ARTICLE pubs.acs.org/jpr
Systems-wide Proteomic Analysis in Mammalian Cells Reveals Conserved, Functional Protein Turnover Sidney B. Cambridge,†,‡ Florian Gnad,† Chuong Nguyen,† Justo Lorenzo Bermejo,§ Marcus Kr€uger,†,|| and Matthias Mann*,† †
Max-Planck-Institute for Biochemistry, Am Klopferspitz 18, 82152 Munich- Martinsried, Germany Interdisciplinary Center for Neurosciences, University of Heidelberg, Im Neuenheimer Feld 307, 69120 Heidelberg, Germany § Institute of Medical Biometry and Informatics, University Hospital Heidelberg, Im Neuenheimer Feld 305, 69120 Heidelberg, Germany Max-Planck-Institute for Heart and Lung Research, Parkstrasse 1, 61321 Bad Nauheim, Germany
)
‡
bS Supporting Information ABSTRACT: The turnover of each protein in the mammalian proteome is a functionally important characteristic. Here, we employed high-resolution mass spectrometry to quantify protein dynamics in nondividing mammalian cells. The ratio of externally supplied versus endogenous amino acids to de novo protein synthesis was about 17:1. Using subsaturating SILAC labeling, we obtained accurate turnover rates of 4106 proteins in HeLa and 3528 proteins in C2C12 cells. Comparison of these human and mouse cell lines revealed a highly significant turnover correlation of protein orthologs and thus high species conservation. Functionally, we observed statistically significant trends for the turnover of phosphoproteins and gene ontology categories that showed extensive covariation between mouse and human. Likewise, the members of some protein complexes, such as the proteasome, have highly similar turnover rates. The high species conservation and the low complex variances thus imply great regulatory fine-tuning of protein turnover. KEYWORDS: proteomics, SILAC, protein turnover, degradation
’ INTRODUCTION Proteins are continuously synthesized and degraded at steadystate levels.1 For decades, researchers have used labeled amino acids and protein chemistry to study the balance of protein synthesis and degradation in an effort to understand cellular protein homeostasis.2 Recently, high-throughput proteomics studies have begun to provide a global view of individual protein dynamics. Protein turnover has been characterized in different biological systems including bacteria,3 yeast,4,5 mammalian cells,68 whole animals,9,10 and even humans.11 Considerable effort to extract functional information from the turnover data in relation to protein size, isoelectric point, surface area, or cellular location did not yield conclusive results.7,12,13 In fact, reported turnover rates ranged from minutes to days,14,15 yet relatively little is still known about the functional impact of protein dynamics on a large scale. Moreover, protein dynamics have been studied mostly in highly proliferative cells in culture even though most cells in adult organisms do not divide anymore. Stable isotope labeling in cell culture (SILAC)16,17 has become a method of choice for quantitative analysis of protein turnover.18 SILAC involves metabolic labeling of cells in culture with either normal (“light”) or 13C or 15N containing (“heavy”) essential amino acids such as arginine and lysine. The metabolic r 2011 American Chemical Society
incorporation of newly added, pulsed or pulse-chased “heavy” amino acid isotopes can then be used as a direct measure of protein synthesis or degradation. With this approach, we obtained in-depth data sets to extract functional information from turnover dynamics including species conservation, regulation by post-translational modifications and degradation motifs, and protein complexes.
’ EXPERIMENTAL SECTION Cell Labeling and Processing for Mass Spectrometry
Dividing HeLa and C2C12 cells were incubated for two weeks with standard SILAC medium19 (DMEM, 10% dialyzed serum) containing a 1:1 mixture of Arg0:Arg10 and Lys0:Lys8 at previously published concentrations.20 Subsequently, cells were arrested while continuing exposure to the 1:1 mixture. HeLa cells were arrested after one day of confluency by exposure to medium with 2% dialyzed fetal bovine serum for two more days. C2C12 cells were differentiated after two days of confluency by exposure to medium with 2% dialyzed horse serum for six more days.21 Arrested cells were shifted to 100% SILAC medium (Lys8, Received: November 26, 2010 Published: November 03, 2011 5275
dx.doi.org/10.1021/pr101183k | J. Proteome Res. 2011, 10, 5275–5284
Journal of Proteome Research Arg10) for 24 h and subsequently harvested and processed for MS analysis. For the measurement of the external versus internal contribution of amino acids, differentiated unlabeled C2C12 cells were exposed to 100% Lys4, Arg10 (Lys4 instead of Lys8 for improved bioinformatics processing) for 1, 2, 4, 8, or 24 h and were then rapidly chilled, lysed, and processed for MS. Cycloheximide (CHX)-dependent turnover analysis was performed by labeling C2C12 cells in parallel either with Lys0, Arg0 or Lys8, Arg10 for two weeks. Following differentiation, cells were exposed to 2.5 mM CHX for 3 h to rapidly block protein synthesis.22 In a crossover experiment, either Lys0, Arg0 cells were exposed to CHX and mixed with control treated Lys8, Arg10 cells or vice versa. The relative CHX-dependent reduction in individual protein abundance compared to untreated cells was averaged for the two experiments and correlated to the respective kdeg values obtained from the C2C12 turnover measurements. All in vitro cell lysates were processed for MS by OFFGEL technology23 using LysC endopeptidase for digestion of the external versus internal analysis and trypsin for all other experiments. In vivo SILAC labeling of 3 male mice (siblings, age 60 days) was done essentially as described24 using a protein and lysine-free diet (Harland) containing 1% Lys6 according to standard mouse nutritional requirements. After 30 days of labeling, animals were sacrificed and their muscles lysed by douncing with several strokes in modified RIPA buffer containing proteasome inhibitors (Roche). Muscle proteins were processed for MS by standard in gel digestion25 using LysC endopeptidase for digestion of the in vivo samples. LCMS/MS
All LCMS/MS experiments were performed by standard procedures. Briefly, peptides were separated using an Agilent 1100 or 1200 nanoflow LC-System. The HPLC system was coupled to an LTQ-Orbitrap or FT-ICR mass spectrometer (Thermo Fisher Scientific). Survey full scan MS spectra (m/z 3002000) were acquired in the Orbitrap analyzer. The five most intense ions from the survey scan were sequenced by collision induced dissociation in the LTQ. Data were acquired using Xcalibur software.
ARTICLE
box: KENXXX[NDEQ]; F box bTrCP: DSGXXS where X can be any amino acid while the brackets show all of the possible amino acids for that one particular position. To identify PEST sequences, we used the online software EPESTFIND (http:// emboss.bioinformatics.nl/). Correlation Analyses
Nonparametric procedures were used to analyze the turnover correlations since the investigated parameters showed a nonnormal distribution even after logarithmic transformation. Spearman’s rho statistic was used as a rank-based measure of association, and proteins with the highest discrepancy in turnover were identified based on this statistic. Correlation analysis between phosphorylation sequence motif matches and turnover values was performed by ranking the proteins according to their values and collecting all ranks of identified sequence motifs. In parallel, 1000 sets of random matches were iteratively created (“bootstrapping”) where each set contained as many random matches as the number of collected ranks and the means of all 1000 random ranking lists produced the bootstrap distribution. The difference between the bootstrap distribution and the mean of the collected ranks was computed in standard deviations and yielded a probability value for a null correlation. The approach was also used for the variance cluster analysis of values in human protein complexes by ranking the values of individual complexes. To address the issue of multiple hypotheses testing, we also applied a Bonferroni correction. Half-life Calculation
Half-life t 1/2 was calculated using the loss of “light” label in a first-order reaction equation: N t ¼ N t0 e-kt
ð1Þ
t 1=2 ¼ ln 2=k
ð2Þ
MaxQuant
with Nt = 1/(r + 1) and Nt0 = 1/(1 + 1) = 0.5, [turnover ratio r = H/L (heavy peak/light peak); e.g. r = 3 = 3/1 = H/L], and k is kdeg, the degradation rate constant. In general, RIA (relative isotope abundance) is given as H/(H + L). Because our samples contain 50% H heavy label at t = 0, we corrected for this by calculating and presenting ΔH/(H + L) so that our RIA values can be compared to standard RIA values where H = ΔH.
Mass spectra were analyzed using the in house developed software MaxQuant.26 The data was searched against the Mouse or Human International Protein Index protein sequence database (IPI, version 3.37) and concatenated with reversed copies of all sequences using Mascot (version 2.2.04, Matrix Science). Initial maximum allowed mass deviation was set to 7 ppm for monoisotopic precursor ions and 0.5 Da for MS/MS peaks. The required false positive rate was set to 1% at the peptide level, the required false discovery rate was set to 1% at the protein level and the minimum required peptide length to 6 amino acids. In addition to the protein false discovery rate threshold, we required at least two sequenced peptides (one unique) for protein identification and three measured peptides for quantitation. For crossover analyses, we only allowed proteins that showed an inverse trend (>1 or
