High-throughput quantitative profiling of serum N-glycome by MALDI-TOF mass spectrometry and N-glycomic fingerprint of liver fibrosis.
Glycomics studies at the whole-tissue level provide a general overview on the glycome, the total glycosylation pattern of glycoproteins, lipids, or other types of biomolecules. Some laboratories have attempted to release all glycans attached to whole serum proteins (the serum glycans/glycome) and compare the profiles of disease and control cases (11-14). Potential diagnostic serum glycan markers have been identified for liver cirrhosis (11, 12), ovarian cancer (13), and breast cancer (14).
Capillary electrophoresis has been applied to profile the serum N-glycome (11) and can quantify carbohydrate molecules with high reproducibility at a CV <5% (15). Compared to mass spectrometry (MS), however, capillary electrophoresis has relatively lower throughput. Recently, Fourier transform ion cyclotron resonance MS has been used (13, 14), but information about its quantification performance has not been available. The MALDI-TOF mass spectrometer, a relatively cheaper instrument, has been used for characterization of carbohydrate structure for many years (12, 16). In the past 5 years, MALDI-TOF MS-based instruments, such as SELDI ProteinChip system (17, 18) and the C1inProt system (19), have been used for quantitative profiling of the serum proteome.
The feasibility of using MALDI-TOF MS for quantitative profiling of the glycome remained uncertain. In this study, we attempted to establish a high-throughput quantitative assay for profiling the serum N-glycome by using the MALDI-TOF MS function and spectrum analysis programs of the SELDI ProteinChip system. We also used the new assay to search for serum N-glycans, the concentrations of which were altered in patients with liver fibrosis, to serve as an example of the clinical potential of the assay.
Materials and Methods
Liver biopsies were obtained from consecutive patients with chronic hepatitis B in the Prince of Wales Hospital, Hong Kong, in 2002 and 2003. Patients included those who had been recruited or screened for other therapeutic trials, as well as patients who were suspected of having active liver disease on the basis of laboratory or radiologic investigations. No patients were undergoing treatment at the time of liver biopsy or had compensated liver disease at the time of recruitment. After obtaining informed consent, we collected fasting blood samples by venipuncture before liver biopsy. Serum was stored at -70[degrees]C. Randomly selected serum samples from 46 patients [35 males and 11 females; mean (SD) age, 48.3 (9.2) years] were archived for this study. The same set of samples had been subjected to proteomic profiling in our previous study (18). Age was calculated to the date of biopsy. Use of these clinical samples for biomarker discovery was approved by the university ethics committee. The data of the histological staging (Ishak score reflecting the degrees of liver fibrosis) and hematologic and biochemical tests [complete blood screens, coagulation tests, bilirubin, total protein, albumin, alanine transaminase, alkaline phosphatase (ALP), and [alpha]-fetoprotein (AFP)] were retrieved from the database of our previous study (18). Ten patients had minimal fibrosis (Ishak score = 1), 9 had mild fibrosis (Ishak score = 2), 10 had moderate fibrosis (Ishak score = 3 or 4), 8 had severe fibrosis and incomplete cirrhosis (Ishak score = 5), and 9 had probable/ definite cirrhosis (Ishak score = 6).
DESIGN OF AN ASSAY FOR QUANTITATIVE PROFILING OF SERUM N-GLYCOME
N-glycans from whole serum proteins in 2 [micro]L serum were released by enzymatic digestion, cleaned up by hydrophilic chromatography, and then quantitatively profiled with a linear MALDI-TOF MS system, which was originally designed for quantitative proteomic profiling. The mass spectra were analyzed with the existing computer programs designed for quantitative proteomic profiling. Fig. 1 (see the Data Supplement that accompanies the online version of this article at http://www.clinchem.org/content/vo153/issue7) is the schematic diagram showing the design of the assay.
DETACHMENT OF N-GLYCANS FROM WHOLE SERUM PROTEINS
We denatured 2 [micro]L of serum at 56[degrees]C in 20 [micro]L of 12.5 mmol/L sodium phosphate buffer (pH 8.3) containing 2.5 mL/L [beta]-mercaptoethanol (Sigma-Aldrich) and 1.25 g/L sodium dodecyl sulfate (Sigma-Aldrich) for 1 h and then added 5 [micro]L of 100 mL/L Triton X-100 (Sigma-Aldrich) to neutralize the denaturing effect of the sodium dodecyl sulfate. We then added 25 [micro]L of 50 mmol/L sodium phosphate buffer (pH 8.3) containing 0.5 U of peptide:N-glycosidase F (PNGase F; EC 188.8.131.52; Sigma-Aldrich). The digestion mixture was incubated at 37[degrees]C for 24 h.
MICROSCALE PURIFICATION OF SERUM N-GLYCOME BY HYDROPHILIC CHROMATOGRAPHY
A method originally developed for purification of glycopeptides (20) was modified in this study to purify N-glycans. We equilibrated 10 mg of preswollen Sepharose 413 gel (GE Healthcare Life Sciences) twice with 0.5 mL of binding solution (butanol:ethanol:7 mmol/L manganese chloride solution = 5:1:1, vol/vol/vol). We diluted 8 [micro]L of the enzymatic digest with 200 [micro]L of binding solution, mixed this with the gel in a 0.5-mL microcentrifuge tube, and incubated the mixture on a blood-tube rotator at room temperature for 30 min. The whole content was then transferred to a Spin-X column (Corning) and washed 4 times with 0.5 mL of washing solution (butanol:ethanol: water = 5:1:1, vol/vol/vol). Finally the N-glycans were eluted with 200 [micro]L of 100 mL/L ethanol, followed by centrifugal vacuum drying. The N-glycans were redissolved with 5 [micro]L of 100 mL/L ethanol as the purified N-glycan preparation and stored at -30[degrees]C before MS analysis.
MALDI-TOF MS-BASED N-GLYCOME PROFILING ASSAY
N-glycan samples and matrix chemical were added on an 8-spot gold ProteinChip array (Ciphergen Biosystems) by undertaking a sandwich spotting approach. Super-2,5-dihydroxybenzoic acid (DHB) in the presence of NaCl was used as the matrix chemical. A stock solution of super-DHB (1.35 g/L DHB, 0.15 g/L 2-hydroxy-5-methoxybenzoic acid, 10 mmol/L NaCl in 100 mL/L ethanol) was prepared, stored at -30[degrees]C, and 6.7-fold diluted with MilliQ water on the day of MS analysis. First, 0.5 [micro]L of super-DHB was spotted and vacuum-dried. Then 0.5 [micro]L of the purified N-glycan preparation was spotted and vacuum-dried, followed by sequential addition and vacuum drying of 0.5 [micro]L of 500 mL/L acetonitrile and 0.5 [micro]L of super-DHB. Finally, 0.5 [micro]L of 500 mL/L ethanol was spotted over the spotting area and vacuum-dried to achieve a homogeneous layer of analyte-matrix cocrystals. The gold ProteinChip arrays were read over the 900-4000 m/z interval on the ProteinChip PBS II reader of a ProteinChip Biomarker System. The ProteinChip reader was equipped with a nitrogen laser operating at 337 run wavelength and at 5 microjoule pulse. The mass spectra were acquired at a positive ion mode and an optimized interval of 1200-3000 m/z. Mass spectrum of each glycan sample was a mean of 610 mass spectra obtained over the sample spot at 61 acquisition positions. The mass spectrum of each sample was smoothed, baseline-subtracted, externally calibrated with the standard N-glycans (see Table 1 in the online Data Supplement), and finally internally calibrated with peaks at 1626.5, 2269.0, and 2947.6 m/z. The glycan peaks with signal-to-noise ratio >3 among the mass spectra were identified and quantified by use of Biomarker Wizard software (Ciphergen Biosystems). The peak intensities were normalized with the total ion current and subsequently with the total peak intensity. On the mass spectra, each nonsialylated glycan appeared as a single peak with an m/z value equivalent to the average molecular mass plus 1.0 mass unit (mass of a proton). For the sialylated glycans, sodium salts were formed and detected as a group of consecutive peaks with a step-wise increment of 22 m/z (see Fig. 2 in the online Data Supplement). For a sialylated glycan with n sialic acid residues, n + 1 consecutive peaks were detected. The normalized peak intensities of all consecutive peaks were summed to obtain the normalized peak intensity for the sodium-free parent glycan. Each sample was analyzed in duplicate. Mean values of the duplicates were used in subsequent data mining.
EVALUATION OF QUANTITATIVE PERFORMANCE OF THE MALDI-TOF MS-BASED N-GLYCOME PROFILING ASSAY
Four standard N-glycans (CalBiochem) were used during evaluation. Structural information of the standard N-glycans is listed in Table 1 in the online Data Supplement. When the reproducibility of the normalized peak intensities was examined, 3 (M3N2, MAN8 and A2) of the standard N-glycans was fixed at 10 pmol. Intraassay CV was determined from 32 identical sample-matrix spots in one single day by use of the same preparations of matrix chemical and reagents. For the interassay CV, the standard glycan mixture was measured in duplicate for 12 consecutive days by using different batches of matrix chemical and reagents, which were prepared freshly on the day of experiment. Linearity relationships between the quantity and normalized peak intensity of each of 4 standard N-glycans were evaluated by varying amounts from 5 to 20 pmol while fixing the amount of the other 3 N-glycans at 10 pmol.
IDENTIFICATION OF POTENTIAL DIAGNOSTIC GLYCANS FOR LIVER FIBROSIS/ CIRRHOSIS
As in our previous study (18), we used 2 criteria to identify the glycans associated only with disease, (a) the normalized peak intensities must be significantly higher/ lower in patients with typical fibrosis/ cirrhosis than in individuals with minimal fibrosis, and (b) the normalized peak intensities must correlate with the degree of fibrosis. The significance analysis of microarray algorithm (Stanford University) (17, 18, 21) was used to identify the glycans that were significantly higher/lower in patients with fibrosis/ cirrhosis by comparing the N-glycome profiles of the cases with minimal fibrosis (Ishak score = 1) with those for cases with typical fibrosis/ cirrhosis (Ishak score >3) at a median false discovery rate of 2.5%. In the significance analysis of microarray analysis, "2-classed, unpaired data" were selected as the data type, and 5000 permutations were performed. Correlations between the degree of fibrosis (Ishak scores) and the peak intensities of the significant glycans were analyzed by the Spearman rank order correlation test.
CONSTRUCTION OF GLYCAN-BASED DIAGNOSTIC MODEL
Log, values of the age, potential diagnostic glycans, and serological markers [total protein, albumin concentration, international normalized ratio, activated partial thromboplastin time (APTT), alanine transaminase, total bilirubin, ALP, AFP, hemoglobin, leukocyte count, and platelet count] were included as independent variables, and Ishak score was set as the dependent output variable. Forward stepwise linear regression analysis was performed using SPSS (SPSS Inc.) to select the variables with independent prediction values for constructing a diagnostic model to calculate the Fibro-Glyco index. Leave-one-out cross-validation was performed to evaluate the performance of the diagnostic model in detecting significant liver fibrosis (Ishak score [greater than or equal to]3) and cirrhosis (Ishak score as we previously described (18).
The Spearman rank order correlation test was used to examine correlation relationships between the N-glycan peak intensities and the clinical /biochemical variables. The sensitivity and specificity were calculated according to the standard formulas. ROC curves were constructed for differentiating patients with or without significant liver fibrosis (Ishak score [greater than or equal to]3) and for differentiating patients with or without liver cirrhosis (Ishak score [greater than or equal to]5). The likelihood ratios (LRs) were calculated by use of the standard formulas: positive LR (LR+) = sensitivity/ (100% - specificity); negative LR (LR-) = (100% sensitivity)/specificity.
QUANTITATIVE PERFORMANCE OF THE MALDI-TOF MS-BASED N-GLYCOME PROFILING ASSAY
A representative mass spectrum of 4 standard N-glycans all at 10 pmol is shown in supplementary Fig. 3. For all 3 standard N-glycans, the Intraassay CVs of the normalized peak intensities were <8% and the interassay CVs were <17% (Table 1). Linearity study showed that the normalized peak intensity was directly proportional to the relative quantity of a glycan, with an [R.sup.2] value [greater than or equal to]0.95 for all 4 standard glycans (see Fig. 4 in the online Data Supplement).
SERUM N-GLYCOME PROFILE AND LIVER FIBROSIS-ASSOCIATED FINGERPRINT
The representative serum N-glycome mass spectra of serum samples from a healthy individual, a patient with minimal liver fibrosis, and a patient with significant fibrosis are shown in Fig. 1, A-C. Fig. 1D shows that when heat-inactivated PNGase F was used to release the N-glycans, no significant peaks were detected. This finding indicates that the majority of the peaks in the mass spectra were contributed by the N-glycans released from the whole serum proteins. We observed that 5 freeze-thaw cycles did not significantly affect the serum N-glycome profiles (see Fig. 5 in the online Data Supplement).
We found that 55 glycans were matched among the serum samples. At a median false-discovery rate of 2.5%, we identified 15 N-glycans that were significantly higher and 8 N-glycans that were significantly lower in the patient group of typical fibrosis/ cirrhosis. Among these 23 significant N-glycans, 13 were positively correlated with the degree of fibrosis, and 8 were negatively correlated (all P values <0.05; Fig. 2). Therefore, only 21 glycans were significantly associated with liver fibrosis. ROC curve analyses showed that 17 of these liver fibrosis-associated glycans had potential value in the detection of either liver fibrosis and/or liver cirrhosis (all P values <0.05). The 1829.7 m/z glycan [area under the curve (AUC) = 0.90] was the best marker for detecting significant liver fibrosis, and the 1444.7 m/z glycan (AUC = 0.83) was the best marker for detecting liver cirrhosis. The results for the 17 potential diagnostic glycans are listed in Table 2.
CORRELATIONS BETWEEN THE N-GLYCAN PEAK INTENSITIES AND SEROLOGICAL MARKERS
Among the 17 potential diagnostic glycans, 5 were correlated with albumin concentration (r = 0.417 to 0.490 and -0.292 to -0.348; all P values <0.05); 5 were correlated with total protein (r = 0.329 to 0.351 and -0.292 to -0.343; all P values <0.05);1 was positively correlated with APTT (r = 0.326; P = 0.027), and 3 were negatively correlated (r = -0.322 to -0.356; all P values <0.05); 4 were correlated with international normalized ratio (r = 0.304 to 0.403 and -0.380 to -0.450; all P values <0.05), resulting in 10 glycans significantly correlated with liver function. Concerning liver inflammation, only the 2606.9 m/z glycan was negatively correlated with alanine transaminase concentration (r = -0.366; P = 0.012). Concerning liver damage, 1 glycan was positively correlated with total bilirubin (r = 0.319; P = 0.031), whereas 1 glycan was negatively correlated (r = -0.316; P = 0.032). None of the glycans were correlated with ALP. Three glycans were positively correlated with AFP (r = 0.327 to 0.431; all P values <0.05). As a result, 76% (13 of 17) of the liver-fibrosis-associated glycans potentially reflected the physiological state of the liver. Finally, it is well known that platelet count is negatively correlated with degrees of fibrosis. We found that 8 glycans were correlated with the platelet count (all P values <0.05). Therefore, 82% (14 of 17) of the glycans forming the liver fibrosis fingerprint were significantly correlated with the serologic markers that directly reflect the liver physiology and/or with those that are well known to be altered in patients with liver fibrosis (Table 2).
LINEAR REGRESSION MODEL FOR DETECTING LIVER FIBROSIS AND CIRRHOSIS
Among all the dependent variables (potential diagnostic glycans, serological markers, and age), only N-glycans of 1341.5, 1829.7, 1933.3, or 2130.3 m/z (all P <0.005) were selected as biomarkers with independent diagnostic values and included in the linear regression model for calculating the Fibro-Glyco index. Leave-one-out cross-validation showed that the Fibro-Glyco index significantly correlated with degrees of liver fibrosis (r = 0.784; P = 0.01). Furthermore, the index was useful in the detection of liver fibrosis and cirrhosis. The ROC curve areas for the Fibro-Glyco index in detection of liver fibrosis and in detection of liver cirrhosis both were 0.91 (Fig. 3). At 84% specificity, the sensitivity for detecting liver fibrosis was 85%, the overall accuracy was 85%, and the LR+ and LR- were 5.4 and 0.18, respectively. At 83% specificity, the sensitivity for detecting liver cirrhosis was 88%, the overall accuracy was 85%, and the LR+ and LR- were 5.1 and 0.14, respectively.
Although MALDI-TOF MS has been widely used to characterize the structure of glycans, this is first study indicating that MALDI-TOF MS is also applicable to quantitative profiling of a complex mixture of glycans. To achieve this quantitative capability, we took 2 important measures, one to ensure a homogeneous layer of glycan-matrix cocrystals and the other to scan the glycan-matrix spot to obtain 610 mass spectra, the mean of which was used as the final mass spectrum for each glycan sample. Previously, we showed that both measures were important to obtain a quantitative mass spectrum for profiling a mixture of protein/peptide molecules (22).
In the present study, using the serum N-glycome profiling assay, we successfully identified a panel of serum N-glycans as potential markers for detecting liver fibrosis and liver cirrhosis; 82% (14 of 17) of the glycans with potential diagnostic values were significantly correlated with the serological markers that directly reflect liver physiology and/or with those that are well known to be altered in patients with liver fibrosis. These findings also serve as indirect evidence indicating that identified diagnostic glycans have clinical meanings associated with the disease being studied.
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
Callewaert et al. (11) applied the DNA sequencer to profile sialidase-treated N-glycans from the whole serum proteins in patients with and without liver cirrhosis. The present study strongly suggested that the serum N-glycome profile is useful in predicting not only liver cirrhosis but also liver fibrosis. As far as we know, this is the first study demonstrating such clinical importance in humans. Because only 46 patients were examined in this study, we are undertaking a similar study with a much larger sample size to confirm the clinical value of the present diagnostic model. Further studies are also needed to determine whether the same serum N-glycan-based model can be used to detect liver fibrosis/ cirrhosis with other underlying causes, such as chronic hepatitis C infection and chronic alcohol abuse.
N-glycans are different from peptides or proteins in terms of sequence-structure analysis. Because human N-linked glycans all contain the same chitobiose core [2 N-acetyl-glucosamines (G1cNAcs) attached to a triman-nose] and are synthesized by a well-defined pathway, an accurate mass measurement will generally permit prediction of structure (23, 24). We attempted to predict the structures of the 4 major diagnostic glycans (1341.5, 1829.7, 1933.3, and 2130.3 m/z) from their mean m/z values by searching the Consortium for Functional Glycomics database and using Expasy GlycoMod (23) with an acceptable error tolerance [less than or equal to]0.05%. The predicted structures of the 4 glycans were successfully matched to those that had been observed in human (Fig. 4).
With the predicted structures, we found that 3 diagnostic N-glycans (1341.5, 1829.7, and 2130.3 m/z) showing positive correlations with fibrosis stages contain a proximal fucose and/or a bisecting G1cNAc. Because liver cirrhosis is the severe stage of liver fibrosis, our prediction was consistent with the observations that degrees of fucosylation on certain serum glycoproteins (25-27) were increased in liver cirrhosis. Recently Morelle et al. (12) also observed that certain serum N-glycan species containing a proximal fucose and/or a bisecting G1cNAc were increased in liver cirrhosis. One of these (1829.7 m/z) was found in the present study. Callewaert et al. (11) also found that a biantennary N-glycan with a proximal fucose and a bisecting G1cNAc was increased in liver cirrhosis and was a major diagnostic marker of liver cirrhosis. Because the m/z values allow us only to predict the structures of the glycans, further experiments are needed to confirm the predicted structures, such as tandem MS analysis (12) and glycosidase sequencing (28).
[FIGURE 3 OMITTED]
[FIGURE 4 OMITTED]
Our serum N-glycome profiling assay has several advantages. The assay can easily be performed by laboratories already equipped with the SELDI ProteinChip system, and only 2 [micro]L of serum is needed for profiling the serum N-glycome. The assay is high-throughput, so a maximum of 96 serum samples can be examined in duplicate within 2 working days. As with the serum proteomic profiling assay, it is not necessary to know the structure of a diagnostic glycan, because a diagnostic N-glycan can be reproducibly recognized and quantified through its unique m/z value. When a MALDI TOF/TOF MS or MALDI Q-TOF MS instrument is used instead of the MALDI TOF MS instrument, the structures of the glycans can be further obtained by tandem MS analysis.
In addition to liver fibrosis, our assay has the potential to be applied to the diagnoses of other diseases, such as CDG. Multiple serum glycoproteins with abnormal glycosylation have been observed in patients with CDG (29, 30). Furthermore, our assay can be modified for other applications, such as quantitative profiling of glycans on a single protein. For example, quantitative profiling of the N-glycans detached from transferrin should be useful in detecting alcohol abuse. Quantitative analysis of specific N-glycans on haptoglobin may be useful in the diagnosis of liver cancer (31).
In conclusion, we have developed a high-throughput assay for quantitative profiling of serum N-glycome using a system originally designed for serum proteomic profiling. Using diagnosis of liver fibrosis as an example, we have demonstrated the potential of applying our assay to biomarker discovery.
Grant/funding support: The project team was supported by the Research Fund for the Control of Infectious Diseases from the Health, Welfare and Food Bureau of the Hong Kong SAR Government and the Li Ka Shing Foundation. Financial disclosures: None declared.
Received January 9, 2007; accepted April 26, 2007. Previously published online at DOI: 10.1373/clinchem.2007.085563
(1.) Raman R, Raguram S, Venkataraman G, Paulson JC, Sasisekharan R. Glycomics: an integrated systems approach to structure-function relationships of glycans. Nat. Methods 2005;2:817-24.
(2.) Freeze HH, Aebi M. Altered glycan structures: the molecular basis of congenital disorders of glycosylation. Curr Opin Struct Biol 2005;15:490-8.
(3.) Seta N, Barnier A, Hochedez F, Besnard MA, Durand G. Diagnostic value of Western blotting in carbohydrate-deficient glycoprotein syndrome. Clin Chim Acta 1996;254:131-40.
(4.) Flahaut C, Michalski JC, Danel T, Humbert MH, Klein A. The effects of ethanol on the glycosylation of human transferring. Glycobiology 2003;13:191-8.
(5.) Shimizu K, Katoh H, Yamashita F, Tanaka M, Tanikawa K, Taketa K, et al. Comparison of carbohydrate structures of serum [alpha]-fetoprotein by sequential glycosidase digestion and lectin affinity electrophoresis. Clin Chim Acta 1996;254:23-40.
(6.) Johnson PJ, Poon TC, Hjelm NM, Ho CS, Blake C, Ho SK. Structures of disease-specific serum alpha-fetoprotein isoforms. Br J Cancer 2000;83:1330-7.
(7.) Helander A, Bergstrom J, Freeze HH. Testing for congenital disorders of glycosylation by HPLC measurement of serum transferrin glycoforms. Clin Chem 2004;50:954-8.
(8.) Legros FJ, Nuyens V, Minet E, Emonts P, Boudjeltia KZ, Courbe A, et al. Carbohydrate-deficient transferrin isoforms measured by capillary zone electrophoresis for detection of alcohol abuse. Clin Chem 2002;48:2177-86.
(9.) Kumada T, Nakano S, Takeda I, Kiriyama S, Sone Y, Hayashi K, et al. Clinical utility of Lens culinaris agglutinin-reactive alpha-fetoprotein in small hepatocellular carcinoma: special reference to imaging diagnosis. J Hepatol 1999;30:125-30.
(10.) Poon TCW, Mok TSK, Chan ATC, Chan CML, Leong V, Tsui SHT, et al. Quantification and utility of monosialylated alpha-fetoprotein in the diagnosis of hepatocellular carcinoma with nondiagnostic serum total alpha-fetoprotein. Clin Chem 2002;48:1021-7.
(11.) Callewaert N, Van Vlierberghe H, Van Hecke A, Laroy W, Delanghe J, Contreras R. Noninvasive diagnosis of liver cirrhosis using DNA sequencer-based total serum protein glycomics. Nat Med 2004; 10:429-34.
(12.) Morelle W, Flahaut C, Michalski JC, Louvet A, Mathurin P, Klein A. Mass spectrometric approach for screening modifications of total serum N-glycome in human diseases: application to cirrhosis. Glycobiology 2006;16:281-93.
(13.) An HJ, Miyamoto S, Lancaster KS, Kirmiz C, Li B, Lam KS, et al. Profiling of glycans in serum for the discovery of potential biomarkers for ovarian cancer. J Proteome Res 2006;5:1626-35.
(14.) Kirmiz C, Li B, An HJ, Clowers BH, Chew HK, Lam KS, et al. A serum glycomics approach to breast cancer biomarkers. Mol Cell Proteomics 2007;6:43-55.
(15.) Lee YH, Lin TI. Determination of carbohydrates by high-performance capillary electrophoresis with indirect absorbance detection. J Chromatogr B Biomed Appl 1996;681:87-97.
(16.) Harvey DJ. Matrix-assisted laser desorption/ionization mass spectrometry of carbohydrates and glycoconjugates. Int J Mass Spectrom 2003;226:1-35.
(17.) Poon TCW, Yip TT, Chan ATC, Yip C, Yip V, Mok TSK, et al. Comprehensive proteomic profiling identifies serum proteomic signatures for detection of hepatocellular carcinoma and its subtypes. Clin Chem 2003;49:752-60.
(18.) Poon TCW, Hui AY, Chan HLY, Ang IL, Chow SM, Wong N, et al. Prediction of liver fibrosis and cirrhosis in chronic hepatitis B infection by serum proteomic fingerprinting: a pilot study. Clin Chem 2005;51:328-35.
(19.) Cheng AJ, Chen LC, Chien KY, Chen YJ, Cheng JT, Wang HM, et al. Oral cancer plasma tumor marker identified with bead-based affinity-fractionated proteomic technology. Clin Chem 2005;51: 2236-44.
(20.) Wada Y, Tajiri M, Yoshida S. Hydrophilic affinity isolation and MALDI multi-stage tandem mass spectrometry of glycopeptides for glycoproteomics. Anal Chem 2004;76:6560-5.
(21.) Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 2001;98:5116-21.
(22.) Pang RTK, Johnson PJ, Chan CIVIL, Kong EKC, Chan ATC, Sung JJY, et al. Technical evaluation of MALDI-TOF mass spectrometry for quantitative proteomic profiling: matrix formulation and application. Clin Proteomics 2004;1:259-70.
(23.) Cooper CA, Gasteiger E, Packer NH. GlycoMod: a software tool for determining glycosylation compositions from mass spectrometric data. Proteomics 2001;1:340-9.
(24.) Mills PB, Mills K, Mian N, Winchester BG, Clayton PT. Mass spectrometric analysis of glycans in elucidating the pathogenesis of CDG type IN. J Inherit Metab Dis 2003;26:119-34.
(25.) Mann AC, Record CO, Self CH, Turner GA. Monosaccharide composition of haptoglobin in liver diseases and alcohol abuse: large changes in glycosylation associated with alcoholic liver disease. Clin Chim Acta 1994;227:69-78.
(26.) Kondo M, Hada T, Fukui K, Iwasaki A, Higashino K, Yasukawa K. Enzyme-linked immunosorbent assay (ELISA) for Aleuria aurantia lectin-reactive serum cholinesterase to differentiate liver cirrhosis and chronic hepatitis. Clin Chim Acta 1995;243:1-9.
(27.) Ryden I, Pahlsson P, Lindgren S. Diagnostic accuracy of alpha(1)acid glycoprotein fucosylation for liver cirrhosis in patients undergoing hepatic biopsy. Clin Chem 2002;48:2195-201.
(28.) Johnson PJ, Poon TCW, Hjelm NM, Ho CS, Ho SKW, Welby C, et al. Glycan composition of serum alpha-fetoprotein in patients with hepatocellular carcinoma and non-seminomatous germ cell tumour. Br J Cancer 1999;81:1188-95.
(29.) Harrison HH, Miller KL, Harbison MD, Slonim AE. Multiple serum protein abnormalities in carbohydrate-deficient glycoprotein syndrome: pathognomonic finding of two-dimensional electrophoresis? Clin Chem 1992;38:1390-2.
(30.) Callewaert N, Schollen E, Vanhecke A, Jaeken J, Matthijs G, Contreras R. Increased fucosylation and reduced branching of serum glycoprotein N-glycans in all known subtypes of congenital disorder of glycosylation I. Glycobiology 2003;13:367-75.
(31.) Ang IL, Poon TCW, Lai PBS, Chan ATC, Ngai SM, Hui AY, et al. Study of serum haptoglobin and its glycoforms in the diagnosis of hepatocellular carcinoma: a glycoproteomic approach. J Proteome Res 2006;5:2691-700.
 Nonstandard abbreviations: CDC, congenital disorders in glycosylation; MS, mass spectrometry; ALP, alkaline phosphatase; AFP, [alpha]-fetoprotein; DHB, 2,5-dihydroxybenzoic acid; LR, likelihood ratio; AUC, area under the curve; G1cNAc, N-acetyl-glucosamine; APTT, activated partial thromboplastin time.
RICHARD K.T. KAM, [1,2] TERENCE C.W. POON, [1,2,3] * HENRY L.Y. CHAN, [2,3] NATHALIE WONG,  ALEX Y. HUI,  JOSEPH J.Y. SUNG [2,3]
 Li Ka Shing Institute of Health Sciences,
 Department of Medicine and Therapeutics,
 Centre for Emerging Infectious Diseases, and
 Department of Anatomical and Cellular Pathology, The Chinese University of Hong Kong, Prince of Wales Hospital, Shatin, Hong Kong SAR.
* Address correspondence to this author at: Li Ka Shing Institute of Health Sciences, Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Prince of Wales Hospital, Shatin, N.T., Hong Kong SAR. Fax 852-2648-8842; e-mail email@example.com.
Table 1. Intra- and interassay reproducibility of the MALDI-TOF MS-based N-glycome profiling assay. Standard N-glycan Intraassay CV, % Interassay CV, % M3N2 8.67 17.11 MAN8 8.67 10.12 A2 8.19 16.59 Table 2. Summary of the 17 serum N-glycans identified that are potential markers for liver fibrosis/cirrhosis. Correlation between peak N-glycan, mean m/z intensity and Ishak score (SD; minimum-maximum) (P value) 1341.4 (0.1; 1341.1-1341.8) 0.447 (b) (0.002) 1363.8 (0.0; 1363.5-1364.0) 0.441 (0.002) 1444.7 (0.1; 1443.3-1445.2) -0.560 ( 0.001) 1465.5 (0.6; 1464.3-1466.7) -0.370 (<0.011) 1488.6 (0.0; 1488.4-1488.8) -0.483 (0.001) 1829.7 (0.0; 1829.6-1829.8) 0.517 (<0.001) 1933.3 (0.3; 1932.1-1934.0) -0.478 (0.001) 1992.0 (0.0; 1991.8-1992.1) 0.327 (0.027) 2130.3 (0.2; 2129.8-2131.0) 0.380 (0.009) 2159.0 (0.2; 2158.3-2159.6) 0.355 (0.016) 2284.6 (0.2; 2284.1-2285.1) 0.292 (0.049) 2431.0 (0.1; 2430.4-2431.6) 0.328 (0.026) 2606.9 (0.2; 2604.0-2607.1) -0.310 (0.036) 2780.4 (0.2; 2779.6-2781.3) 0.383 (0.009) 2881.4 (0.2; 2880.6-2882.3) -0.437 (0.002) 3027.6 (0.3; 3026.7-3028.6) 0.357 (0.015) 3285.0 (0.2; 3283.8-3286.1) -0.357 (0.015) Correlation between peak N-glycan, mean m/z intensity and serological (SD; minimum-maximum) markers (a) 1341.4 (0.1; 1341.1-1341.8) LD+ 1363.8 (0.0; 1363.5-1364.0) LF-, WBC- 1444.7 (0.1; 1443.3-1445.2) LF+, LD-, PLC+ 1465.5 (0.6; 1464.3-1466.7) PLC+, WBC+ 1488.6 (0.0; 1488.4-1488.8) LF+, PLC+ 1829.7 (0.0; 1829.6-1829.8) LF-, PLC- 1933.3 (0.3; 1932.1-1934.0) LF+, PLC+ 1992.0 (0.0; 1991.8-1992.1) LF-, PLC- 2130.3 (0.2; 2129.8-2131.0) NS 2159.0 (0.2; 2158.3-2159.6) AFP+ 2284.6 (0.2; 2284.1-2285.1) NS 2431.0 (0.1; 2430.4-2431.6) NS 2606.9 (0.2; 2604.0-2607.1) LI- 2780.4 (0.2; 2779.6-2781.3) LF-, AFP+ 2881.4 (0.2; 2880.6-2882.3) LF+, PLC+ 3027.6 (0.3; 3026.7-3028.6) LF-, AFP+ 3285.0 (0.2; 3283.8-3286.1) LF+, PLC+ Mean (SD) peak intensity of N-glycans N-glycan, mean m/z Patients with Ishak (SD; minimum-maximum) score <3 1341.4 (0.1; 1341.1-1341.8) 0.228 (0.365) 1363.8 (0.0; 1363.5-1364.0) 0.266 (0.255) 1444.7 (0.1; 1443.3-1445.2) 2.698 (1.741) 1465.5 (0.6; 1464.3-1466.7) 2.550 (1.136) 1488.6 (0.0; 1488.4-1488.8) 0.942 (0.678) 1829.7 (0.0; 1829.6-1829.8) 1.021 (0.193) 1933.3 (0.3; 1932.1-1934.0) 10.827 (1.099) 1992.0 (0.0; 1991.8-1992.1) 1.265 (0.197) 2130.3 (0.2; 2129.8-2131.0) 1.173 (0.162) 2159.0 (0.2; 2158.3-2159.6) 0.635 (0.124) 2284.6 (0.2; 2284.1-2285.1) 4.905 (0.821) 2431.0 (0.1; 2430.4-2431.6) 1.369 (0.640) 2606.9 (0.2; 2604.0-2607.1) 1.502 (0.254) 2780.4 (0.2; 2779.6-2781.3) 0.650 (0.152) 2881.4 (0.2; 2880.6-2882.3) 5.871 (1.295) 3027.6 (0.3; 3026.7-3028.6) 1.685 (0.819) 3285.0 (0.2; 3283.8-3286.1) 0.396 (0.123) Mean (SD) peak intensity of N-glycans Patients with Ishak N-glycan, mean m/z score [greater than (SD; minimum-maximum) or equal to]3 1341.4 (0.1; 1341.1-1341.8) 0.440 (0.752) 1363.8 (0.0; 1363.5-1364.0) 0.405 (0.410) 1444.7 (0.1; 1443.3-1445.2) 1.067 (0.923) 1465.5 (0.6; 1464.3-1466.7) 1.944 (0.912) 1488.6 (0.0; 1488.4-1488.8) 0.354 (0.312) 1829.7 (0.0; 1829.6-1829.8) 1.478 (0.425) 1933.3 (0.3; 1932.1-1934.0) 9.968 (1.158) 1992.0 (0.0; 1991.8-1992.1) 1.507 (0.311) 2130.3 (0.2; 2129.8-2131.0) 1.283 (0.235) 2159.0 (0.2; 2158.3-2159.6) 0.852 (0.244) 2284.6 (0.2; 2284.1-2285.1) 5.376 (0.889) 2431.0 (0.1; 2430.4-2431.6) 1.463 (0.693) 2606.9 (0.2; 2604.0-2607.1) 1.399 (0.284) 2780.4 (0.2; 2779.6-2781.3) 0.753 (0.263) 2881.4 (0.2; 2880.6-2882.3) 4.941 (1.479) 3027.6 (0.3; 3026.7-3028.6) 2.068 (1.017) 3285.0 (0.2; 3283.8-3286.1) 0.318 (0.106) ROC curve area (95% CI) N-glycan, mean m/z (SD; minimum-maximum) Detection of fibrosis 1341.4 (0.1; 1341.1-1341.8) 0.684 (0.515-0.853) 1363.8 (0.0; 1363.5-1364.0) 0.690 (0.522-0.858) 1444.7 (0.1; 1443.3-1445.2) 0.795 (0.667-0.924) (c) 1465.5 (0.6; 1464.3-1466.7) NS (d) 1488.6 (0.0; 1488.4-1488.8) 0.776 (0.633-0.918) (c) 1829.7 (0.0; 1829.6-1829.8) 0.897 (0.802-0.991) 1933.3 (0.3; 1932.1-1934.0) 0.713 (0.563-0.864) (c) 1992.0 (0.0; 1991.8-1992.1) 0.745 (0.595-0.895) 2130.3 (0.2; 2129.8-2131.0) NS 2159.0 (0.2; 2158.3-2159.6) 0.778 (0.645-0.911) 2284.6 (0.2; 2284.1-2285.1) 0.702 (0.545-0.858) 2431.0 (0.1; 2430.4-2431.6) NS 2606.9 (0.2; 2604.0-2607.1) 0.766 (0.625-0.907) (c) 2780.4 (0.2; 2779.6-2781.3) NS 2881.4 (0.2; 2880.6-2882.3) 0.692 (0.540-0.844) (c) 3027.6 (0.3; 3026.7-3028.6) NS 3285.0 (0.2; 3283.8-3286.1) 0.688 (0.531-0.845) (c) ROC curve area (95% CI) N-glycan, mean m/z (SD; minimum-maximum) Detection of cirrhosis 1341.4 (0.1; 1341.1-1341.8) 0.696 (0.534-0.857) 1363.8 (0.0; 1363.5-1364.0) 0.696 (0.536-0.856) 1444.7 (0.1; 1443.3-1445.2) 0.834 (0.719-0.948) (c) 1465.5 (0.6; 1464.3-1466.7) 0.740 (0.595-0.885) (c) 1488.6 (0.0; 1488.4-1488.8) 0.785 (0.654-0.916) (c) 1829.7 (0.0; 1829.6-1829.8) 0.702 (0.552-0.852) 1933.3 (0.3; 1932.1-1934.0) 0.740 (0.593-0.888) (c) 1992.0 (0.0; 1991.8-1992.1) NS 2130.3 (0.2; 2129.8-2131.0) 0.719 (0.640-0.942) 2159.0 (0.2; 2158.3-2159.6) NS 2284.6 (0.2; 2284.1-2285.1) NS 2431.0 (0.1; 2430.4-2431.6) 0.677 (0.515-0.840) 2606.9 (0.2; 2604.0-2607.1) NS 2780.4 (0.2; 2779.6-2781.3) 0.777 (0.636-0.918) 2881.4 (0.2; 2880.6-2882.3) 0.753 (0.607-0.898) (c) 3027.6 (0.3; 3026.7-3028.6) 0.789 (0.656-0.922) 3285.0 (0.2; 3283.8-3286.1) NS (a) LF+, peak intensity positively correlated with liver function (positively correlated with Alb concentration and/or total protein concentration; negatively with INR and/or APTT). LF-, peak intensity negatively correlated with liver function (negatively correlated with Alb concentration and/or total protein concentration; positively with INR and/or APTT). LI-, peak intensity negatively correlated with liver inflammation (negatively correlated with ALT concentration). LD , peak intensity positively correlated with liver damage (positively correlated with total bilirubin concentration). LD+, peak intensity negatively correlated with liver damage (negatively correlated with total bilirubin concentration). AFP+, peak intensity positively correlated with AFP concentration. PLC+, peak intensity positively correlated with platelet count. PLC-, peak intensity negatively correlated with platelet count. WBC+, peak intensity positively correlated with white blood cell count. WBC-, peak intensity negatively correlated with white blood cell count. (b) Correlation coefficient calculated by the Spearman rank order correlation test. (c) ROC curve area was calculated by using 1/normalized peak intensity. (d) NS, Not statistically significant, i.e., P >0.05 (two tails).
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Proteomics and Protein Markers|
|Author:||Kam, Richard K.T.; Poon, Terence C.W.; Chan, Henry L.Y.; Wong, Nathalie; Hui, Alex Y.; Sung, Joseph|
|Date:||Jul 1, 2007|
|Previous Article:||Profiling the antibody immune response against blood stage malaria vaccine candidates.|
|Next Article:||B-type natriuretic peptide concentrations predict the progression of nondiabetic chronic kidney disease: The mild-to-moderate kidney disease study.|