High-throughput phosphoproteomics reveals in vivo insulin signaling dynamics

Journal name:
Nature Biotechnology
Volume:
33,
Pages:
990–995
Year published:
DOI:
doi:10.1038/nbt.3327
Received
Accepted
Published online

Mass spectrometry has enabled the study of cellular signaling on a systems-wide scale, through the quantification of post-translational modifications, such as protein phosphorylation1. Here we describe EasyPhos, a scalable phosphoproteomics platform that now allows rapid quantification of hundreds of phosphoproteomes in diverse cells and tissues at a depth of >10,000 sites. We apply this technology to generate time-resolved maps of insulin signaling in the mouse liver. Our results reveal that insulin affects ~10% of the liver phosphoproteome and that many known functional phosphorylation sites, and an even larger number of unknown sites, are modified at very early time points (<15 s after insulin delivery). Our kinetic data suggest that the flow of signaling information from the cell surface to the nucleus can occur on very rapid timescales of less than 1 min in vivo. EasyPhos facilitates high-throughput phosphoproteomics studies, which should improve our understanding of dynamic cell signaling networks and how they are regulated and dysregulated in disease.

At a glance

Figures

  1. Scalable, EasyPhos phosphoproteomics platform for single-run analysis of phosphoproteomes compared with conventional workflows.
    Figure 1: Scalable, EasyPhos phosphoproteomics platform for single-run analysis of phosphoproteomes compared with conventional workflows.

    (a) Conventional phosphoproteomics workflows generally require around 10 mg of sample lysate, use FASP (filter-assisted sample preparation) or urea-based protein digestion, followed by peptide desalting and lyophilization, peptide fractionation using strong cation exchange or high-pH reversed-phase chromatography. Phosphopeptides are enriched by IMAC or TiO2, sometimes multiple times, resulting in numerous LC-MS/MS measurements per biological sample to be analyzed. (b) The phosphoproteomics workflow described here requires minimal starting materials, no fractionation, and uses TFE-based digestion, eliminating the need for peptide desalting before streamlined phosphopeptide enrichment in tubes or 96-well format. Phosphopeptides are subsequently analyzed by single-run LC-MS/MS measurements. (c,d) Rapid, parallel enrichment in 96-well plate format (c) and StageTip-based elution (d).

  2. Experimental design for time-resolved map of in vivo insulin signaling in the liver.
    Figure 2: Experimental design for time-resolved map of in vivo insulin signaling in the liver.

    (a) Experimental design for phosphopeptides LC-MS analysis of control (PBS-treated) and insulin-treated mice. The mouse hepatic portal vein was perfused with either PBS or insulin for different lengths of time to yield four 'early' and seven 'intermediate' time points. (b) Unsupervised clustering of time-resolved phosphorylation profiles reveal functionally linked nodes in insulin signaling. Each temporal profile is color coded according to its distance from the respective cluster center. Shown on the right are proximal nodes in the canonical insulin signaling pathway; tyrosine phosphorylation of the insulin receptor and the insulin receptor substrate (Irs1), the activation loop of Akt, and Akt substrates. Shown in bold are several insulin-regulated substrates not previously implicated in insulin signaling, representing candidate PI3K-Akt targets in the liver.

  3. Features of insulin signaling in the liver.
    Figure 3: Features of insulin signaling in the liver.

    (a) Multisite insulin-stimulated phosphorylation of Akt2 at S474 by mTORC2 and S478 by an unknown kinase, in the C-terminal hydrophobic motif (A and B, left), produces distinct temporal profiles (A and B, right). (b) Phosphorylation of residues on adaptor protein Frs2 in response to insulin indicate communication between the insulin and Fgfr pathways, occurring through unknown intermediates (multi-arrow lines) (A and B, left). Temporal profiles (A and B) are shown on right. (c) Temporal relationship (A, B and C, right) between the insulin receptor (Insr) (A, left) and insulin-regulated phosphorylation of the scaffolding protein Gab1 in the liver (B and C, left). All data plotted are the median fold-changes compared with PBS control mice from six to ten biological replicates. All data points plotted are derived from at least two values, and where shown, error bars denote s.e.m. from at least three values.

  4. Time-course of insulin-induced PI3K-Akt and MAPK network activity in vivo.
    Figure 4: Time-course of insulin-induced PI3K-Akt and MAPK network activity in vivo.

    (a) Insulin signaling pathway members activated in liver are rapidly phosphorylated upon insulin treatment. The pathway figure represents summarized literature-curated knowledge of proximal regions of the canonical PI3K-Akt and MAPK signaling networks. (b) Temporal profiles for phosphorylation changes of key signaling nodes are depicted in the top panel, after insulin treatment. Data are median fold-changes compared with PBS control mice from six to ten biological replicates. All data points plotted are derived from at least two values, and where shown, error bars denote s.e.m. from at least three values.

  5. Precision of the phosphoproteomics workflow.
    Supplementary Fig. 1: Precision of the phosphoproteomics workflow.

    Two identical SILAC-labelled HeLa cell populations were mixed a immediately after cell scraping or b after the phosphorylation enrichment workflow and immediately before LC-MS analysis. c-d Distribution and standard deviations of SILAC ratios of phosphopeptides from the experiments shown in the panel above are shown. e Sources of variability in phosphoproteome platform (1) biological, (2) workflow, or (3) LC-MS, investigated by replicates at each level as indicated. f Box-plot of Coefficient of Variation (%) of phosphopeptide intensities calculated from replicate sample data in e. g Average Pearson correlation coefficients (in box inset) for the replicate experiments, and density plots depicting the log2 transformed intensities of two representative replicates. Representative samples shown in plots were chosen based on the closest match between pairwise correlation coefficients and average correlation coefficients for all replicates. h Performance of match between runs (MBR) in transferring identification of phosphopeptides and quantification between replicate workflow samples. Percent valid values are calculated by dividing the number of unique phosphopeptides quantified in all replicates by the total number of possible quantification points, i.e. (∑ phosphopeptides quantified / (total number of phosphopeptides identified x number of replicates)) x 100). Error bars denote SD.

  6. Phosphoproteome analysis of mouse cell lines and tissues.
    Supplementary Fig. 2: Phosphoproteome analysis of mouse cell lines and tissues.

    a Experimental design for phosphoproteome analysis of control (PBS) and insulin-treated mouse liver cell lines (100 nM insulin, 15 minutes) and unstimulated (fed animals) liver, kidney and brain mouse tissues with six biological replicates each. Samples were enriched using the described phosphoproteomics pipeline, and the analysis was performed with single-run measurements lasting 0.5-1.5 day for all replicates of the respective cell or tissue typeper cell or tissue type. b Summary of the identified phosphorylation sites, including enrichment specificity (number of unique phosphopeptides identified divided by the total number of unique peptides identified). c Rate of phosphopeptide identification over the LC gradient in a single sample (Hepa 1-6), by sequencing (MS/MS) or match between runs (MBR) in MaxQuant. d Comparison of phosphosites (Class 1) quantified in this dataset with those quantified in the data of our previous liver and Hepa 1-6 cell line study (Monetti et al., 2011) analyzed with the same version of MaxQuant (1.5.1.1).

  7. Coverage and overlap compared with published tissue phosphoproteome datasets.
    Supplementary Fig. 3: Coverage and overlap compared with published tissue phosphoproteome datasets.

    a Quantified localized phosphosites reported in deep rodent tissue phosphoproteome studies. b Comparison of the tissue-specific components of mouse tissue phosphoproteomes in this dataset with data published in a large study by Huttlin et al., reveals high overlap of tissue phosphoproteomes with the single-run approach.

  8. Quantitative analysis of tissue phosphoproteomes.
    Supplementary Fig. 4: Quantitative analysis of tissue phosphoproteomes.

    a Heat map of Pearson correlation coefficients, and multi-scatter plots showing reproducibility between cell line and tissue phosphoproteomes. b Principal component analysis (PCA) of mouse cell lines and tissue phosphoproteomes, revealing liver and kidney are similar, while brain is highly distinct. c Quantitative enrichment analysis of mouse brain, kidney and liver phosphoproteomes. Number of significantly enriched (t-test with Permutation-based FDR < 0.01) phosphorylation sites in each tissue compartment, and percentage of the phosphoproteome this represents.

  9. Deep, time-resolved atlas of insulin stimulation in the liver.
    Supplementary Fig. 5: Deep, time-resolved atlas of insulin stimulation in the liver.

    a-b Heat map of Pearson correlation coefficients showing reproducibility between phosphoproteomes of mouse livers stimulated with insulin or PBS (‘Fasted’) for a four ‘early’ and b seven ‘intermediate’ time points. c-d PCA of insulin stimulated mouse liver. e Summary of quantified unique phosphoproteins, phosphopeptides and phosphorylation sites (Class 1) from the complete liver insulin time-series study. f Overlap between the phosphorylation sites quantified in this study, and the combined data from 9 mouse tissues from a large tissue-specific phosphorylation atlas (Huttlin et al., 2010). Data were kindly provided by the authors, and analyzed using MaxQuant using identical settings to enable likewise comparison. g Quantification coverage of unique phosphopeptide species (singly, doubly or greater phosphopeptides) containing only localized (Class 1) phosphosites. Inner circle denotes phosphopeptides quantified in ≥ 3 biological replicates across all insulin stimulated time points for the respective study, and average quantification coverage is shown for this core phosphoproteome (77% and 85% respectively). h Distribution of phosphosite localization probabilities for all phosphosites in the liver time series study, and for Class 1 sites the proportion of phosphoserine (pS), phosphothreonine (pT) and phosphotyrosine (pY) sites. i Dynamic range of MS-signals of phosphopeptides from the liver insulin time-series study span 7 orders of magnitude. Enrichment of protein GO terms (biological process and cellular component) in each intensity abundance range was assessed by Fisher’s exact test (Benjamini-Hochberg FDR < 0.02). The position of GO terms along the horizontal axis represents enrichment of these terms within the respective abundance quartile.

  10. A time-resolved map of liver insulin signaling in vivo.
    Supplementary Fig. 6: A time-resolved map of liver insulin signaling in vivo.

    a Assembly of insulin signaling proteins from multiple database sources. The combined list contains the majority of known insulin signaling molecules, however not all proteins in the database are known to be regulated by phosphorylation. b The canonical insulin signaling network, comprising the PI3K-Akt and MAPK signaling branches. Proteins were curated from multiple database sources in a, as well as from literature. Shown on these proteins are known key insulin-regulated phosphorylation sites. Circles denote phosphorylated residue types (purple = Serine, Blue = Threonine, yellow = Tyrosine, grey = Not quantified) and phosphorylation site position (mouse). Below these are the average response of the phosphosite (fold change compared with PBS control, log2), at 0.5 and 6 min. in our time-resolved liver phosphoproteome data.

Accession codes

Primary accessions

Proteomics Identifications Database

References

  1. Olsen, J.V. & Mann, M. Status of large-scale analysis of post-translational modifications by mass spectrometry. Mol. Cell. Proteomics 12, 34443452 (2013).
  2. Purvis, J.E. & Lahav, G. Encoding and decoding cellular information through signaling dynamics. Cell 152, 945956 (2013).
  3. Kanshin, E., Bergeron-Sandoval, L.P., Isik, S.S., Thibault, P. & Michnick, S.W. A cell-signaling network temporally resolves specific versus promiscuous phosphorylation. Cell Rep. 10, 12021214 (2015).
  4. Olsen, J.V. et al. Global, in vivo, and site-specific phosphorylation dynamics in signaling networks. Cell 127, 635648 (2006).
  5. Vaga, S. et al. Phosphoproteomic analyses reveal novel cross-modulation mechanisms between two signaling pathways in yeast. Mol. Syst. Biol. 10, 767 (2014).
  6. de Graaf, E.L., Giansanti, P., Altelaar, A.F. & Heck, A.J. Single-step enrichment by Ti4+-IMAC and label-free quantitation enables in-depth monitoring of phosphorylation dynamics with high reproducibility and temporal resolution. Mol. Cell. Proteomics 13, 24262434 (2014).
  7. Humphrey, S.J. et al. Dynamic adipocyte phosphoproteome reveals that Akt directly regulates mTORC2. Cell Metab. 17, 10091020 (2013).
  8. Russell, W.K., Park, Z.Y. & Russell, D.H. Proteolysis in mixed organic-aqueous solvent systems: applications for peptide mass mapping using mass spectrometry. Anal. Chem. 73, 26822685 (2001).
  9. Wang, H. et al. Development and evaluation of a micro- and nanoscale proteomic sample preparation method. J. Proteome Res. 4, 23972403 (2005).
  10. Dickhut, C., Feldmann, I., Lambert, J. & Zahedi, R.P. Impact of digestion conditions on phosphoproteomics. J. Proteome Res. 13, 27612770 (2014).
  11. Cox, J. et al. Accurate proteome-wide label-free quantification by delayed normalization and maximal peptide ratio extraction, termed MaxLFQ. Mol. Cell. Proteomics 13, 25132526 (2014).
  12. Monetti, M., Nagaraj, N., Sharma, K. & Mann, M. Large-scale phosphosite quantification in tissues by a spike-in SILAC method. Nat. Methods 8, 655658 (2011).
  13. Lundby, A. et al. Quantitative maps of protein phosphorylation sites across 14 different rat organs and tissues. Nat. Commun. 3, 876 (2012).
  14. Huttlin, E.L. et al. A tissue-specific atlas of mouse protein phosphorylation and expression. Cell 143, 11741189 (2010).
  15. Wisniewski, J.R., Nagaraj, N., Zougman, A., Gnad, F. & Mann, M. Brain phosphoproteome obtained by a FASP-based method reveals plasma membrane protein topology. J. Proteome Res. 9, 32803289 (2010).
  16. Villén, J., Beausoleil, S.A., Gerber, S.A. & Gygi, S.P. Large-scale phosphorylation analysis of mouse liver. Proc. Natl. Acad. Sci. USA 104, 14881493 (2007).
  17. Wilson-Grady, J.T., Haas, W. & Gygi, S.P. Quantitative comparison of the fasted and re-fed mouse liver phosphoproteomes using lower pH reductive dimethylation. Methods 61, 277286 (2013).
  18. Dubois, F. et al. Differential 14–3–3 affinity capture reveals new downstream targets of phosphatidylinositol 3-kinase signaling. Mol. Cell. Proteomics 8, 24872499 (2009).
  19. Narvekar, P. et al. Liver-specific loss of lipolysis-stimulated lipoprotein receptor triggers systemic hyperlipidemia in mice. Diabetes 58, 10401049 (2009).
  20. Alessi, D.R. et al. Characterization of a 3-phosphoinositide-dependent protein kinase which phosphorylates and activates protein kinase Balpha. Curr. Biol. 7, 261269 (1997).
  21. Sarbassov, D.D., Guertin, D.A., Ali, S.M. & Sabatini, D.M. Phosphorylation and regulation of Akt/PKB by the rictor-mTOR complex. Science 307, 10981101 (2005).
  22. Liu, P. et al. Cell-cycle-regulated activation of Akt kinase by phosphorylation at its carboxyl terminus. Nature 508, 541545 (2014).
  23. Hsu, P.P. et al. The mTOR-regulated phosphoproteome reveals a mechanism of mTORC1-mediated inhibition of growth factor signaling. Science 332, 13171322 (2011).
  24. Mok, J. et al. Deciphering protein kinase specificity through large-scale analysis of yeast phosphorylation site motifs. Sci. Signal. 3, ra12 (2010).
  25. Frias, M.A. et al. mSin1 is necessary for Akt/PKB phosphorylation, and its isoforms define three distinct mTORC2s. Curr. Biol. 16, 18651870 (2006).
  26. Ferrell, J.E. Jr. & Ha, S.H. Ultrasensitivity part II: multisite phosphorylation, stoichiometric inhibitors, and positive feedback. Trends Biochem. Sci. 39, 556569 (2014).
  27. Berglund, E.D. et al. Fibroblast growth factor 21 controls glycemia via regulation of hepatic glucose flux and insulin sensitivity. Endocrinology 150, 40844093 (2009).
  28. Kouhara, H. et al. A lipid-anchored Grb2-binding protein that links FGF-receptor activation to the Ras/MAPK signaling pathway. Cell 89, 693702 (1997).
  29. Ortiz-Padilla, C. et al. Functional characterization of cancer-associated Gab1 mutations. Oncogene 32, 26962702 (2013).
  30. Matsumoto, M., Pocai, A., Rossetti, L., Depinho, R.A. & Accili, D. Impaired regulation of hepatic glucose production in mice lacking the forkhead transcription factor Foxo1 in liver. Cell Metab. 6, 208216 (2007).
  31. Brunet, A. et al. 14–3–3 transits to the nucleus and participates in dynamic nucleocytoplasmic transport. J. Cell Biol. 156, 817828 (2002).

Download references

Author information

Affiliations

  1. Department of Proteomics and Signal Transduction, Max Planck Institute of Biochemistry, Martinsried, Germany.

    • Sean J Humphrey,
    • S Babak Azimifar &
    • Matthias Mann

Contributions

S.J.H. and M.M. conceived the project, interpreted data and wrote the manuscript, S.J.H. developed methods, and performed MS experiments, S.B.A. and S.J.H. designed and performed animal experiments. All authors read and approved the manuscript.

Competing financial interests

The authors declare no competing financial interests.

Corresponding author

Correspondence to:

Author details

Supplementary information

Supplementary Figures

  1. Supplementary Figure 1: Precision of the phosphoproteomics workflow. (485 KB)

    Two identical SILAC-labelled HeLa cell populations were mixed a immediately after cell scraping or b after the phosphorylation enrichment workflow and immediately before LC-MS analysis. c-d Distribution and standard deviations of SILAC ratios of phosphopeptides from the experiments shown in the panel above are shown. e Sources of variability in phosphoproteome platform (1) biological, (2) workflow, or (3) LC-MS, investigated by replicates at each level as indicated. f Box-plot of Coefficient of Variation (%) of phosphopeptide intensities calculated from replicate sample data in e. g Average Pearson correlation coefficients (in box inset) for the replicate experiments, and density plots depicting the log2 transformed intensities of two representative replicates. Representative samples shown in plots were chosen based on the closest match between pairwise correlation coefficients and average correlation coefficients for all replicates. h Performance of match between runs (MBR) in transferring identification of phosphopeptides and quantification between replicate workflow samples. Percent valid values are calculated by dividing the number of unique phosphopeptides quantified in all replicates by the total number of possible quantification points, i.e. (∑ phosphopeptides quantified / (total number of phosphopeptides identified x number of replicates)) x 100). Error bars denote SD.

  2. Supplementary Figure 2: Phosphoproteome analysis of mouse cell lines and tissues. (574 KB)

    a Experimental design for phosphoproteome analysis of control (PBS) and insulin-treated mouse liver cell lines (100 nM insulin, 15 minutes) and unstimulated (fed animals) liver, kidney and brain mouse tissues with six biological replicates each. Samples were enriched using the described phosphoproteomics pipeline, and the analysis was performed with single-run measurements lasting 0.5-1.5 day for all replicates of the respective cell or tissue typeper cell or tissue type. b Summary of the identified phosphorylation sites, including enrichment specificity (number of unique phosphopeptides identified divided by the total number of unique peptides identified). c Rate of phosphopeptide identification over the LC gradient in a single sample (Hepa 1-6), by sequencing (MS/MS) or match between runs (MBR) in MaxQuant. d Comparison of phosphosites (Class 1) quantified in this dataset with those quantified in the data of our previous liver and Hepa 1-6 cell line study (Monetti et al., 2011) analyzed with the same version of MaxQuant (1.5.1.1).

  3. Supplementary Figure 3: Coverage and overlap compared with published tissue phosphoproteome datasets. (394 KB)

    a Quantified localized phosphosites reported in deep rodent tissue phosphoproteome studies. b Comparison of the tissue-specific components of mouse tissue phosphoproteomes in this dataset with data published in a large study by Huttlin et al., reveals high overlap of tissue phosphoproteomes with the single-run approach.

  4. Supplementary Figure 4: Quantitative analysis of tissue phosphoproteomes. (465 KB)

    a Heat map of Pearson correlation coefficients, and multi-scatter plots showing reproducibility between cell line and tissue phosphoproteomes. b Principal component analysis (PCA) of mouse cell lines and tissue phosphoproteomes, revealing liver and kidney are similar, while brain is highly distinct. c Quantitative enrichment analysis of mouse brain, kidney and liver phosphoproteomes. Number of significantly enriched (t-test with Permutation-based FDR < 0.01) phosphorylation sites in each tissue compartment, and percentage of the phosphoproteome this represents.

  5. Supplementary Figure 5: Deep, time-resolved atlas of insulin stimulation in the liver. (650 KB)

    a-b Heat map of Pearson correlation coefficients showing reproducibility between phosphoproteomes of mouse livers stimulated with insulin or PBS (‘Fasted’) for a four ‘early’ and b seven ‘intermediate’ time points. c-d PCA of insulin stimulated mouse liver. e Summary of quantified unique phosphoproteins, phosphopeptides and phosphorylation sites (Class 1) from the complete liver insulin time-series study. f Overlap between the phosphorylation sites quantified in this study, and the combined data from 9 mouse tissues from a large tissue-specific phosphorylation atlas (Huttlin et al., 2010). Data were kindly provided by the authors, and analyzed using MaxQuant using identical settings to enable likewise comparison. g Quantification coverage of unique phosphopeptide species (singly, doubly or greater phosphopeptides) containing only localized (Class 1) phosphosites. Inner circle denotes phosphopeptides quantified in ≥ 3 biological replicates across all insulin stimulated time points for the respective study, and average quantification coverage is shown for this core phosphoproteome (77% and 85% respectively). h Distribution of phosphosite localization probabilities for all phosphosites in the liver time series study, and for Class 1 sites the proportion of phosphoserine (pS), phosphothreonine (pT) and phosphotyrosine (pY) sites. i Dynamic range of MS-signals of phosphopeptides from the liver insulin time-series study span 7 orders of magnitude. Enrichment of protein GO terms (biological process and cellular component) in each intensity abundance range was assessed by Fisher’s exact test (Benjamini-Hochberg FDR < 0.02). The position of GO terms along the horizontal axis represents enrichment of these terms within the respective abundance quartile.

  6. Supplementary Figure 6: A time-resolved map of liver insulin signaling in vivo. (814 KB)

    a Assembly of insulin signaling proteins from multiple database sources. The combined list contains the majority of known insulin signaling molecules, however not all proteins in the database are known to be regulated by phosphorylation. b The canonical insulin signaling network, comprising the PI3K-Akt and MAPK signaling branches. Proteins were curated from multiple database sources in a, as well as from literature. Shown on these proteins are known key insulin-regulated phosphorylation sites. Circles denote phosphorylated residue types (purple = Serine, Blue = Threonine, yellow = Tyrosine, grey = Not quantified) and phosphorylation site position (mouse). Below these are the average response of the phosphosite (fold change compared with PBS control, log2), at 0.5 and 6 min. in our time-resolved liver phosphoproteome data.

PDF files

  1. Supplementary Text and Figures (2,214 KB)

    Supplementary Figures 1–6;

  2. Supplementary Information (296 KB)

Excel files

  1. Supplementary Table 1 (4,110 KB)
  2. Supplementary Table 2 (20,388 KB)
  3. Supplementary Table 3 (4,044 KB)
  4. Supplementary Table 4 (35,311 KB)

Additional data