SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

We report a comprehensive and quantitative analysis of the mouse liver and plasma proteomes. The method used is based on extensive fractionation of intact proteins, further separation of proteins based on their abundance and size, and high-accuracy mass spectrometry. This analysis reached a depth in proteomic profiling not reported to date for a mammalian tissue or a biological fluid, with 7099 and 4727 proteins identified with high confidence in the liver and in the corresponding plasma, respectively. This method allowed for the identification in both compartments of low-abundance proteins such as cytokines, chemokines, and receptors and for the detection in plasma of proteins in the pg/mL concentration range. This method also allowed for semiquantitation of all identified proteins. The calculated abundance scores correlated with the abundance of the corresponding transcripts for the large majority of the proteins identified in the liver. Finally, comparison of the liver and plasma datasets demonstrated that a significant number of proteins identified in the liver can be detected in plasma. These included proteins involved in complement and coagulation, in fatty acid, purine and pyruvate metabolism, in gluconeogenesis and glycolysis, in protein ubiquitination, and in insulin, interleukin-4, epidermal growth factor, and platelet-derived growth factor signaling. Conclusion: This in-depth analysis of the mouse liver and corresponding plasma proteomes provides a strong basis for investigations of liver pathobiology and biology that employ mouse models of hepatic diseases in an effort to better understand, diagnose, treat, and prevent human hepatic diseases. (HEPATOLOGY 2008.)

Mapping of the liver proteome content is expected to facilitate the understanding of hepatic biological processes and their dysregulation in disease.1 In addition, the discovery of novel protein biomarkers for the early detection of liver disease and for the assessment of a therapy's efficacy and side effects is urgently needed. Plasma represents a promising source of such markers because it has the advantage of being easily accessible. However, unless we have knowledge of which proteins compose the plasma in health and which proteins of the liver are also present in plasma, critical elements of how to design biomarker studies will evade us.

To date, reported proteome profiling of complex biological mixtures such as biological fluids, tissues, or cells has resulted in the identification of only a small percentage of the expressed proteins, missing most proteins of low abundance. The use of advanced mass spectrometers with high mass accuracy and high sensitivity in combination with extensive sample pre-fractionation has significantly increased the coverage of complex proteomes. In addition, developments in immunoaffinity depletion of highly abundant proteins and in higher dimensional separation strategies have enabled the plasma proteome to be profiled with greater dynamic range of coverage, allowing proteins at the ng/mL levels to be identified confidently.2–5 The ability to estimate the abundance of the identified proteins is also an important goal for proteomic studies aimed at discovering disease biomarkers. Several studies have shown that the total number of peptide hits per protein can be used as a semiquantitative approach.6–10 We have developed a pipeline based on multidimensional separation of intact proteins with particular attention to the isolation of low-abundance proteins and exhaustive liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. We applied this pipeline to the analysis of the mouse liver and of the corresponding plasma, resulting in comprehensive proteomic profiling of an organ and of a biological fluid. Protein abundance was estimated for each of the identified proteins. Finally, integration between the liver and the plasma proteome datasets and integration between the liver proteome and transcriptome were performed.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Mouse Samples.

Four adult C57BL/6 mice were used for this study. The mice were housed and sacrificed according to institutional guidelines. Blood samples were obtained by cardiac puncture and collected in ethylenediaminetetraacetic acid tubes, and plasma was processed by centrifugation at 1300g for 10 minutes. Liver tissue samples from the same 4 mice were obtained by necropsy and immediately snap-frozen in liquid nitrogen.

Protein Extraction and Separation.

The 4 liver tissue samples were separately ground on dry ice and an equal amount from each sample was subsequently pooled. Proteins were extracted twice from 40 mg of the pooled sample in 1 mL lysis buffer (5 M urea, 2 M thiourea, 2% wt/vol n-Octyl-β-D-glucopyranoside, 40 mM Trisbase, and 1 mM phenylmethylsulfonyl fluoride). After centrifugation at 16,100g at 4°C for 1 hour, the pellet fraction was solubilized in Laemmli buffer, and the proteins from the supernatant were separated using the Alliance 2-D Bioseparations System (Waters Corporation, Milford, MA). An anion exchange column, BioSuite Q 10 μm (Waters Corporation, Milford, MA) was used for the first dimension with 40 mM Trisbase/4 M Urea/6% isopropanol (pH 7.8) as the first eluent and 1 M NaCl/40 mM Trisbase/4 M Urea/6% isopropanol (pH 7.8) as the second eluent. Eight stepwise gradients were performed consisting of 0, 100 mM, 200 mM, 350 mM, 500 mM, 650 mM, 800 mM, and 1000 mM NaCl. The reversed-phase columns, Symmetry300 C4 3.5 μm (Waters Corporation, Milford, MA), were used for separation of the fractions obtained from the first dimension steps. Two reversed-phase columns were switched through the column selector. The first eluent (A) was 0.1% trifluoroacetic acid in high-performance liquid chromatography (HPLC) grade water, and the second eluent (B) was 0.1% trifluoroacetic acid in acetonitrile. The gradients were run at a flow rate of 0.5 mL/minute according to the following steps: 0 minutes 10% B, 15 minutes 10% B, 50 minutes 95% B, 52.5 minutes 95% B, 55 minutes 5% B, 65 minutes 5% B. A total of 238 fractions were collected. Some adjacent fractions were combined, leading to a final number of 34 samples. All fractions were lyophilized and resuspended in Laemmli buffer.

Equal volumes of the 4 plasma samples were pooled. A total of 320 μL pooled sample was immunodepleted of albumin, transferrin, and immunoglobulin G using the Ms-3 immunoaffinity column (Agilent Technologies, Santa Clara, CA). Proteins from the immunodepleted fraction were separated using the same method as for liver tissue proteins but for the following exceptions. For separation in the first dimension, 25 mM Trisbase/1 M Urea/3% isopropanol (pH 8.7) was used as the first eluent, and 1 M NaCl/25 mM Trisbase/1 M Urea/3% isopropanol (pH 8.7) was used as the second eluent. A total of 253 fractions were collected. Some adjacent fractions were combined, leading to a final number of 35 samples. All fractions were lyophilized and resuspended in Laemmli buffer.

Proteins obtained from the 2-dimensional (2-D) HPLC separation of the immunodepleted plasma, from the 2-D HPLC separation of the liver tissue, and from the pellet of the liver tissue were further separated by 12% sodium dodecyl sulfate polyacrylamide gel electrophoresis. The gels were stained with colloidal Coomassie blue G-250, and each lane was cut into pieces. Gel pieces were combined into 43 individual samples for plasma and into 37 individual samples for liver tissue according to protein size and abundance. After destaining with 50% acetonitrile and incubation in 250 μL digestion buffer (10% acetonitrile, 50 mM ammonium bicarbonate), the gel pieces were dehydrated with 100% acetonitrile and dried using a speed vacuum. Gel pieces were incubated with 10 μL 6.7 ng/μL trypsin in digestion buffer overnight at 37°C. The reaction was stopped with 15 μL extraction buffer (2% formic acid/3% acetonitrile), and the supernatants were collected.

Mass Spectrometry and Database Searching.

The generated peptide samples were desalted using Symmetry C18 de-salt columns (Waters Corporation, Milford, MA) and subjected in duplicate to nanoflow LC-MS/MS analysis with a nano-UPLC system (Waters Corporation, Milford, MA) coupled to a hybrid 7-Tesla linear ion-trap Fourier-transform ion cyclotron resonance mass spectrometer (LTQ-FT, Thermo Scientific, Waltham, MA). Peptides were separated on a reversed-phase column (75 μm × 250 mm) packed with Magic C18AQ (5-μm 100Å resin; Michrom Bioresources, Auburn, CA) and directly mounted on the electrospray ion source. We used a 60-minute gradient from 10% to 40% acetonitrile in 0.1% formic acid at a flow rate of 300 nL/minute. A spray voltage of 1600 V was applied. The LTQ-FT instrument was operated in the data-dependent mode, switching automatically between MS survey scans in the Fourier-transform ion cyclotron resonance (FTICR; target value 1,600,000, resolution 100,000, and injection time 1.5 seconds) with MS/MS spectra acquisition in the linear ion trap. The 5 most intense ions from the Fourier-transform (FT) full scan were selected for fragmentation in the linear ion trap by collision-induced dissociation with a normalized collision energy of 30% at a target value of 10,000 (injection time, 400 milliseconds). Selected ions were dynamically excluded for 60 seconds. The absolute average mass accuracy for the parent ion was less than 5 ppm. The data files were converted to the m/z XML generic format and searched against the nonredundant mouse International Protein Index protein sequence database (version 3.20, http://www.ebi.ac.uk/IPI/) using the X!Tandem Search Engine. The following search criteria were used in all cases: trypsin specificity, 2.5 Da of mass accuracy for the parent ion and methionine oxidation as a variable modification.

Proteomic Data Analysis.

To obtain reliable protein identifications from the search results, ProteinProphet, a statistical model that computes protein probabilities based on peptides assigned to MS/MS spectra,11 was used. Protein identifications from ProteinProphet can result in either a single protein or a protein group. The scores of the peptides associated with these identified proteins were computed using PeptideProphet. Relative protein abundance scores were calculated based on total peptide counts normalized by the length of the protein, as previously described.7 This ratio was further normalized to account for the total amount of protein in the mixture. The Ingenuity Systems bioinformatics tool (www.ingenuity.com; version 5.0) was used to analyze pathways associated with proteins identified in both liver tissue and plasma. Canonical pathways with P values less than 0.05 were selected.

Transcriptomic Analysis.

Total RNA was extracted from the same pool of liver samples used for the proteomic analysis. The first-strand complementary DNA, the double-strand complementary DNA, and complementary RNA (cRNA) were synthesized, and cRNA was fragmented using Affymetrix (Santa Clara, CA) kits and guidelines. All cRNA final products were tested for amount and integrity before microarray hybridization. cRNA samples were processed on Affymetrix GeneChip Mouse Genome 430 2.0 Array with strict adherence to the labeling, fragmentation, and hybridization protocols provided by Affymetrix. GeneChip image analysis was performed using GCOS v1.4 (Affymetrix, Santa Clara, CA). Probe-level analysis, pre-processing, and normalization steps were carried out using GeneTraffic 3.2.-11 (Iobion Stratagene Microarray Analysis Software, Iobion, La Jolla, CA). The MATLAB “Plot and Calculate Confidence Bounds” function (www.mathworks.com) was used to calculate confidence bounds for the linear regression data.

Results

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Comprehensive and Quantitative Proteome Profiling of Mouse Liver Tissue.

To comprehensively profile the mouse liver proteome, we applied an approach based on 3-dimensional separation of intact proteins. Equal amounts of ground liver tissue from 4 healthy adult C57BL/6 mice were pooled and, after protein extraction and centrifugation, the supernatant and pellet fractions were collected. Proteins from the supernatant were separated by 2-D HPLC (anion exchange in tandem with reversed-phase chromatography) followed by 1-dimensional sodium dodecyl sulfate polyacrylamide gel electrophoresis (Fig. 1). Proteins were then extracted based on their size and abundance, resulting in 37 fractions. The corresponding peptides generated by in-gel trypsin digestion were analyzed in duplicate by LC-MS/MS. A total of 7099 proteins (7090 unique proteins and 9 protein groups), 4336 proteins (4332 unique proteins and 4 protein groups), and 2860 proteins (2857 unique proteins and 3 protein groups) were identified with a ProteinProphet score ≥ 0.9 and with at least 2, 3, or 4 peptides, respectively (Fig. 2A) (Supplementary Table 1). In this data set, a 0.9 ProteinProphet score corresponded to a rate of misclassification of 1.2%.

thumbnail image

Figure 1. Schematic view of the protein extraction and 3-dimensional separation process. On-line 2-D HPLC separation of the supernatant fraction consisted of 8 stepwise gradients in the 1st dimension AEX liquid chromatography separation, each of them further separated in the 2nd dimension by reversed-phase liquid chromatography separation. Proteins were further separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis. The fractionation process generated approximately 40 samples analyzed in duplicate by LC-MS/MS. AEX, anion exchange.

Download figure to PowerPoint

thumbnail image

Figure 2. Liver tissue protein identifications and quantitation. (A) Number of protein identifications classified by number of peptides used for identification. (B) Protein abundance distribution and associated functions.

Download figure to PowerPoint

A relative abundance score was calculated for each of the 7090 unique proteins identified with at least 2 peptides, using peptide counts normalized by the protein length, as described7 (Fig. 2B). Among the 7090 proteins identified were growth factors, cytokines, and chemokines, receptors, transcription factors, and liver-enriched enzymes. The growth factors, cytokines, and chemokines included epidermal growth factor, interleukin-16, interferon-beta, fibroblast growth factor-15, tumor necrosis factor, interleukin-17, transforming growth factor beta 1, and platelet-derived growth factor C, and their abundance spanned the lowest abundance score interval. The receptors included interleukin-6 receptor, interleukin-12 receptor, interleukin-1 receptor, interleukin-18 receptor, transforming growth factor beta receptor, fibroblast growth factor receptor, epidermal growth factor receptor, interferon gamma receptor, platelet-derived growth factor beta receptor, and toll-like receptors. The transcription factors included signal transducer and activator of transcription 2, CCAAT enhancer binding protein alpha, and retinoic acid receptor. The abundance scores for both groups of proteins spanned low abundance score intervals. The liver-enriched enzymes included fatty acid synthase, cytochrome P450s, liver carboxylesterases, alanine aminotransferase, methionine adenosyltransferase I, alpha, argininosuccinate lyase and synthase, carbamoyl-phosphate synthase, alcohol dehydrogenase, glutathione S-transferases, aspartate aminotransferase, fructose-1,6-bisphosphatase, and glutathione peroxidase, and spanned high abundance score intervals. In conclusion, this approach provided an extensive coverage of the mouse liver proteome with the identification of proteins over a wide range of expression levels. This approach also provided an accurate estimate of the abundance of the identified proteins.

Comprehensive and Quantitative Proteome Profiling of the Corresponding Plasma.

We applied a similar approach to profile the proteome of the plasma collected from the same 4 mice (Fig. 1). Equal volumes of plasma from each of the same 4 healthy adult C57BL/6 mice used for the liver tissue analysis were pooled and, after immunodepletion of the abundant proteins, albumin, transferrin, and immunoglobulin G, proteins were separated and analyzed as for the tissue. In our plasma data set, a 0.9 ProteinProphet score corresponded to a rate of misclassification of 1.7%. A total of 4727 proteins (4721 unique proteins and 6 protein groups), 2463 proteins (2461 unique proteins and 2 protein groups), and 1343 proteins (1341 unique proteins and 2 protein groups) were identified with at least 2, 3 or 4 peptides, respectively (Fig. 3A) (Supplementary Table 2).

thumbnail image

Figure 3. Plasma protein identifications and quantitation. (A) Number of protein identifications classified by number of peptides used for identification. (B) Protein abundance distribution and associated functions.

Download figure to PowerPoint

As for the tissue, abundance scores were calculated for the 4721 unique proteins identified (Fig. 3B). Proteins such as A disintegrin-like and metalloproteinases with thrombospondin motifs, A disintegrin and metalloproteinase proteins, cadherins, catenins, semaphorins, colony-stimulating factor 1, leukemia inhibitory factor, interleukin-33, fibroblast growth factor-15, and interferon alpha were identified with low abundance scores; insulin-like growth factor binding proteins, interleukin-3, transforming growth factor beta 3, apolipoprotein A5 (ApoA5), alpha-fetoprotein, and chemokine (C-X-C motif) ligand 14 were identified with intermediate abundance scores; and apolipoproteins, complement components, alpha-1-antitrypsins and fibrinogens were identified with high abundance scores. To further validate the quantitation method used, we compared the calculated abundance with published concentrations for 50 plasma proteins spanning the range of concentrations from 10 pg/mL (inhibin alpha) to 2.65 mg/mL (fibrinogen). These 50 proteins included 4 proteins in the pg/mL range, 11 proteins in the ng/mL range, 29 proteins in the μg/mL range, and 6 proteins in the mg/mL range (Fig. 4). The calculated relative abundance correlated remarkably well with the reported concentration of the corresponding human proteins in human plasma (R = 0.82) (Fig. 4). Overall, these data demonstrate that the method used correctly estimated the abundance of proteins of high abundance (mg/mL range) such as fibrinogen, apolipoprotein A-I, complement C3, and alpha-1-antitrypsin. Our method also correctly estimated the abundance of proteins of low-abundance (pg/mL to ng/mL range) such as macrophage colony-stimulating factor 1, inhibin alpha, interferon alpha, interleukin-3, urokinase-type plasminogen activator, alpha-fetoprotein, insulin-like growth factor 1, and hepatocyte growth factor activator. Therefore, this method constitutes a robust approach for the identification and simultaneous quantitation of plasma proteins with level dynamics over 9 orders of magnitude, some of them at low abundance and of potential clinical relevance.

thumbnail image

Figure 4. Comparison between calculated abundance and reported concentrations for 50 plasma proteins. The correlation between relative abundance and reported concentrations for 50 proteins identified in plasma is shown. Abbreviations: CSF1, colony-stimulating factor 1; PLAU, urokinase-type plasminogen activator; IGF1, insulin-like growth factor 1; HGFAC, hepatocyte growth factor activator; PROZ, vitamin K-dependent protein Z; IGFALS, insulin-like growth factor-binding protein complex acid labile chain; RBP4, retinol binding protein 4; SERPING1, C1 esterase inhibitor. Most reported concentrations were obtained from Specialty Laboratories (www.specialtylabs.com). The remaining reported concentrations were obtained from the literature as referenced: interleukin-328; CSF129; PLAU30; beta-enolase31; enolase 1, alpha32; IGF133; uromodulin34; lactotransferrin35; HGFAC, serum amyloid A protein, IGFALS, factor XII, apolipoprotein C-II4; beta-2-microglobulin36; PROZ37; protein C38; apolipoprotein D39; apolipoprotein C-III40; vitamin D-binding protein41; alpha-2-HS-glycoprotein42; hemopexin.43

Download figure to PowerPoint

Comparison Between the Liver and the Plasma Proteomes.

We compared the proteins identified in liver tissue with those identified in plasma. A total of 1818 proteins were common to both data sets and corresponded to 25.6% of the proteins identified in the liver tissue and to 38.5% of the proteins identified in plasma (Fig. 5A). These results demonstrated a significant overlap between the liver and plasma proteomes. The abundance distributions of the 1818 proteins common to liver and plasma proteomes was overall very similar to the abundance distribution of all proteins in both compartments (Fig. 5B compared with Fig. 2B and Fig. 5C compared with Fig. 3B). However, only a weak correlation was observed between the abundance of these proteins in the liver tissue and in the plasma (R = 0.52) (Fig. 5D).

thumbnail image

Figure 5. Number of proteins commonly identified in liver tissue and plasma and relative abundance. (A) Overlap between proteins identified in liver tissue and plasma. (B) Protein abundance distribution in liver tissue for the common proteins. (C) Protein abundance distribution in plasma for the common proteins. (D) Comparison between relative abundance in liver tissue and relative abundance in plasma.

Download figure to PowerPoint

To further characterize the biological significance of these 1818 proteins to the liver, we used the Ingenuity Systems data mining tool. Twenty-three metabolic and signaling pathways were associated with statistical significance (P < 0.05), with the proteins commonly detected in the liver tissue and plasma (Fig. 6). These included major pathways associated with known liver functions such as purine metabolism (xanthine oxidase and urate oxidase); complement and coagulation (factor V, protein C, fibrinogens, complement components, and mannose-binding lectin 2); fatty acid metabolism (cytochrome P450s and acyl-coenzyme A dehydrogenase, very long chain); estrogen receptor signaling (phosphoenolpyruvate carboxykinase 1 and peroxisome proliferator-activated receptor binding protein); pyruvate metabolism (pyruvate carboxylase and glyoxalase I); insulin receptor signaling (insulin receptor substrate 1 and glycogen synthase kinase 3 alpha); glycolysis/gluconeogenesis (glucose phosphate isomerase); epidermal growth factor signaling (epidermal growth factor receptor, mitogen-activated protein kinase kinase 4, and inositol 1,4,5-triphosphate receptors); platelet-derived growth factor signaling (inositol polyphosphate phosphatase-like 1 and tyrosine kinase 2); and Janus kinase/signal transducer and activators of transcription pathway (signal transducer and activation of transcription 2). In addition, the proteins common to liver tissue and plasma included very-low-abundance proteins such as cadherins, A disintegrin-like and metalloproteinases with thrombospondin motifs, A disintegrin and metalloproteinase proteins, and fibroblast growth factor-15.

thumbnail image

Figure 6. Pathways associated with proteins common to liver tissue and plasma. Proteins common to liver tissue and plasma were analyzed using Ingenuity and pathways with P-values < 0.05 were selected.

Download figure to PowerPoint

Correlation Between Protein and Transcript Abundance.

Finally, we analyzed the relationship between protein and transcript abundance for the 7090 proteins identified in the liver tissue. Total RNA was prepared from the same pooled liver tissue sample used for the proteomics analysis. Comparison between the genes associated with the 7090 proteins identified in liver tissue and the 45,037 probe sets present on the microarray identified 5680 genes common to both. The correlation coefficient between protein abundance and transcript intensity for these 5680 genes (R = 0.62) demonstrated a positive correlation between abundance at the protein and at the transcript level (Fig. 7A). To assess whether the data come from a normal distribution, a normal probability plot was generated (Fig. 7B). The dots show the empirical probability versus the data value for each protein/gene in our dataset. Because the line connects the 10th and the 99th percentiles of the data and represents a robust linear fit, insensitive to the extremes of the dataset, and because all the data points fall near the line, the plot is clear evidence that the distribution is normal.

thumbnail image

Figure 7. Comparison between the proteome and the transcriptome of liver tissue. (A) Plot of relative abundance for individual proteins and associated transcript hybridization intensities. The trend line (black) and 95% confidence bounds (dashed) are shown on the scatter plot. (B) Normal probability plot of ratios of transcript hybridization intensities to relative protein abundance.

Download figure to PowerPoint

Of the 1818 proteins common to liver tissue and plasma, 1529 corresponding genes were represented on the microarray. The correlation between protein abundance and transcript intensity for these genes was lower (R = 0.58) (data not shown). Confidence bounds corresponding to 95% confidence interval were calculated for the complete data set (Fig. 7A). Only 214 genes were found outside the confidence interval, with 145 genes above the upper boundary line and 69 genes below the lower boundary line. Remarkably, 56% of the above outliers and 26% of the inliers corresponded to extracellular proteins. Altogether, these results indicate a reduced correlation between protein abundance and transcript intensity for proteins found in the plasma compared with proteins more confined to liver tissue. Some of the genes in the above outliers region include afamin, alpha-2-HS-glycoprotein, apolipoproteins, complement components, collagen, carboxypeptidase N, fibrinogens, vitamin D binding protein, haptoglobin, histidine-rich glycoprotein, protein C, retinol binding protein 4, serum amyloids, alpha-1-antitrypsin, and vitronectin.

Discussion

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Mouse models have been developed to better understand the underlying mechanisms of hepatic diseases such as liver fibrosis, steatohepatitis, and hepatocellular carcinoma. These genetically engineered mouse models of human diseases have been shown to be helpful for understanding human disease progression.12, 13 They are also helpful for minimizing background genetic variations between target and control samples, making them attractive for studies aimed at discovering novel diagnostics and therapeutics. A comprehensive understanding of the mouse liver and plasma proteomes is critical for studies using mouse models aimed at identifying protein markers of liver diseases.

Given the tremendous complexity of organ and plasma proteomes and the dynamic range of protein abundance (6-8 orders of magnitude in cells and 10-12 orders of magnitude in plasma14), different strategies (such as depletion of highly abundant proteins, extensive separation, specific enrichment of subproteome) have been developed. These approaches used in combination with high-resolution mass spectrometry allowed for the identification of plasma proteins in the low ng/mL range3, 4 and for the identification in the liver of between 2000 and 2500 proteins.15, 16 We are reporting a method based on extensive 3-dimensional separation of intact proteins according to their charge, hydrophobicity, and molecular mass that significantly improved detection and allowed for an even deeper coverage of the liver and plasma proteomes. While increasing analysis time, such broad fractionation resulted in the largest organ proteome to date and reached impressive sensitivity levels with the detection of proteins that span at least 9 orders of magnitude in concentration in plasma (pg/mL to mg/mL). Such results were obtained using lower amounts of material compared with previous studies. In addition, separations of intact proteins on the basis of their different masses or other properties can be used to identify different protein isoforms.5, 17

Quantitative analysis based on the number of peptides identified per protein is increasingly used.6–10 An important consideration with spectrum counting is the fact that small proteins tend to have fewer peptides identified per protein compared with large proteins. Therefore, it is important to take into consideration the length of a protein when determining protein abundance using spectrum counting. This report further demonstrates that spectral counting is a powerful quantitative proteomic approach because the calculated abundance scores strongly correlated with the reported concentrations in plasma for 50 proteins with concentrations ranging from pg/mL to mg/mL. The overall concordance between expression levels of the identified proteins and corresponding transcripts was also found to be good, suggesting that the transcript patterns were highly predictive of the corresponding protein levels. Lack of correlation was observed largely for extracellular proteins, with more than 50% of the proteins showing high transcript level and low protein abundance corresponding to proteins detected in plasma, in agreement with a recent report showing discordance between transcript and protein levels for proteins such as complement components.18

A big debate in the clinical proteomics field is which type of sample—blood or diseased tissue—should be used to discover serologic biomarkers.19, 20 Proteomic studies using blood as a source of proteins still have not resolved the question of whether plasma or serum can serve as a window into the state of a patient's disease, and there is currently no scientific evidence that a linear relationship exists between protein changes in diseased tissue and protein changes in blood. This is the first study comparing the proteomes of an organ and plasma obtained from the same animal or individual. The proteins common to liver tissue and plasma were associated with 23 metabolic and signaling pathways, some of them associated with well-known liver functions. Because the liver is responsible for producing most of the plasma proteins involved in the complement and coagulation pathway, it was not surprising to have identified numerous proteins associated with this pathway among the proteins common to liver tissue and plasma. Other proteins involved in major metabolic pathways of relevance to the liver and identified in both the liver and the plasma included proteins involved in gluconeogenesis and glycolysis and proteins involved in fatty acid metabolism. Examples include pyruvate carboxylase, which in the liver provides oxaloacetate for gluconeogenesis,21 acyl-coenzyme A dehydrogenase, very long chain, which has been linked to fatty changes in the liver,22 and xanthine oxidase, which has been hypothesized to be responsible for the increased lipid peroxidation associated with transplantation of steatotic livers.23 Proteins associated with signaling pathways of relevance to liver biology (platelet-derived growth factor, interleukin-4, epidermal growth factor, and insulin receptor signaling pathways) were also identified among the proteins common to liver tissue and plasma. They included peroxisome proliferator-activated receptor binding protein, an important player in liver regeneration,24 epidermal growth factor receptor, involved in the proliferation and mitogenesis of hepatocytes,25 inositol 1,4,5-triphosphate receptor, which has been found to be decreased in bile ducts of patients with primary biliary cirrhosis,26 and signal transducer and activator of transcription 2, which promotes antiviral defense in the liver.27 Finally, 10 members of the A disintegrin and metalloproteinase protein and A disintegrin-like and metalloproteinase with thrombospondin motifs families were identified. Members of these families are cell surface proteins with a unique structure possessing both potential adhesion and protease function. Proteolysis of the extracellular matrix plays a critical role in establishing tissue architecture during development and in tissue degradation in diseases such as cancer and a variety of inflammatory conditions. These families of proteases have been poorly studied in the liver.

In conclusion, this study is a comprehensive liver and plasma proteome profiling comparing these 2 proteomes. We have devised a method that allows for the simultaneous identification and quantitation of liver tissue and plasma proteins, many of them in the abundance range within which clinically relevant plasma biomarkers are believed to reside. As such, our study should allow for better integrated systems biology-based studies of mouse models of disease, which are relevant to developing an improved understanding of human hepatic processes and to discovering novel diagnostics and therapeutics for liver diseases.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

We thank Jason M. Hogan of the Fred Hutchinson Cancer Research Center Proteomics Facility for collecting the mass spectrometry data used in this work. Partial funding for purchase of the LTQ-FT mass spectrometer used in this work was generously provided by the M.J. Murdock Charitable Trust.

References

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Supplementary material for this article can be found on the H EPATOLOGY Web site ( http://interscience.wiley.com/jpages/0270-9139/suppmat/index.html ).

FilenameFormatSizeDescription
hep22123-SupplementaryTable1.pdf3613KSupplementary Table 1. Liver Tissue
hep22123-SupplementaryTable2.pdf2413KSupplementary Table 2. Plasma

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.