat. Commun. Nature Publishing Group 6 (2015) 7103, https://doi.org/10.1038/ncomms8103.