A Dataset of Human Liver Proteins Identified by Protein Profiling Via Isotope-coded Affinity Tag (ICAT) and Tandem Mass Spectrometry

This human liver proteomic dataset consists of three subdatasets generated from three protein profiling experiments using the following samples: human liver carcinoma cells (Huh7), primary cultures of human fetal hepatocytes (HFH)¹ (4), and an immortalized cell line derived from human fetal hepatocytes (HH4).²

The Huh7 and HFH cells were selected to study the interferon response in transformed (Huh7) and nontransformed (HFH) human liver cells, respectively. About 2 × 10⁷ cells were either interferon-α_2b- (400 IU/ml Intron-A; Schering-Plough Co., Kenilworth, NJ) or mock- treated for 16 h before harvest. The cells were lysed, and cell lysates were fractionated into cytosolic, membrane, and nuclear fractions by sequential differential centrifugation at 3,000 × g (nuclear fraction from the pellet) and 100,000 × g (cytosolic fraction from the supernatant and membrane fraction from the pellet). Proteins from each subcellular fraction were labeled with isotopically light- (¹²C, for stimulated cells) or heavy- (¹³C, for control cells) ICAT reagents following the manufacturer’s protocol (Applied Biosystems, Foster City, CA). Corresponding isotopically light- and heavy-labeled samples were then combined and digested with trypsin (Promega, Madison, WI). The resulting peptides were separated by strong cation exchange chromatography, as previously described (5), and affinity purified by avidin cartridges following the manufacturer’s protocol (Applied Biosystems), through which the cysteine (Cys)-containing peptides were enriched. The Cys-containing peptides from ∼20 fractions purified above were then subjected to μLC-ESI-MS/MS using an LCQ-DECA-XP ion-trap mass spectrometer (ThermoFinnigan, San Jose, CA) as previously described (6, 7).

All observed MS/MS spectra were subsequently subjected to search against the human International Protein Index (IPI) database (www.ebi.ac.uk/IPI/IPIhelp.html) (v2.28) using the SEQUEST™ software. Search parameters for the cleavable ICAT-labeled samples used in this study were the following: +227.13 Da for static modification on cysteine residues labeled with cleavable ICAT, +9 Da for ¹³C isotopic ICAT-labeled cysteine, +16 Da for oxidized methionine; mass tolerance ± 3 Da; restriction on Cys-containing peptides; and no proteolytic enzyme specified. Accuracy of the SEQUEST assignments of MS/MS spectra to peptide sequences was estimated by the PeptideProphet™ software based on a statistical model (2). For each identified peptide, a probability score was computed on a scale of 0 (for “incorrect”) to 1 (for “correct”) based on match of the peptide sequence to the tandem mass spectra and the trypsin proteolytic pattern. These assigned peptides were then subjected to ProteinProphet™ analysis to assign a protein probability score for each identified protein or related protein group inferred from the peptide data (3). The protein probabilities, again on a scale of 0 to 1, discriminate correct (p = 1) from incorrect (p = 0) protein identifications. Validation of initial data base search results on the basis of statistical modeling allows the presentation of large-scale proteomics datasets with known sensitivity for positive identifications and error rates for false positive identifications.

In the Huh7 cells, 23,310 peptides, with a PeptideProphet probability score p ≥ 0.05, were obtained and included for subsequent ProteinProphet analysis. The sequences of the assigned peptides, together with their IPI reference name, PeptideProphet probability, and calculated and measured mass, are presented as reference for future proteomics studies (Supplemental Table Ia). Using ProteinProphet software, 1,146 proteins or related protein groups were identified with an arbitrary probability cut-off of p ≥ 0.9 (Supplemental Table IIa). This value corresponded to 87.5% sensitivity (i.e. 87.5% of all possible identifications were made) and a false positive error rate of 0.7% (Supplemental Fig. 1A). This type of analysis also allows the investigator to compute the implications of changing the probability value on sensitivity and false positive error rate. For example, a reduction of the protein probability from 0.9 to 0.5 in this subdataset increased the number of protein identifications to 1,296, increased the sensitivity to 95.7%, and also increased the error rate to 3.8%.

Similarly, a total of 31,641 peptides (p ≥ 0.05) were obtained from the analysis of the HFH cells (Supplemental Table Ib). These assigned peptides were used for subsequent ProteinProphet analysis and lead to the identification of 1,235 proteins and related protein groups (p ≥ 0.9) (Supplemental Table IIb), which corresponded to 86.5% sensitivity and 0.8% error rate (Supplemental Fig. 1B). A reduction of protein probability threshold to 0.5 resulted in 1,430 protein identifications with 96% sensitivity and 4.6% error rate.

The HH4 cells are immortalized human hepatocytes. Two HH4-based cell lines were constructed, based on an ecdysone-inducible expression system (8), to induce expression of the entire hepatitis C virus (HCV) ORF or green fluorescence protein, respectively.³ The ecdysone-regulated gene expression system consists of a modified ecdysone receptor (a heterodimer of VgEcR and RXR) that binds to its recognition sequence (5xE/GRE) and associates with transcription corepressors to repress the downstream promoter. Upon induction by a plant-derived ecdysone analog ponasterone A (ponA), ponA binds to the VgEcR to release the corepressors and recruit cotransactivators to activate transcription of the downstream target genes. We performed two ICAT labeling experiments to investigate HCV-mediated protein expression profiles in human hepatocytes. The first compared total cell extracts from HCV ORF-induced cells (ponA+, light-labeled) with noninduced cells (ponA-, heavy-labeled). The second experiment compared total cell extracts from cells with induced HCV proteins (light-labeled) with cells carrying induced green fluorescence protein (heavy-labeled). The labeled samples were subjected to the same analyses as described above. From HH4 cells we obtained a total of 28,029 peptide assignments with p ≥ 0.05 (Supplemental Table Ic), which contributed to identification of 1,318 (p ≥ 0.9) or 1,476 proteins and related protein groups (p ≥ 0.5) (Supplemental Table IIc and Supplemental Fig. 1C).

Taken together, proteomics analyses of the three human liver cells of both transformed and nontransformed cells lead to a total of 2,159 (p ≥ 0.9) or 2,486 (p ≥ 0.5) unique protein identifications. Among them, 496 (p ≥ 0.9) or 540 (p ≥ 0.5) were found in all three liver cells, while 337/397, 364/457, and 414/456 proteins and related protein groups were uniquely observed in the Huh7, HFH, and HH4 cells, respectively (p ≥ 0.9/p ≥ 0.5). Comparison of the three proteomics subdatasets from human liver cells (Supplemental Table IIIa for p ≥ 0.9 and IIIb for p ≥ 0.5) are also shown as a Venn Diagram (Fig. 1) using the on-line Create-A-Venn system at www.venndiagram.com. This human liver proteomics datasets with more than 2,000 protein identifications, presented in a statistically validated and transparent way, describes a possible mechanism for publishing large-scale protein identification datasets in the literature and for data comparison from different experiments.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Comparison of the protein identifications of the three subdatasets from the Huh7, HFH, and HH4 cells. Identified protein or related protein groups from the three proteomics subdatasets (Supplemental Table IIa for Huh7, IIb for HFH, and IIc for HH4) were compared based on the IPI identification of each entry. The results were displayed by Venn Diagram (www.venndiagram.com) at the protein probability threshold of 0.9 and 0.5, respectively.

• © 2004 The American Society for Biochemistry and Molecular Biology