Glycan profiling of endometrial cancers using lectin microarray

Authors

  • Yoshihiro Nishijima,

    1. Department of Reproductive Biology, National Research Institute for Child Health and Development, Setagaya-ku, Tokyo, Japan
    2. Department of Obstetrics and Gynecology, Specialized Clinical Science, Tokai University School of Medicine, Isehara-shi, Kanagawa, Japan
    Search for more papers by this author
  • Masashi Toyoda,

    1. Tokyo Metropolitan Institute of Gerontology, Itabashi-ku, Tokyo, Japan
    Search for more papers by this author
  • Mayu Yamazaki-Inoue,

    1. Department of Reproductive Biology, National Research Institute for Child Health and Development, Setagaya-ku, Tokyo, Japan
    Search for more papers by this author
  • Taro Sugiyama,

    1. Department of Obstetrics and Gynecology, Specialized Clinical Science, Tokai University School of Medicine, Isehara-shi, Kanagawa, Japan
    Search for more papers by this author
  • Masaki Miyazawa,

    1. Department of Obstetrics and Gynecology, Specialized Clinical Science, Tokai University School of Medicine, Isehara-shi, Kanagawa, Japan
    Search for more papers by this author
  • Toshinari Muramatsu,

    1. Department of Obstetrics and Gynecology, Specialized Clinical Science, Tokai University School of Medicine, Isehara-shi, Kanagawa, Japan
    Search for more papers by this author
  • Kyoko Nakamura,

    1. Department of Reproductive Biology, National Research Institute for Child Health and Development, Setagaya-ku, Tokyo, Japan
    Search for more papers by this author
  • Hisashi Narimatsu,

    1. Research Center for Medical Glycoscience, National Institute of Advanced Industrial Science and Technology, Tsukuba, Ibaraki, Japan
    Search for more papers by this author
  • Akihiro Umezawa,

    Corresponding author
    • Department of Reproductive Biology, National Research Institute for Child Health and Development, Setagaya-ku, Tokyo, Japan
    Search for more papers by this author
  • Mikio Mikami

    1. Department of Obstetrics and Gynecology, Specialized Clinical Science, Tokai University School of Medicine, Isehara-shi, Kanagawa, Japan
    Search for more papers by this author

  • Communicated by: Takashi Tada

Correspondence: umezawa@1985.jukuin.keio.ac.jp

Abstract

Cell surface glycans change during the process of malignant transformation. To characterize and distinguish endometrial cancer and endometrium, we performed glycan profiling using an emerging modern technology, lectin microarray analysis. The three cell lines, two from endometrial cancers [well-differentiated type (G1) and poorly differentiated type (G3)] and one from normal endometrium, were successfully categorized into three independent groups by 45 lectins. Furthermore, in cancer cells, a clear difference between G1 and G3 type was observed for the glycans recognized with six lectins, Ulex europaeus agglutinin I (UEA-I), Sambucus sieboldiana agglutinin (SSA), Sambucus nigra agglutinin (SNA), Trichosanthes japonica agglutinin I (TJA-I), Amaranthus caudatus agglutinin (ACA), and Bauhinia purpurea lectin (BPL). The lectin microarray analysis using G3 type tissues demonstrated that stage I and stage III or IV were distinguished depending on signal pattern of three lectins, Dolichos biflorus agglutinin (DBA), BPL, and ACA. In addition, the analysis of the glycans on the ovarian cancer cells showed that only anticancer drug-sensitive cell lines had almost no activities to specific three lectins. Glycan profiling by the lectin microarray may be used to assess the characteristics of tumors and potentially to predict the success of chemotherapy treatment.

Introduction

Glycosylation is one of the most common post- and co-translational modification of eukaryotic proteins and is known to have important functions as mutual cellular recognitions in biological processes (Opdenakker et al. 1993). The glycans expressed on the cell membrane especially vary among cell types and reflect cellular conditions during development, differentiation, activation, inflammation, and even malignant transformation. Therefore, glycan profiling provides us with a lot of information of cells and tissues and is attracting attention in the fields of reproductive medicine and oncology. Antibodies have been used to recognize glycan markers in cancer. However, lectins recognize specific and well-defined glycan epitopes, such as for blood group typing, tissue staining, lectin-probed blotting, and flow cytometry. A modern technology to discriminate glycan profiling is lectin microarray, an emerging technology that enables ultrasensitive detection of multiplex lectin–glycan interactions based on a unique principle; that is, the evanescent-field fluorescence-detection principle, which has been used extensively for biosensors to study real-time binding events on the glass slide surfaces (Kuno et al. 2005). Thus, the evanescent-field methods have greater advantage to analyze relatively weak interactions between lectins and glycans in a liquid phase at equilibrium. Owing to its extremely high sensitivity and accuracy, the lectin microarray system is the best tool for a ‘cell profiler’ to distinguish cancer, and it is expected to be applicable for selection of cancer-specific lectins.

Endometrial cancer is one of the most common malignancies of the female genital tract. A mass screening contributes to early detection of malignant disorders of the uterus and ovaries, and the progress of further diagnosis and treatment has been expected. In endometrial cancer, endometrioid carcinoma is classified into subtypes based on histological observation. One is the well-differentiated type, which is typically hormone-sensitive and low-stage, and has excellent prognosis. The other is the poorly differentiated type, with a poor prognosis with the tendency to recur, even at the early stage. We employ histological grading from Grade 1 to Grade 3 to endometrial cancer, based on the area of well-formed glands (well-differentiated type) versus solid nongland area (poorly differentiated type). Therefore, the histological diagnosis is extremely important and is now the only method for grading of endometrial cancers. Because histology is time-consuming and low-throughput, another precise and convenient methods for diagnosis of the endometrial cancer type would be preferred, but has not yet been achieved. For this purpose, we previously analyzed the glycolipids of uterine cancer and reported that sulfated glycolipids were specifically expressed in well-differentiated uterine cancer, which may have a role in tumor cell differentiation (submitted). In the present study, we analyzed the oligosaccharides of glycoproteins in endometrial cancers by lectin microarray. Moreover, cell surface glycans are considered to be good targets for analyzing subtle alterations of cancer cells such as sensitivity for anticancer drugs. We performed lectin microarrays using ovarian cancer cell lines including anticancer drug-sensitive cell lines and resistant ones and demonstrated a clear difference in the glycans between these cell lines. Glycan profiling by lectin microarray may therefore enable us to assess the characteristics of cancers from different patients and to recommend proper therapeutic treatment with anticancer drugs.

Results

It is well known that carbohydrate chains of glycolipids and glycoproteins on the cell surface are altered by carcinogenesis and transformation. We first performed glycan profiling using the endometrial cancer cell lines HEC-6 and HEC-50B. HEC-6 is a well-differentiated type (G1: Grade 1), and HEC-50B is a poorly differentiated type (G3: Grade 3). To evaluate the differences in their glycan processing, we carried out lectin microarray analysis of the membrane protein in these two cell lines (Fig. 1A). The binding of Cy3-labeled glycoprotein in the microarray wells and the data obtained are in triplicate for each of the 45 different lectins. The triplicate wells of each lectin showed almost same intensity, and we confirmed that the intensity was within the linear range. We then quantified lectin signal using ‘ArrayPro Analyzer’ software and calculated the net intensity of three spots for each lectin (Fig. 1B). Hierarchical clustering analysis (HCA) on all 45 lectins showed that three cell lines were reproducibly categorized into independent groups (Fig. 1C). HEC-50B was categorized closer to UtE1104 rather than HEC-6. Principal component analysis (PCA) on all lectins was performed, showing that the three cell lines were apparently categorized into three independent groups (Fig. 1D).

Figure 1.

Lectin microarray analysis of human endometrial cancer cell lines using 45 lectins. HEC-6, HEC-50B, and UtE1104 are the cell lines prepared from well-differentiated cancer cells (Grade 1), poorly differentiated cancer cells (Grade 3) and normal endometrium, respectively. (A) Scan images of the three cell lines for all 45 lectins. Fluorescent signal obtained are triplicate for each of the 45 different lectins. The names of the lectins used are shown in the lower part of the scan image. Four dots representing the positive control are at the left of lectin area. (B) Heat map of lectin microarray on the three cell lines. Sample numbers 1–3: HEC-6; 4–6: HEC-50B; 7 and 8: UtE1104. The samples from HEC-6 and HEC-50B were obtained from independent triplicate experiments, and the samples from UtE1104 were obtained from independent duplicate experiments. (C) Hierarchical clustering analysis of lectin microarray on the three cell lines. (D) Principal component analysis of lectin microarray on the three cell lines.

We then determined whether the lectin array data could be used to separate these cell lines into two groups, endometrial cancer (HEC-6, HEC-50B) and endometrium (UtE1104). Six lectins, Aspergillus oryzae lectin (AOL), Trichosanthes japonica agglutinin II (TJA-II), Aleuria aurantia lectin (AAL), Jacalin, Concanavalin A (ConA), and Lycopersicon esculentum lectin (LEL) that could distinguish endometrial cancer from normal endometrium, were identified with p-value <0.01. Each lectin signal was indicated in Fig. 2A as relative intensity. The AOL signals of HEC-6 and HEC-50B were remarkably higher than that of UtE1104, and the signals of TJA-II, AAL, Jacalin, and ConA on HEC cell lines were apparently higher than that on UtE1104. AOL, AAL, and TJA-II recognize nonreducing terminal fucose. Fucosylation by α1,2 fucosyltransferase is regulated by FUT1 (H type), and FUT2 (Se type) in human. Mutation of FUT1 is rare in the Japanese and 16% of the Japanese are homozygotes of the negative FUT2 alleles (Kudo et al. 1996; Kaneko et al. 1997). Lack of reactivity to AOL, AAL, and TJA-II in UtE1104 was not attributed to the negative FUT 2 alleles because endometrial tissue did not express the FUT 2 gene by the UniGene database and RT-PCR. In contrast, LEL signal intensities show that UtE1104 cells have higher binding activity than HEC cell lines. We then performed HCA and PCA with six lectins, which demonstrated that endometrial cancer cells and endometrium cells were clearly categorized into two independent groups (Fig. 2B). PCA also indicated the independency of these cell lines (Fig. 2C).

Figure 2.

Lectin microarray analysis of human endometrial cancer cell lines using six lectins. (A) Relative binding intensity of the cancer cell lines and the normal endometrium cell line, for six lectins (Aspergillus oryzae lectin, TJA-II, Aleuria aurantia lectin, Jacalin, Concanavalin A, and Lycopersicon esculentum lectin). Sample numbers 1–3: HEC-6; 4–6: HEC-50B; 7 and 8: UtE1104. The samples from HEC-6 and HEC-50B were obtained from independent triplicate experiments, and the samples from UtE1104 were obtained from independent duplicate experiments. (B) Hierarchical clustering analysis of lectin microarray on the three cell lines using six lectins. (C) Principal component analysis of lectin microarray on the three cell lines using six lectins.

We investigated the ability of the lectins to distinguish well-differentiated cells (HEC-6) from poorly differentiated cells (HEC-50B). We identified six lectins whose signals were significantly different between the two cell lines (Fig. 3A). The Ulex europaeus agglutinin I (UEA-I) signal was higher in HEC-6 than in HEC-50B. In contrast, Sambucus sieboldiana agglutinin (SSA), Sambucus nigra agglutinin (SNA), Trichosanthes japonica agglutinin I (TJA-I), Amaranthus caudatus agglutinin (ACA), and Bauhinia purpurea lectin (BPL) signals were higher in HEC-50B than in HEC-6. Because TJA-I, SSA, and SNA recognize not only sialyl α2-6Galβ1-4GlcNAc but also sulfate-6Galβ1-4GlcNAc, we performed sialidase treatment of HEC-50B. After sialidase treatment, the relative binding intensity of each of these three lectins was reduced (Fig. 3B). In contrast, intensities of the five lectins, Ricinus communis agglutinin I (RCA120), TJA-II, Phytolacca americana agglutinin (PWM), Jacalin and Wisteria floribunda Agglutinin (WFA) that recognize bare glycan structures because of removal of the sialyl group, increased (Fig. 3B). These data show that the reactivity of SSA, SNA, and TJA-I in HEC-50B is attributed to the presence of sialyl α2-6Galβ1-4GlcNAc. We then performed HCA and PCA to investigate cell variability during cultivation (Fig. 3C,D). Signal intensities of the six lectins show that HEC-6 and HEC-50B did not vary significantly by cultivation.

Figure 3.

Lectin microarray analysis of human endometrial cancer cell lines using six lectins. (A) Relative binding intensity of cancer cell lines, for six lectins (UEA-I, Sambucus sieboldiana agglutinin, Sambucus nigra agglutinin TJA-I, Amaranthus caudatus agglutinin, and Bauhinia purpurea lectin). Sample numbers 1–3: HEC-6; 4–6: HEC-50B. The samples from HEC-6 and HEC-50B were obtained from independent triplicate experiments. (B) Relative binding intensity of HEC-50B after sialidase treatment for each lectin. The samples (Cy3-labeled glycoprotein in probing buffer) were incubated with or without 5 mU sialidase (Glyco Sialidase A, ProZyme) for 2 h at 37 °C and were analyzed by lection microarray. The sialidase-treated samples were obtained from independent duplicate experiments. (C) Hierarchical clustering analysis of lectin microarray on two cell lines using six lectins. (D) Principal component analysis of lectin microarray on two cell lines using six lectins.

Endometrial cancer tissues as well as endometrial cancer cell lines were subjected to lectin microarray analysis (Fig. 4A). In addition to histopathological grading, samples were categorized into clinical staging (early stage I and advanced stages III/IV), based on cancer progression and patient prognosis. Clinical staging is very important in poorly differentiated endometrial cancer because of its therapeutic difficulty. A significant difference in lectin signal pattern between Stage I and Stages III/IV was detected. We identified three lectins, Dolichos biflorus agglutinin (DBA), BPL, and ACA, with P-value <0.01 (Fig. 4B). As DBA recognizes blood group A antigen, blood types of donors were investigated. For donors of all blood types (A, B, O, AB), the DBA signal in Stage III/IV was consistently higher regardless of blood type, suggesting that DBA reactivity in endometrioid cancer is independent of blood type. The signals from DBA, BPL, and ACA for Stage III/IV tissues were higher than those for Stage I. HCA using three lectins gave good separation, dependent on staging (Fig. 4C). Endometrial cancer tissue rated Grade 3 were apparently categorized into two different groups, early stage (Stage I) and advanced stage (Stage III/IV).

Figure 4.

Lectin microarray analysis of endometrial cancer tissues on different stages in Grade 3. (A) Heat map of endometrial cancer on different stages in Grade 3. Sample numbers 1–7: G3(I); 8–10: G3(III); 11: G3(IV). Each sample was obtained from a different donor: seven samples from G3(I), three samples from G3(III), and one sample from G3(IV). (B) Relative binding intensity of endometrial cancer on different stages in Grade 3, for three lectins (Dolichos biflorus agglutinin, Bauhinia purpurea lectin, and Amaranthus caudatus agglutinin). (C) Hierarchical clustering analysis of lectin microarray on endometrial cancer with different stages in Grade 3, using three lectins.

Along with successful categorization of endometrial cancer, we further extended the analysis to ovarian cancer in terms of anticancer drug effectiveness (Fig. 5A). We used a human ovarian cancer cell line (KF28) and anticancer drug-resistant cell lines that are cisplatin (CD)-resistant (KFr13, C13), taxol-resistant (KF28TX, 2008/PX24), and CD- and taxol- (CD&TX) resistant (KFr13TX). KF28TX and KFr13 are taxol- and CD-resistant cell lines that were established by repeated exposure of the parent KF28 cells to taxol and CD, respectively. We employed three lectins; Helix pomatia agglutinin (HPA), DBA and Psophocarpus tetragonolobus lectin I (PTL-I) (Fig. 5B). Interestingly, drug-sensitive cell subclones had almost no activities to these three lectins, but, drug-resistant subclones exhibited higher reactivities to these lectins. The CD-resistant subclones (TX&CD and CD) specially showed higher reactivities to these lectins, compared with taxol-resistant subclones. Hierarchical clustering analysis showed clear grouping among drug-sensitive and drug-resistant cells (Fig. 5C). Together, these results suggest that glycan information on the cancer cells with lectin microarrays can help us to assess tumor characteristics and possibly determine therapeutic strategy using anticancer drugs.

Figure 5.

Lectin microarray analysis of a human ovarian cancer cell line and the anti-cancer drug-resistant lines. Sample numbers 1 and 2: KF28 cisplatin (CD)- and taxol (TX)-sensitive cell line; 3 and 4: 2008/PX24 taxol-resistant cell line; 5 and 6: KF28TX taxol-resistant cell line; 7 and 8: KFr13TX CD- and taxol-resistant cell line; 9 and 10: KFr13 CD-resistant cell line; 11 and 12: C13 CD-resistant cell line. The samples from each cell line were obtained from independent duplicate experiments. (A) Heat map of an ovarian cancer cell line and anti-cancer drug-resistant lines. (B) Relative binding intensity of a human ovarian cancer cell line and the anti-cancer drug-resistant lines for three lectins (Helix pomatia agglutinin, Dolichos biflorus agglutinin and PTL-I). (C) Hierarchical clustering analysis of lectin microarray on a human ovarian cancer cell line and the anticancer drug-resistant lines, using 3 lectins.

Discussion

Changes in glycosylation status have been implicated in cancer (Hakomori 2002). The glycans on cell surface reflect the cellular condition and they are information-rich and of particular interest in cancer research. As glycan profiling tools, liquid chromatography, capillary electrophoresis, and mass spectrometry have been commonly used (Vanderschaeghe et al. 2010), and the combination of these techniques further enable to determine detailed structures of each glycan even in a very small amount. However, these methods are relatively low-throughput and require complicated experimental steps, including purification of glycans, and these methods present the difficulty of identifying O-glycans. Furthermore, these techniques may not be suitable for initial glycan profiling of cancer, because the object of analysis is not to identifying each glycan structure, but to determine the presence of cancer, based on information of the cells' whole repertoire of N- and O- linked glycans. Since 2005, when lectin microarray technology with an evanescent-field fluorescence-detection was reported, this sensitive and high-through put technology has been extensively used for cancer cells (Matsuda et al. 2008, 2010; Kuno et al. 2010; Narimatsu et al. 2010; Fry et al. 2011). In addition, this technology has recently applied to pluripotent and somatic stem cell types based on differentiation stage and potential (Tateno et al. 2011; Toyoda et al. 2011).

Nonreducing terminal fucose in endometrial cancer

Endometrial cancer tissue comprises well-differentiated cells and poorly differentiated cells, and the tissues are graded as G1 to G3 by histological observation. Among lectins on the lectin microarray chip, specific lectins that recognized endometrial cancer were determined, which include AOL, TJA-II, AAL, Jacalin, and ConA. AOL, AAL and TJA-II recognize nonreducing terminal fucose. Fucose is a component of many different classes of glycans, including N-glycans, mucin-type O-glycans, and can be directly linked to the hydroxyl group of serine or threonine residues as O-linked fucose. Fucosylation is typically found as a terminal modification of N- or O-glycans, is observed in colon and pancreatic cancer, and is used as a tumor marker (Haltiwanger 2009; Osumi et al. 2009; Miyoshi et al. 2010). In addition to these cancers, nonreducing terminal fucose can be used as a marker for endometrial cancer, which can include core fucosylation or O-fucosylation.

Lectin reactivities in endometrial cancer grading

Endometrial cancer grading is well correlated with prognosis, and thus the association between reactivity levels of six lectins (SSA, SNA, TJA-I, ACA, BPL, and UEA-I) and grading of endometrial cancer that we found is quite interesting. Among the lectins that correlated with the grading, SSA, SNA, and TJA-I recognize terminal sialic acid (Siaα2-6Gal/GalNAc). These lectin specificities include the α2-6-sialylated lactosamine as well as the sialic acid α2-6-linked to N-acetylgalactosamine. The lectin-reactive carbohydrates are products that are generated by the specific enzyme, β-galactoside α2-6-sialyltransferase (ST6Gal.I). This enzyme is one of the principal enzymes responsible for the addition of α-2-6-linked sialic acids to the Galβ1-4GlcNAc disaccharide (Dall'Olio 2000) and is overexpressed in many types of human cancers, including colon (Dall'Olio et al. 1989; Sata et al. 1991; Gessner et al. 1993; Lise et al. 2000; Petretti et al. 2000), breast (Recchi et al. 1998), ovarian (Wang et al. 2005), gastric (Gretschel et al. 2003), oral (Shah et al. 2008), cervical (Lopez-Morales et al. 2009), choriocarcinoma (Fukushima et al. 1998), leukemia (Mondal et al. 2010), and brain tumors (Kaneko et al. 1996). Sialic acid with high expression of ST6Gal.I positively correlates with tumor invasiveness, metastasis, differentiation state, and poor prognosis (Bresalier et al. 1990; Harvey et al. 1992; Le Marer & Stehelin 1995; Recchi et al. 1998; Lise et al. 2000; Zhu et al. 2001; Lin et al. 2002; Seales et al. 2005; Christie et al. 2008; Hedlund et al. 2008; Shah et al. 2008; Shaikh et al. 2008). Terminal sialic acid can also be used to predict endometrial cancer grading using the lectin microarray, especially signals of SSA, SNA, and TJA-I.

The lectin binding described above allowed us to successfully distinguish Grade 1 and Grade 3 endometrial cancer based on data obtained using endometrial cancer cell lines. However, in endometrial cancer tissues obtained directly from patients, pathological grading is sometimes difficult, especially Grade 2. Detailed grading is important for future therapeutic treatment, and glycan profiling by lectin microarray may allow for the grading without conventional histology.

Oligosaccharide structure in differential staging

Clinical staging is important issue in addition to grading because therapeutic approach is dependent on determination of a stage. Three lectins, ACA, BPL, and DBA, were identified as stage-specific lectins: ACA and BPL recognize Galβ1-3GalNAc; DBA recognizes α-linked GalNAc. Characteristic expression of glycolipids is reported to be associated with prognosis in the ovarian cancer cells used in this study (Kiguchi et al. 2006). Causal sequence between change in carbohydrate composition and cancer staging remains unclear, but the tight correlation between lectin reactivities and endometrial cancer staging may illustrate a novel approach toward determination of therapeutic regimen and prediction of prognosis. Indeed, DBA was also identified as one of the chemosensitivity-specific lectins for ovarian cancer (Fig. 5). Using stage- and chemoresistance-specific lectins, other technologies such as immunohistochemistry will be useful to estimate ratio of lectin-positive and lectin-negative area because lectin microarray technology is employed using cell membrane extract of specimens.

Gene expression, epigenetics, and genome have been comprehensively analyzed in cancer. In this study, sugar structure was analyzed in gynecological cancer. Flow cytometry and immunochemistry have been commonly used to analyze sugar structure in endometrial cancer. In addition to these approaches, lectin microarray has emerged as a cell profiler to discriminate glycan profiling. Owing to its extremely high sensitivity and accuracy, the lectin microarray system is launched to characterize cancer, and it is expected to be applicable to select specific lectins of other cancers.

Experimental procedures

Cells and tissues

Two cell lines established from human endometrial cancers, HEC-6 and HEC-50B, are available through the Japanese Collection of Research Bioresources-Cell Bank. HEC-6 is well-differentiated cancer (G1: Grade 1), and HEC-50B is poorly differentiated cancer (G3: Grade 3). UtE1104 prepared from endometrium was the mesenchymal cells (Toyoda et al. 2011). Endometrial cancer tissues of the poorly differentiated type (G3) were collected from extirpated specimens and classified into Stage I–IV. When the specimen was taken in the operating room, the cancerous portion was excised, put in an Eppendorf tube and stored at −80 °C until use. The Ethics Committee of Tokai University Hospital specifically approved this study. Signed informed consent was obtained from the patients, and the surgical specimens were de-identified. All experiments that involved the handling human cells and tissues were performed in line with the Tenets of the Declaration of Helsinki.

The cell lines, KFr13, KFr13TX, KF28 and KF28TX, were kindly donated by Prof. Y. Kikuchi (National Defense Medical College, Saitama, Japan). KF28, a human ovarian cancer cell line, was established from serous carcinoma. A CD-resistant cell line, KFr13, and a taxol-resistant line, KF28TX, were derived from the parental cell line, KF28. A paclitaxel- and CD-resistant line, KFr13TX was established from KFr13 (Yamamoto et al. 2000; Iwamori et al. 2007). Serous carcinoma lines, C13 and 2008/PX24, were CD-resistant and taxol-resistant lines, respectively. The cells were cultured in Dulbecco-modified MEM medium supplemented with 10% FCS, 100 μU/mL penicillin and 0.1 μg/mL streptomycin, in a humidified incubator at 37 °C under a 5% CO2 atmosphere.

Sample preparation and lectin microarray analysis

The cells (0.1 − 1 × 106) and the collected tissues from extirpated specimens were washed with PBS, and the tissues were then cut into small pieces (3 mm × 3 mm). Lectin microarray analysis was performed as previously described (Kuno et al. 2005, 2008). Briefly, the membrane fractions were prepared using a CelLytic MEM Protein Extraction kit (Sigma, St. Louis, MO, USA), and a small aliquot of the fraction (200 ng as protein) was labeled with Cy3-succinimidyl ester (Amersham Biosciences) in 0.1 m sodium bicarbonate buffer (pH 9.3) for 1 h. Cy3-labeled glycoprotein thus obtained were then subjected to the lectin chip (LecChip; GP BioSciences, Yokohama, Japan) with 45 lectins for human cells. One hundred microliters of Cy3-labeled glycoprotein solution in probing buffer (TBS containing 0.05% Triton X-100) were applied to each well with immobilized lectins, at a concentration of 0.25 and 0.5 μg/mL. Incubation was performed at 4 °C, until the binding reached equilibrium. After the incubation, we acquired a fluorescence image of the array using an evanescent-field fluorescence scanner, GlycoStation Reader 1200, SC-Profiler (GP BioScience). We calculated the net intensity value for each spot by subtracting a background value from signal intensity and then averaged the signal net intensity values of three spots. Lectin microarray data on each cell type and tissue were processed by the microarray system using a max-normalization procedure after a gain-merging process (Kuno et al. 2008). All of the data were analyzed with the Array Pro analyzer version 4.5 (Media Cybernetics, Inc.).

Hierarchical clustering analysis and principal component analysis

To analyze the lectin microarray data, we used agglomerative HCA and PCA (Sharov et al. 2005). HCA classifies data by similarity and their results are represented by dendrograms. PCA is a multivariate analysis technique that finds major patterns in data variability.

Acknowledgements

We express our sincere thanks to M. Yamada for fruitful discussion, Y. Takahashi and M. Harasawa for providing expert technical assistance, to C. Ketcham for reviewing the manuscript and K. Saito and Y. Suehiro for secretarial work. This research was supported by grants from the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan; by Ministry of Health, Labour and Welfare Sciences (MHLW) research grants; by a Research Grant on Health Science focusing on Drug Innovation from the Japan Health Science Foundation; by the program for the promotion of Fundamental Studies in Health Science of the Pharmaceuticals and Medical Devices Agency; by the Grant of National Center for Child Health and Development to A.U. This work was also supported in part by Grants-in-aid for scientific research from the Ministry of Education, Culture, Sports, Science, and Technology, Japan (No. 23592465), and Tokai University Research Aid to M.M.

Ancillary