Strategy for Glycoproteomics: Identification of Glyco-Alteration Using

Jan 29, 2009 - Hiromi Ito,‡,# Atsushi Kuno,‡,# Hiromichi Sawaki,‡,# Maki Sogabe,‡ Hidenori Ozaki,‡. Yasuhito Tanaka,§ Masashi Mizokami,§ J...
0 downloads 0 Views 1MB Size
Download PDF
Strategy for Glycoproteomics: Identification of Glyco-Alteration Using Multiple Glycan Profiling Tools† Hiromi Ito,‡,# Atsushi Kuno,‡,# Hiromichi Sawaki,‡,# Maki Sogabe,‡ Hidenori Ozaki,‡ Yasuhito Tanaka,§ Masashi Mizokami,§ Jun-ichi Shoda,⊥ Takashi Angata,‡ Takashi Sato,‡ Jun Hirabayashi,‡ Yuzuru Ikehara,‡ and Hisashi Narimatsu*,‡ Research Center for Medical Glycoscience, National Institute of Advanced Industrial Science and Technology (AIST), 1-1-1 Umezono, Tsukuba, Ibaraki 305-8568, Japan, Department of Clinical Molecular Informative Medicine, Nagoya City University Graduate School of Medical Sciences, Kawasumi, Mizuho, Nagoya 467-8601, Japan, and Department of Gastroenterology, Graduate School of Comprehensive Human Sciences, University of Tsukuba, Tsukuba, Ibaraki 305-8577, Japan Received September 7, 2008

Glycan alterations of proteins, a common feature of cancer cells, are associated with carcinogenesis, invasion and metastasis. Glycomics, the study of glycans and glycan-binding proteins in various biological systems, is an emerging field in the postgenome and postproteomics era. However, systematic and robust strategies for glycomics are still not fully established because the structural analysis of glycans, which comprise different patterns of branching, various possible linkage positions as well as monomer anomericity, is technically difficult. Here, we introduce a new strategy for glyco-alteration analysis of glycoproteins by using multiple glycan profiling tools. To understand glycan alterations of proteins by correlating the glycosyltransferase expression profile with the actual glycan structure, we systematically used three glycan profiling tools: (1) multiplex quantitative PCR (qPCR) array format for profiling the expression pattern of glycogenes, (2) lectin microarray as a multiplex glycan-lectin interaction analysis system for profiling either a pool of cell glycoproteins or a target glycoprotein, and (3) tandem mass spectrometry for identifying the glycan structure connected to a target glycoprotein. Using our system, we successfully identified glycan alterations on alpha-fetoprotein (AFP), including a novel LacdiNAc structure in addition to previously reported alterations such as R1,6 fucosylation. Keywords: glycoproteomics • glycomics • multiplex quantitative PCR array • lectin microarray • tandem mass spectrometry • glycoproteins

Introduction Glycosylation plays an important role in regulating properties of proteins or lipids on the cell surface. Glycan interactions mediate biological events such as trafficking, signaling, folding and adhesion, in addition to processes such as development, immunity and diseases.1-5 Glycan biosynthetic pathways involve many glycan related enzymes encoded by glycogenes, including glycosyltransferases, sulfotransferases, which add a sulfate moiety to the glycan, and sugar-nucleotide transporters.6,7 The complex interplay of these enzyme systems leads to a rich diversity of glycan structures. Furthermore, disease-associated alteration of glycogene expression can change these glycan † Originally submitted and accepted as part of the “Glycoproteomics” special section, published in the February 2009 issue of J. Proteome Res. (Vol. 8, No. 2). * To whom correspondence should be addressed. Hisashi Narimatsu, Tel: +81-29-861-3200. E-mail: [email protected]. ‡ Research Center for Medical Glycoscience, National Institute of Advanced Industrial Science and Technology (AIST). # These authors contributed equally to this study. § Department of Clinical Molecular Informative Medicine, Nagoya City University Graduate School of Medical Sciences. ⊥ Department of Gastroenterology, Graduate School of Comprehensive Human Sciences, University of Tsukuba.

1358 Journal of Proteome Research 2009, 8, 1358–1367 Published on Web 01/29/2009

structures. For example, glycoproteins of cancer cells undergo an abnormally high degree of N-glycan branching, extension with polylactosamine chains, fucosylation and sialylation. Such glycan associated alterations can affect the biological characteristics of the tumor cells.8,9 Additionally, compared with O-glycans from normal cells, mucin type O-glycans from cancer cells can be highly sialylated but less sulfated. Cancer associated O-glycans are often truncated and usually contain Tn, T, sialylTn (STn) and sialyl-T antigens.10 Thus, alterations to N- and/ or O-glycans found in glycoproteins derived from cancerous tissue are a common feature and can be used as a convenient biomarker. Therefore, it is clinically important to be able to identify subtle changes to glycan structures. Furthermore, a systematic understanding of glycan alterations to proteins will be achieved by correlating the glycosyltransferase expression profile with the actual glycan structure. Gene expression profiling has become a common method for understanding a broad range of biological and medically related phenomena. Genome researches have successfully completed DNA microarray platforms for comprehensive transcription level analyses, which carry probes covering over 22 000 human genes.11 Indeed, the new technology has realized 10.1021/pr800735j CCC: $40.75

 2009 American Chemical Society

research articles

Systematic Strategy for Glyco-Alteration Analysis 12

outstanding throughput for biological studies. Many statistical analyses and programs were developed for post-processing and interpreting DNA microarray data.13 Furthermore, validation processes were also examined and are now considered to be necessary due to the low-quantitative capability of DNA microarray.14 However, any validation methods are relatively time-consuming and decrease total throughput of gene expression profiling. Quantitative PCR (qPCR), being both sensitive and accurate, is one of the most popular validation protocols for DNA microarray analysis.15 Our interest is focused on glycogenes, which encode enzymes involved in glycan biosynthetic pathways. The human genome contains several hundred glycogenes.6,7 To improve the total throughput for glycogene expression profiling, we omitted DNA microarray analysis and performed multiplex qPCR array analysis for glycogene expression profiling. An important advantage of gene expression analysis is that it constitutes a simple and uniform process for material handling. By contrast, glycan analyses of N- and O-glycans from glycoproteins, glycolipids and their sulfated or sialylated derivatives involve diverse protocols. Lectin microarray is an emerging glycotechnology that enables ultrasensitive glycan profiling in a high-throughput format on the basis of multiplex lectin-glycan interaction analyses.16-18 A number of different protocols have been developed for detecting alterations to the glycosylation profile of cell surfaces, viruses and bacteria.19-22 Recent advances using an ultrasensitive lectin microarray23 facilitates differential glycan analysis by targeting a restricted region of one-dot sections on a formalin-fixed tissue array.24 Furthermore, a practical data-mining strategy with a gain-merging process and a max-normalization procedure makes it possible to perform a feasible statistical analysis using a data set derived from a lectin microarray.25 These technologies are straightforward and have the potential to detect a series of novel glyco-alterations. Recently, novel MS methodologies have been developed as a powerful tool for glycan structural analysis.26-28 MS provides higher sensitivity and more rapid glycan analysis than NMR, but requires various pretreatment procedures for detailed structural characterization. In MS analysis, ionization efficiency of glycans, especially acidic oligosaccharides (e.g., sialylated or sulfated glycans), is generally low in contrast to that of peptides and proteins. To ensure the highest sensitivity and to facilitate unambiguous sequencing in tandem MS experiments, glycans are usually derivatized by methylation (i.e., permethylation and esterification)29,30 or by reducing end tagging (e.g., 2-aminopyridine (PA) and 2-aminobenzamide (AB))31,32 before MS analysis. Additionally, structural assignment of glycans could easily be performed by exploiting the data analysis tools such as the interpretation of MS data and MSn spectra libraries for highthroughput MSn analyses.33-37 Although each profiling methodology (i.e., qPCR array, lectin microarray and MS) has a certain advantage in detecting glycan alterations of proteins, it is still difficult to elucidate detailed glycan structures by relying on a single technique alone. By using a combination of several analytical methods, however, it should be possible to determine the glycan structures and better understand glyco-alterations that occur on cell surfaces and/or on proteins. Here, we introduce a systematic strategy for glyco-alteration analysis by using multiple glycan profiling tools, that is, multiplex qPCR array, lectin microarray and mass spectrometry. Additionally, we demonstrate our approach as an example of N-glycan structural analysis using alpha-fetoprotein (AFP) from two hepatic cell lines.

Experimental Section Materials and Reagents. All reagents were purchased from QIAGEN (Hilden, Germany), Roche Diagnostics (Mannheim, Germany), Eurogentec (Lie`ge, Belgium), Sigma-Aldrich (St. Louis, MO), Wako Chemicals (Osaka, Japan), Nacalai Tesque (Kyoto, Japan), Takara Bio Inc. (Otsu, Japan) and Fluka (St. Louis, MO) unless otherwise noted. AFP from human placenta was from Cosmo Bio, Inc., Tokyo, Japan. Rabbit anti-human AFP polyclonal antibody (anti-AFP) was purchased from DakoCytomation (Glostrup, Denmark). Hepatic cell lines (HepG2 as hepatoblastoma cell line and HuH-7 as hepatocellular carcinoma cell line) were obtained from Riken Cell Bank (Ibaraki, Japan). These cell lines were cultured at 37 °C in a humidified atmosphere of 5% CO2 and 95% air in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum, 4 mM L-glutamine, 4.5 g/L glucose, 1000 U/mL penicillin and 1 mg/mL streptomycin. Expression Profiling of Genes Related to Glycan Synthesis. HepG2 and HuH-7 cells were cultured in the above conditions. A total of 1 × 107 cells in sparse condition were rinsed with phosphate buffered saline (PBS) and lysed on a plate with lysis buffer containing RNeasy plus mini kit (QIAGEN). First strand cDNA was synthesized with QuantiTect Rev. Transcription kit (QIAGEN) from 4 µg of total RNA. Oligonucleotide DNAs as PCR primer with hydrolysis probes were designed for expression analysis of 186 glycogenes, related to glycan-biosynthesis, and 3 housekeeping genes. PCR primers were designed using ProbeFinder via the Web (https://www.roche-applied-science. com/). Candidates with high score values were validated using the corresponding plasmid DNAs as templates for PCR, which comprise identical nucleotide sequences to each primer set and probe. Assays indicating an improperly large cycle value or inadequate amplification curve were replaced with better candidates using the procedure described by Mouritzen et al.38 A total of 189 assays were carried out in duplicate using 384well formatted plate on LightCycler 480 (Roche Diagnostics) with the following cycle conditions; 50 °C for 2 min, 95 °C for 2 min, and 55 cycles of 95 °C for 15 s and 60 °C for 1 min 20 s. Individual reaction volume was 7.5 µL, which consisted of 3.75 µL of 2-fold enzyme mixture; qPCR Quick GoldStar Mastermix Plus (Eurogentec), 0.1 µL of 10 µM of forward and reverse primers and hydrolysis probe selected from Universal ProbeLibrary (Roche Diagnostics). Assays with the qPCR array and reference template were performed in triplicate. Our results revealed that analytical errors were negligibly small across measurements; that is, coefficient of variance (CV%) of each assay was smaller than 3.1, and crossing mean of CV%s was 0.41. Coefficient values for 189 individual assays were calibrated from the results at two different concentrations using a 100 000 and 1000 copies of each of the 189 plasmids. Main parts of the plasmid collection consisting of glycogenes are the product of a previous project by us.7 Details of each assay are listed in Supporting Infromation Table 1. For analysis of unknown samples, cDNA aliquots equivalent to 7.5 ng of total RNA were subjected to individual reaction. To compare expression profiles of the two cells, the copy number of each transcript was normalized against arithmetical mean of all the glycogene transcripts in each cell. The normalization procedure is similar to the average-scaling method commonly used for DNA microarray analysis. Lectin Microarray Analysis of Hepatic Cell Lines. For cell lysate analysis, HepG2 and HuH-7 cells were cultured in the Journal of Proteome Research • Vol. 8, No. 3, 2009 1359

research articles

Ito et al.

above conditions. A total of 1 × 10 cells in sparse condition were rinsed with PBS and lysed with 0.2% SDS/PBS (designated as whole cell lysate). The protein concentration of the whole cell lysate was measured using a Micro BCA Protein Assay kit (PIERCE, Rockford, IL). A small aliquot of the obtained protein (100 ng) was labeled with 10 µg of Cy3-succimidyl ester at room temperature for 1 h in the dark. To block the functional group of the fluorescent reagent, the probing buffer (500 mM glycine in Tris buffered saline (TBS) containing 1.0% Triton X-100) was added and further incubated at room temperature for 2 h in the dark. The resultant Cy3-labeled glycoprotein solution (60 µL, 2.0 µg/mL) was applied to the lectin microarray. After incubation at 20 °C for 12 h, the glass slide was washed three times with the probing buffer and scanned by an evanescentfield fluorescence scanner, GlycoStation (Moritex, Co., Tokyo, Japan). All the data were analyzed with the Array Pro analyzer version 4.5 (Media Cybernetics, Inc., Bethesda, MD). The net intensity value for each spot was calculated by subtracting the background value from the signal intensity values of three spots. For culture supernatant analysis, the 90% confluent adherent cells were rinsed with protein free medium (i.e., DMEM) and further cultivation was carried out in DMEM. After 48 h, the conditioned medium was harvested and cells were removed by centrifugation at 4500g. The resultant culture supernatant was subjected to the protein quantitation and the lectin microarray analysis as described earlier. Immunoprecipitation. AFP was purified from cultured supernatant using immnoaffinity chromatography. Briefly, cultured supernatant (200 mL) filtered through a 0.45 µm filter cartridge was mixed with NaN3 (final concentration 0.1%) and then pretreated with 2 mL of Sepharose 4B to eliminate nonspecific proteins absorbed to the resin. The pass-through fraction was subsequently loaded onto an anti-AFP Sepharose 4B column (1.3 mg IgG/mL gel, column volume: 2 mL). After washing the column with 20 mL of 50 mM Tris-HCl (pH 8.0) containing 0.15 M NaCl and 0.1% NaN3, the bound AFP was eluted with 0.1 M glycine (pH 2.5). The eluate was immediately neutralized with 3 M Tris-HCl (pH 8.5) and pooled. Western-Blot Analysis. The purified AFPs were electrophoresed under reducing conditions on 10% polyacrylamide gels. The separated proteins were transferred to a PVDF membrane. After blocking with 5% skimmed milk in PBS containing 0.1% Tween 20 (PBST), the membrane was incubated with an anti-AFP (1/1000 diluted), and then with peroxidase-conjugated anti-rabbit IgG (1/3000 diluted; GE Healthcare UK Ltd., Little Chalf-ont, U.K.). Cross-reacting bands were detected with Western Lightning Chemiluminescence Plus (Perkin-Elmer LAS, Inc., Boston, MA). Antibody-Overlay Lectin Microarray. Antibody-overlay lectin microarray was performed essentially as described previously.39,40 Briefly, the purified proteins were diluted to 60 µL with PBS containing 0.1% Triton X-100 (PBSTx) and then applied to the lectin microarray, which we developed previously.18,23 After incubation at 20 °C for 12 h, 20 µg of human serum polyclonal IgG was added to the glass slide followed by a 30-min incubation. The reaction solution was discarded and the glass slide was washed three times with PBSTx. A total of 60 µL of biotinylated antibody solution in PBSTx was applied to the array and then incubated at 20 °C for 1 h. After washing three times with PBSTx, 60 µL of Cy3-labeled streptavidin (GE Healthcare) solution in PBSTx was added to the array and then incubated at 20 °C for 30 min. The glass slide was rinsed with PBSTx and scanned by the Glycostation (Moritex, Co., Tokyo, 7

1360

Journal of Proteome Research • Vol. 8, No. 3, 2009

Figure 1. Identification of glycosylation changes (glyco-alteration) in a glycoprotein using multiple glycan profiling tools.

Japan). All of the data were analyzed with the Array Pro analyzer version 4.5. The net intensity value for each spot was calculated by subtracting the background value from the signal intensity values of three spots. To obtain a glycan profile of desialylated AFP, equal amounts of AFPs were incubated with Sialidase A from Arthrobacter ureafaciens (ProZyme, Inc., San Leandro, CA) at 37 °C for 2 h and then subjected to the lectin microarray as described above. Release and Reduction of N-Glycans from the Purified AFPs. The buffer for AFP, which was purified from cultured supernatant using anti-AFP polyclonal antibody, was changed by dialysis against 3 × 2.5 L of 50 mM ammonium bicarbonate (pH 8.5) at 4 °C for 48 h. After dialysis, the sample was lyophilized. All samples were dissolved in 30 µL of distilled water. Three microliters of denaturation buffer (100 µL of 1% SDS in 1 M Tris-HCl (pH 8.6)/1.5 µL of 2-mercaptoethanol) was added to each sample solution (30 µL) and heated at 100 °C for 5 min. Subsequently, 2.5 mU PNGaseF (Takara Bio Inc.) was added to the solutions, followed by incubation at 37 °C for 15 h to remove N-glycans from each AFP. The released N-glycans were purified using a reverse-phase column Oasis HLB (10 mg/mL; Waters) and then subjected to reduction. A

Systematic Strategy for Glyco-Alteration Analysis

research articles

Figure 4. Electrogram of AFPs purified from culture supernatants. AFPs were purified from culture supernatants of HuH-7 and HepG2 cell lines with anti-AFP polyclonal antibody immobilized resin. AFP in each elution fraction (lanes 2 and 5 for HepG2; lanes 3 and 6 for HuH-7) was subjected to SDS-PAGE and then qualified by an electrogram with silver staining (A) and Western blotting (B). As a positive control, placental AFP (lanes 1 and 4) was also applied to the same gel.

Figure 2. Differential expression profiling of glycogenes between HepG2 and HuH-7 cells. Absolute copy number of transcripts in 7.5 ng of total RNA was scatter-plotted using a log-scale. Differentially expressed genes are indicated by gene symbols beside their markers. The copy numbers of B4GALNT3 were plotted even though its numbers are lower than 100.

total of 40 µL of 500 mM NaBD4 in 50 mM NaOH was applied to each purified N-glycans and incubated at 50 °C for 10 h. Each reaction solution was neutralized by adding 20 µL of 10% aqueous acetic acid. The reduced N-glycans were desalted by a cation-exchange column Oasis MCX (60 mg/3 mL; Waters, Milford, MA). Alditols were eluted with 750 µL of distilled water

and dried with a Speed-Vac. The remaining borate was removed by the addition of 50 µL of 1% acetic acid in methanol and dried under vacuum several times. Permethylation of N-Glycans Taken from Each AFP Sample. Permethylation was performed using the solid NaOH technique.41,42 Small NaOH pellets (approximately 50 mg; Fluka) were first mixed with 250 µL of DMSO (∞Pure; Wako). The released N-glycans (alditols) from the AFP samples were dried in a glass tube and approximately 50 µL of the NaOH/DMSO slurry was added to the sample followed by 50 µL of iodomethane (Wako). The reaction mixture was agitated at room temperature for 30 min. The reaction was terminated by addition of 1 mL of ice-cold distilled water. The diluted reaction

Figure 3. Differential glycan profiling between two hepatic cell lines (HuH-7 and HepG2). (A) Protein mixtures in culture supernatant were labeled with Cy3-SE and subjected to the lectin microarray. Fluorescent intensities of 43 lectins with the glycoproteins (2 µg/mL) on the lectin array are indicated with white bars. The fluorescent intensities, obtained using an equal amount of Sialidase A-treated glycoprotein, are indicated with black bars. (B) Protein mixtures in whole cell lysate treated with 0.2% SDS were labeled with Cy3-SE and applied to the lectin microarray. Fluorescent intensities of 43 lectins with the glycoproteins (2 µg/mL) are indicated with black bars. Journal of Proteome Research • Vol. 8, No. 3, 2009 1361

research articles

Figure 5. Focused differential profiling of AFPs. Differential glycan profiling among AFPs. The microarray was incubated with AFPs purified from culture supernatants of HepG2 (B) and HuH-7 (C) cells together with control AFP from placenta (A). AFP-binding signals on the array were detected with anti-AFP polyclonal antibody as described previously.39,40 Relative intensities of 8 lectins with the glycoproteins determined from the ratio to the maximal fluorescent intensity on the lectin array are indicated with white bars (top). Detailed structural information of glycans from the various AFPs was obtained using Sialidase A-treated samples. The resulting asialo-AFPs were subjected to antibodyoverlay lectin microarray. Relative intensities of 13 lectins with the asialo-AFPs were determined from the ratio of the maximal fluorescent intensity on the lectin array (indicated with black bars at the bottom).

solution was applied to a reverse-phase column Oasis HLB (10 mg/mL; Waters), followed by elution of permethylated alditols with 600 µL of acetonitrile and dried with a Speed-Vac. MALDI MS and MSn Analysis of Permethylated N-Glycans. The mass spectra and MSn spectra were carried out in reflectron positive ion mode with a Reflex IV MALDI-TOF (Bruker-Daltonik GmbH, Bremen, Germany) and an AXIMAQIT MALDI quadrupole ion trap TOF instruments (Shimadzu Corp., Kyoto, Japan), respectively. For sample preparation, the dried permethylated sample was resuspended in 25 µL of acetonitrile. A total of 0.5 µL of matrix solution (1 mg of 2,5DHB dissolved in 1 mL of 30% ethanol) and 0.5-1 µL of the diluted analyte solution were spotted on the plate and mixed on the plate. Finally, the dried matrix-analyte mixture was recrystallized with 1 µL of ethanol. All MS and MSn spectra were obtained from Na+ adduct ions.

Results and Discussion Strategy for Glyco-Alteration Analysis of Glycoproteins by Multiple Glycan Profiling Tools. To discover glycoalteration of a glycoprotein even though the alteration occurs 1362

Journal of Proteome Research • Vol. 8, No. 3, 2009

Ito et al. in minor population of the glycoprotein, a systematical and logical strategy should be employed using multiple technologies for glycan analysis. In this study, two analytical phases, that is, cell-level analyses on whole cells followed by molecularlevel glycan analyses on a specific glycoprotein as a model, were sequentially and efficiently carried out (Figure 1). First, a differential analysis of the cellular whole glycome was performed. Expression levels of the glycogenes were analyzed using a multiplex qPCR array format, and a lectin microarray system was used for glycan profiling of a pool of cell glycoproteins. Next, we examined that the glyco-alteration thus identified was expressed in a secretory glycoprotein. We performed glycan profiling of the target protein by antibody-overlay lectin microarray and detailed glycan structural assignments using tandem MS. Finally, by combining each data set obtained from the qPCR array, lectin microarray and MSn analyses, we were able to verify glyco-alterations to specific glycoproteins. To evaluate this concept, two hepatic cell lines, HepG2 and HuH-7 were used as model cell lines because it had been reported that these cell lines have almost the same properties. As both cell lines abundantly express AFP, AFP was selected as a model glycoprotein for analysis of glycan structural differences on a specific glycoprotein. By focusing on a target glycoprotein such as AFP, we successfully achieved the identification of different glycan structures on AFP between two cell lines, including a novel finding, that is, LacdiNAc structure, as in the following results. Differences in the Expression Profile of Glycogenes in Two Hepatic Cell Lines. For the purpose of investigating glycogene expression, we selected qPCR which was considered to be suitable for gene-expression analysis. Because the expression levels of most glycogenes are believed to be low, we were concerned as to whether DNA microarray analysis was a sufficiently sensitive technique. Indeed, Comelli et al. attempted to detect 46 glycogene transcripts in mice liver using DNA microarray, but succeeded in detecting only 15 genes.43 The following experiments were designed to validate the methodology. By using a qPCR protocol, we were able to detect 34 out of 46 genes from human liver (data not shown). Therefore, we consider that the higher sensitivity of qPCR is necessary for expression analysis of glycogenes. The expression profile of 186 glycogenes were analyzed and compared between two cell lines, HepG2 and HuH-7 (Figure 2). Absolute copy numbers of transcripts in 7.5 ng of total RNA were plotted on a log-scale. For the starting material, RNA was extracted from 107 cells and the equivalent amount from approximately 106 cells was subjected to expression analysis. The equivalent amounts of total RNA from approximately 103 cells were analyzed in individual reaction wells. Genes having a value smaller than 1000 in Figure 2 were expressed in a limited population of cells. Therefore, genes with a smaller number of transcripts than 100 were omitted from our results. Glycogene expression profiling indicated 18 genes were differentially expressed between the two hepatic cell lines, HepG2 and HuH-7. B3GNT3 was expressed only in HepG2 cells with a copy number >100. Nine glycogenes were expressed only in HuH-7 cells: HS3ST4, B3GNT5, UGT8, B4GALNT2, HAS2, CHST11, B3GALNT1, GALNT12 and ST6GALNAC1. Four glycogenes, ChGn, FUT6, GALNAC4S-6ST and HS6ST1, were expressed at an elevated level (i.e., >10-fold higher) in HepG2 compared with HuH-7. Conversely, four glycogenes, B3GALT1, GALNT7, GALNT3, GCNT1, were expressed at a higher level in HuH-7 compared with HepG2 cells. For an initial interpretation

Systematic Strategy for Glyco-Alteration Analysis

research articles

Figure 6. MALDI-TOF MS spectra of N-glycans released from AFPs. (A) Placenta, (B) HepG2, (C) HuH-7. The signals indicated with Arabic numerals are summarized in Table 1.

of the glycogene expression profiling, we employed a glycoproteomic approach to analyze N-glycan synthesis and modification enzymes. N-glycan carbohydrates undergo a variety of terminal modifications that are suitable for analysis using lectin array and MS techniques. FUT6, which is significantly expressed in HepG2 cells, encodes R1,3-fucosyltransferase that synthesizes the Lewis X (Lex) and sialyl Lex carbohydrate structure on nonreducing end of glycans. Thus, increased abundance of the various N-glycan modifications in the two cell lines was predicted from our glycogene expression profiling. Differential Glycan Profiling of Two Hepatic Cell Lines Using a Lectin Microarray. To obtain glycan profiles of a mixture of glycoproteins secreted from either HepG2 or HuH-7 cell lines, various concentrations of Cy3-labeled glycoprotein solution (60 µL, 0.5, 1.0, 2.0, 4.0, and 8.0 µg/mL) prepared from the cultured supernatants were subjected to lectin microarray analysis. For robust analysis, the obtained scan data were processed to expand the dynamic range of signal intensity using a novel data-mining system, which we recently developed.25 A dose-dependent increment of signal intensity was observed on the positive-spots when analyzing both cell lines (Supporting Information Figure 1). We acquired a cell-specific glycan profile within the range of protein concentration from 1.0 to 4.0 µg/ mL. However, the signals of some lectins (TJA-I, DSA, RCA120 and LEL) in HepG2 cells were saturated at 8 µg/mL. Thus, a glycan profile of 2 µg/mL of the cell glycoprotein solution was used for the following differential analysis (Figure 3A). To eliminate the effect of sialylation on the glycan profile, equal amounts of desialylated glycoprotein solution prepared by Sialidase A treatment were simultaneously applied to the lectin microarray. As a result, any obvious differences in signal pattern for N-glycosylation were evident by comparison between the HepG2 and HuH-7 cell lines as summarized below: (1) higher terminal fucosylation in HepG2 cells as revealed by signals on AAL and AOL, (2) incremental increase in LacNAc structures evident from signals of ECA, RCA120 and PHA(E), (3) increased level of branching of N-glycans in HepG2 estimated by signals

of PHA(L) and DSA. A significant difference in O-glycosylation is explicable in terms of altered sialylation of the core 1 structure represented by signals on Jacalin, MPA, ABA, ACA and MAH. In addition, existence of a unique LacdiNAc structure in HuH-7 was predicted by observation of the signal enhancement on WFA and WGA after sialidase treatment (Figure 3A). These results correlated well with the quantitative analysis of glycogene expression by RT-qPCR array. The two cell lines were further analyzed using whole cell lysate with 0.2% SDS treatment. The obtained profile was quite different from that of the cultured supernatant, that is, highcontent of high-mannose type and/or immature N-glycans in the whole cell lysate revealed by signals of NPA, GNA, HHL and UDA (Figure 3B). However, this result was not consistent with the above glycogene expression analysis. These results suggest that to detect glyco-alteration, based on a multiple use of glycan profiling tools, a target glycoprotein should be isolated from the cultured supernatant to act as an analyte. Focused Differential Glycan Profiling of AFP. Next, we examined whether the observed glyco-alterations between the two hepatic cell lines HepG2 and HuH-7 are reflected in the glycosylation status of a target glycoprotein. To validate the glyco-alteration discovery system, AFP, which bears an N-glycan without any O-glycosylation, was used as a model N-glycoprotein during the following study. AFP was chosen because the glyco-alteration of this glycoprotein has been welldocumented; that is, increment of R1,6-fucosylated AFP (L3 fraction) in serum correlated well with hepatic tumor progression. About 500 µg of AFPs was efficiently enriched from cultured supernatant of each cell line by affinity column chromatography with an anti-AFP polyclonal antibody Sepharose 4B column (see electrogram in Figure 4A). After protein quantification by Western blotting (Figure 4B), 1/10 000 of the obtained AFPs (approximately 50 ng) were subjected to antibodyoverlay lectin microarray assisted with the anti-AFP polyclonal antibody detection system as described previously.39,40 A placental AFP was also analyzed as a control, that is, less Journal of Proteome Research • Vol. 8, No. 3, 2009 1363

research articles

Ito et al.

Table 1. Summary of N-Glycan Structures Identified by MALDI-TOF MS and MALDI-QIT-TOF MS of Each AFP

a All observed ion correspond to [M + Na]+.. Green circle, Man; blue square, GlcNAc; yellow circle, Gal; yellow square, GalNAc; red triangle, Fuc; red diamond, Neu5Ac.

abundant R1,6-fucosylated form. As anticipated, the apparent difference in N-glycosylation of AFP between hepatic cells and placenta corresponded to the signals on R1,6-fucose binders, AOL, LCA and PSA (see top of Figure 5). Furthermore, the terminal-fucose binder AAL tends to have a high affinity for HepG2 AFP rather than HuH-7 AFP. This result suggests that HepG2 AFP bears a terminal fucosylation, including blood type H, Lex, Ley and sialyl Lex antigens. Indeed, recent targeted glycoproteomics analysis revealed that HepG2 cells express high levels of sialyl Lex on secreted glycoproteins, such as transferrin and β2-glycoprotein, in contrast to HuH-7 cells.44 Interestingly, taken together with the profiles of the desialylated derivatives (see bottom of Figure 5), the alternative forms of glycosylation 1364

Journal of Proteome Research • Vol. 8, No. 3, 2009

found on signals for WGA and WFA indicated the presence of a LacdiNAc structure. This glycan epitope represented by partial substitution of the LacNAc unit of the N-glycans might effect both the signal enhancement on GalNAc binders, BPL, MPA and Jacalin, and the signal reduction on Lac/LacNAc binder RCA120 and biantennary asialo-N-glycan (NA2) binder PHA(E) (see Supporting Information Table 2). This novel finding can be easily estimated from our two different profiling techniques (i.e., glycogene expression profiling by qPCR array and cell glycoprotein profiling by lectin microarray). Glycan Structural Analysis of AFPs Using Tandem Mass Spectrometry. As mentioned earlier, the purified AFP (HepG2 and HuH-7) and a placental AFP were subjected to analysis of

Systematic Strategy for Glyco-Alteration Analysis

research articles

Figure 7. Determination of the detailed N-glycan structures released from AFPs. MS/MS spectra of peak 7, 8, 11 and 13 by tandem MS are shown. Structural analyses of all signals indicated in Figure 6 are summarized in Table 1. The Na+ adduct was chosen as a parent ion for all MS/MS spectra.

Figure 8. Expression levels of the relevant genes for the observed glyco-alterations. (A) Six FUT genes encode enzymes synthesizing the Lex and sialyl Lex structures. Absolute copy number of transcripts; closed bar, HepG2; open bar, HuH-7. FUT6 was predominantly expressed in HepG2 cells. (B) Expression of two genes encoding LacdiNAc synthase: B4GALNT3 and B4GALNT4. The level of expression of B4GALNT3 was 10-fold higher in HuH-7 than in HepG2 cells. Values were obtained from experiments performed in duplicate. Each result is listed in Supporting Information Table 3.

N-glycans by mass spectrometry. Additionally, two glycan analysis procedures were applied in this study. (1) Reduction of the released N-glycans was performed to enable distinction of fragment ions, that is, reducing end GlcNAc and nonreducing end HexNAc. (2) Permethylation method was used because of the analytical advantages, such as stabilization of sialic acid residue and increased ionization efficiency. Figure 6 shows the MS spectra of N-glycans released from each AFP and the observed MS signals of the glycans (peaks 1-15 in Figure 6). For identification of the detailed glycan structures, all MS signals in Figure 6 were subjected to MSn analysis and the assigned signals as N-glycan are summarized in Table 1.

Several major differences in the N-glycans of AFP derived from hepatic cells and placenta were detected. Most N-glycans from hepatic cells (HepG2 and HuH-7) AFP had an R1,6-fucose residue at the reducing terminus (core-fucose), whereas Nglycans from placental AFP mainly lacked the R1,6-fucose residue. This conclusion is consistent with the MS/MS spectra, where fragment ions [M + Na - 468]+ indicate the presence of an R1,6-fucose residue and [M + Na - 294]+ indicate the absence of an R1,6-fucose residue from the reducing end (Figure 7). Moreover, N-glycan structures of AFP from two hepatic cell lines (HepG2 and HuH-7) were also different. In HepG2 AFP, Lex and sialyl Lex structures at the nonreducing Journal of Proteome Research • Vol. 8, No. 3, 2009 1365

research articles 44

terminus were observed. MS/MS spectra of such molecular ions revealed signals at m/z 660 and 1021, supporting the presence of Lex and sialyl Lex structures, respectively (Figure 7). A sialyl LacdiNAc structure was observed as a novel feature of HuH-7 AFP. Indeed, the fragment ion at m/z 888 provides evidence of the sialyl LacdiNAc structure (Figure 7). This is the first study to describe the unique LacdiNAc structure in HuH-7 AFP. In conclusion, the results of our MSn structural analyses were entirely consistent with those of the glycan profiling using qPCR array and lectin microarray analyses. Expression of Causative Genes for the Observed GlycoAlteration on AFP Detected by Lectin Microarray and MS Analyses. To support the results of glycan analysis using lectin microarray and mass spectrometry, we focused on genes whose expression was related to the observed glycan structural differences, that is, Lex, sialyl Lex and LacdiNAc structures. Lex and sialyl Lex epitopes contain fucosylated GlcNAc, which is generated by R1,3 fucosyltransferase activity. The absolute copy number of 6 gene transcripts (FUT3, FUT4, FUT5, FUT6, FUT7 and FUT9) that encode R1,3 fucosyltransferase are shown in Figure 8A. FUT6 was more highly expressed in HepG2 than in HuH-7 cells (i.e., 70-fold difference). Indeed, FUT6 is known to be the major R1,3 fucosyltransferase in hepatic cells.44 Thus, we fully anticipated the down-regulation of FUT6 expression in HuH-7 cells. LacdiNAc is a unique N-glycan structure found in glycoproteins, which is synthesized by either β-1,4-Nacetylgalactosaminyltransferase III (B4GALNT3)45 or β-1,4-Nacetylgalactosaminyltransferase IV (B4GALNT4).46 The transcript levels of the genes encoding these two enzymes were relatively low and were missed in the overall analysis (Figure 2). Measurement results by duplicated analysis are compared in Figure 8B. B4GALNT3 expression was 10-fold higher in HuH-7 by comparison to HepG2, whereas B4GALNT4 displayed similar levels of expression in the two cell lines. Neither of these genes are expressed in normal liver tissue.45,46 Thus, ectopic expression of B4GALNT3 in HuH-7 may have caused the emergence of the unusual LacdiNAc moiety on AFP observed in this study. In conclusion, we have successfully predicted glyco-alterations by glycogene expression profiling, and identified the relevant genes from only 4 µg of total RNA. Thus, expression profiling of 186 glycogenes from a small specimen sample potentially provides an overview of glycan alterations. We believe glycogene expression profiling is a fully supportive tool for searching glycan alterations.

Conclusion To analyze alterations to glycan structures on the cell surface and/or target proteins, it is necessary to identify differences in each glycoform. Here, we propose a systematic approach for glyco-alteration using three glycan profiling tools. First, the expression pattern of 186 human glycogenes (e.g., glycosyltransferases, sulfotransferases and sugar-nucleotide transporters) are analyzed using a multiplex qPCR array platform to test for regulation of glycan structures at the transcriptional level. A genetic approach to glycomics and glycoproteomics is useful in obtaining information concerning the relationship between the glycogene expression profile and the actual glycan structure. Second, glycan alteration analysis is performed by lectin microarray as a multiple glycan-lectin interaction profile system prior to MS analysis. Recent improvements in the lectin microarray system facilitate profiling of glycoforms using only a nanogram-order of analyte. With this highly sensitive system, 1366

Journal of Proteome Research • Vol. 8, No. 3, 2009

Ito et al. we were able to analyze structural differences on a specific target glycoprotein, AFP. The detailed structure of a specific glycan from the targeted glycoprotein (e.g., AFP) is determined by tandem MS analysis. MSn techniques are especially useful for defining the linkage information between carbohydrate residues. By comparing the results of glycan structural analyses derived from lectin microarray and MSn data with those from the qPCR array, we are able to understand the biosynthetic pathways involved in these glycan alterations. In this way, a combination of these glycan profiling tools will facilitate a new systems glycomic and glycoproteomic approach to research.

Acknowledgment. We thank S. Unno, N. Uchiyama, Y. Kubo, J. Murakami, and M. Sukegawa for their technical assistance. This work was performed as part of the R&D Project of the Industrial Science and Technology Frontier Program supported by the New Energy and Industrial Technology Development Organization (NEDO). Supporting Information Available: Supplementary Tables 1-3, and Figure 1. This material is available free of charge via the Internet at http://pubs.acs.org. References (1) Varki, A. Glycobiology 1993, 3, 97–130. (2) Rudd, P. M.; Elliott, T.; Cresswell, P.; Wilson, I. A.; Dwek, R. A. Science 2001, 291, 2370–2376. (3) Dove, A. Nat. Biotechnol. 2001, 19, 913–917. (4) Haltiwanger, R. S.; Lowe, J. B. Annu. Rev. Biochem. 2004, 73, 491– 537. (5) Mitra, N.; Sinha, S.; Ramya, T. N.; Surolia, A. Trends. Biochem. Sci. 2006, 31, 156–163. (6) Kikuchi, N.; Narimatsu, H. Biochim. Biophys. Acta 2006, 1760, 578– 583. (7) Narimatsu, H. Glycoconjugate J. 2004, 21, 17–24. (8) Dennis, J. W.; Granovsky, M.; Warren, C. E. Biochim. Biophys. Acta 1999, 1473, 21–34. (9) Zhao, Y. Y.; Takahashi, M.; Gu, J. G.; Miyoshi, E.; Matsumoto, A.; Kitazume, S.; Taniguchi, N. Cancer Sci. 2008, 99, 1304–1310. (10) Brockhausen, I. EMBO Rep. 2006, 7, 599–604. (11) Yang, R. A. Cell 2000, 102, 9–15. (12) Taniguchi, M.; Miura, K.; Iwao, H.; Yamanaka, S. Genomics 2001, 71, 34–39. (13) Epstein, C. B.; Butow, R. A. Curr. Opin. Biotechnol. 2000, 11, 36– 41. (14) Rajeevan, M. S.; Ranamukhaarachchi, D. G.; Vernon, S. D.; Unger, E. R. Methods 2001, 25, 443–51. (15) Chuaqui, R. F.; Bonner, R. F.; Best, C. J.; Gillespie, J. W.; Flaig, M. J.; Hewitt, S. M.; Phillips, J. L.; Krizman, D. B.; Tangrea, M. A.; Ahram, M.; Linehan, W. M.; Knezevic, V.; Emmert-Buck, M. R. Nat. Genet. 2002, 32, 509–514. (16) Angeloni, S.; Ridet, J. L.; Kusy, N.; Gao, H.; Crevoisier, F.; Guinchard, S.; Kochhar, S.; Sigrist, H.; Sprenger, N. Glycobiology 2005, 15, 31–41. (17) Pilobello, K. T.; Krishnamoorthy, L.; Slawek, D.; Mahal, L. K. ChemBioChem 2005, 6, 985–989. (18) Kuno, A.; Uchiyama, N.; Koseki-Kuno, S.; Ebe, Y.; Takashima, S.; Yamada, M.; Hirabayashi, J. Nat. Methods 2005, 2, 851–856. (19) Ebe, Y.; Kuno, A.; Uchiyama, N.; Koseki-Kuno, S.; Yamada, M.; Sato, T.; Narimatsu, H.; Hirabayashi, J. J. Biochem. 2006, 139, 323–327. (20) Hsu, K. L.; Pilobello, K. T.; Mahal, L. K. Nat. Chem. Biol. 2006, 2, 153–157. (21) Pilobello, K. T.; Slawek, D. E.; Mahal, L. K. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 11534–11539. (22) Tateno, H.; Uchiyama, N.; Kuno, A.; Togayachi, A.; Sato, T.; Narimatsu, H.; Hirabayashi, J. Glycobiology 2007, 139, 323–327. (23) Uchiyama, N.; Kuno, A.; Tateno, H.; Kubo, Y.; Mizuno, M.; Noguchi, M.; Hirabayashi, J. Proteomics 2008, 8, 3042–3050. (24) Matsuda, A.; Kuno, A.; Ishida, H.; Kawamoto, T.; Shoda, J.; Hirabayashi, J. Biochem. Biophys. Res. Commun. 2008, 370, 259– 263. (25) Kuno, A.; Itakura, Y.; Toyoda, M.; Takahashi, Y.; Yamada, M.; Umezawa, A.; Hirabayashi, J. J. Proteomics Bioinf. 2008, 1, 68–72. (26) Dell, A.; Morris, H. R. Science 2001, 291, 2351–2356. (27) Harvey, D. J. Expert Rev. Proteomics 2005, 2, 87–101.

research articles

Systematic Strategy for Glyco-Alteration Analysis (28) Zaia, J. Mass Spectrom. Rev. 2004, 23, 161–227. (29) Ciucanu, I.; Costello, C. E. J. Am. Chem. Soc. 2003, 125, 16213– 16219. (30) Powell, A. K.; Harvey, D. J. Rapid Commun. Mass Spectrom. 1996, 10, 1027–1032. (31) Hase, S. Methods Enzymol. 1994, 230, 225–237. (32) Bigge, J. C.; Patel, T. P.; Bruce, J. A.; Goulding, P. N.; Charles, S. M.; Parekh, R. B. Anal. Biochem. 1995, 230, 229–238. (33) Ethier, M.; Krokhin, O.; Ens, W.; Standing, K. G.; Wilkins, J. A.; Perreault, H. Rapid Commun. Mass Spectrom. 2005, 19, 721–727. (34) Tang, H.; Mechref, Y.; Novotny, M. V. Bioinformatics 2005, 21, i431– i439. (35) Lapadula, A. J.; Hatcher, P. J.; Hanneman, A. J.; Ashline, D. J.; Zhang, H.; Reinhold, V. N. Anal. Chem. 2005, 77, 6271–6279. (36) Ceroni, A.; Maass, K.; Geyer, H.; Geyer, R.; Dell, A.; Haslam, S. M. J. Proteome Res. 2008, 7, 1650–1659. (37) Kameyama, A.; Kikuchi, N.; Nakaya, S.; Ito, H.; Sato, T.; Shikanai, T.; Takahashi, Y.; Takahashi, K.; Narimatsu, H. Anal. Chem. 2005, 77, 4719–4725. (38) Mouritzen, P.; Noerholm, M.; Nielsen, P. S.; Jacobsen, N.; Lomholt, C.; Pfundheller, H. M.; Tolstrup, N. Nat. Methods 2005, 2, 313– 316. (39) Kato, Y.; Kaneko, M. K.; Kuno, A.; Uchiyama, N.; Amano, K.; Chiba, Y.; Hasegawa, Y.; Hirabayashi, J.; Narimatsu, H.; Mishima, K.; Osawa, M. Biochem. Biophys. Res. Commun. 2006, 349, 1301–1307.

(40) Kuno, A.; Kato, Y.; Matsuda, A.; Kaneko, M. K.; Ito, H.; Amano, K.; Chiba, Y.; Narimatsu, H.; Hirabayashi, J. Mol. Cell. Proteomics 2009, 8 (1), 99–108. (41) Dell, A.; Khoo, K.-H.; Panico, M.; McDowell, R. A.; Etienne, A. T.; Reason, A. J.; Morris, H. R. In Glycobiology: A Practical Approach; Fukuda, M., Kobata, A., Eds.; Oxford University Press: Oxford, 1993; pp 187-222. (42) Lemoine, J.; Chirat, F.; Domon, B. J. Mass Spectrom. 1996, 31, 908– 912. (43) Comelli, E. M.; Head, S. R.; Gilmartin, T.; Whisenant, T.; Haslam, S. M.; North, S. J.; Wong, N. K.; Kudo, T.; Narimatsu, H.; Esko, J. D.; Drickamer, K.; Dell, A.; Paulson, J. C. Glycobiology 2006, 16, 117–131. (44) Higai, K.; Shibukawa, K.; Muto, S.; Matsumoto, K. Anal. Sci. 2003, 19, 85–92. (45) Sato, T.; Gotoh, M.; Kiyohara, K.; Kameyama, A.; Kubota, T.; Kikuchi, N.; Ishizuka, Y.; Iwasaki, H.; Togayachi, A.; Kudo, T.; Ohkura, T.; Nakanishi, H.; Narimatsu, H. J. Biol. Chem. 2003, 278, 47534–47544. (46) Gotoh, M.; Sato, T.; Kiyohara, K.; Kameyama, A.; Kikuchi, N.; Kwon, Y. D.; Ishizuka, Y.; Iwai, T.; Nakanishi, H.; Narimatsu, H. FEBS Lett. 2004, 562, 134–140.

PR800735J

Journal of Proteome Research • Vol. 8, No. 3, 2009 1367