Glycans exhibit characteristic changes in their structures during development and thus have been used as markers for stem/progenitor cells. However, the glycan structures unique to cancer stem cells (CSC) remain unknown. In the present study, we examined glycan structures in CD133+CD13+ CSC, which were recently found to have a high CSC ability, by means of a lectin microarray. Seven sialylated glycan-recognizing lectins, MAL-I, SNA, SSA, TJA-I, ACG, ABA and MAH, showed higher affinity to CD133+CD13+ CSC than CD133+ cells with a lower CSC ability. In addition, we demonstrated that CD133+SSA+ cells isolated from Huh7 cells had a significantly higher ability to form tumors in non-obese diabetic/severe combined immunodeficiency disease (NOD/SCID) mice and spheres under serum-free conditions than CD133+SSA− cells. These results suggest that hepatic CSC highly express sialylated glycans and that SSA lectin can be used as a tool for isolating CSC. This study is the first report to demonstrate the characteristic glycan structures in CSC and to indicate a new methodology involving lectins for isolating CSC. (Cancer Sci 2011; 102: 1164–1170)
A growing body of evidence has suggested that tumors are frequently composed of heterogeneous cell types, and that tumor initiation and growth are driven by a small subset of cells, termed cancer stem cells (CSC) or tumor-initiating cells.(1,2) Cancer stem cells can self-renew and also give rise to more differentiated progeny that comprise the bulk of a tumor.(3) Furthermore, several lines of research have indicated that CSC can be preferentially resistant to many current therapies, including various chemotherapeutic agents and radiation treatment.(4,5) Thus, therapeutic strategies that effectively target CSC could have a major impact on the survival of cancer patients. Evidence of tumor heterogeneity and CSC was first obtained from acute myeloid leukemia(6) and more recently extended to several human solid tumors, for example, breast,(7) brain,(8,9) prostate,(10) colon(11–13) and pancreatic(14) cancers. Over the last decade, a large body of literature has implicated CD133 as a bona fide marker for CSC in some cancers.(8,11,12) However, the validity of CD133 has been a matter of debate since recent studies showed the limitation of isolating CSC using only CD133 as a CSC marker.(15) We reported that the CD133+CD44+ population in colon cancer exhibited higher tumorigenicity than the CD133+CD44− one,(16) suggesting that it would be much better to use CD133 in combination with other markers to identify CSC. Under these circumstances, we recently found that CD13 was a novel marker for hepatic CSC. The CD133+CD13+ population in hepatoma cell lines and hepatocellular carcinoma (HCC) tissues exhibits higher tumorigenic and self-renewal properties than the CD133+CD13− one, indicating that HCC-initiating cells are highly enriched in the CD133+CD13+ population.(17)
Glycans often become attached to proteins and lipids on the cell surface, and then structurally and functionally modify these molecules. Glycans consist of several kinds of monosaccharides and show great structural diversity. Research in the field of glycobiology has revealed diverse and complex biological roles for these glycans.(18) The structures and amounts of glycan present on the cell surface dramatically change during development and differentiation.(19) Some of the glycan structures are specific to undifferentiated embryonic stem (ES) cells and thus can be used as markers for these cells. Stage-specific embryonic antigen-1 (SSEA-1) is known to be a marker for mouse ES cells.(20) In addition, human ES cells express SSEA-3, SSEA-4, TRA-1-60 antigen and TRA-1-81 antigen, which can be used as markers for human ES cells.(21) However, the typical glycan structures in CSC remain to be elucidated.
Sialic acids comprise one of the building blocks of glycans, and are generally found at the outermost ends of the glycan chains of glycoproteins and glycolipids. Thus, sialic acids are associated with many physiological and pathological events, including binding with infectious pathogens, regulation of immune responses and tumor malignancy.(22) In particular, alteration of sialic acids is associated with cancer cell behavior, such as invasiveness and metastasis.(23–25)
In the present study, we investigated the characteristic glycan structures in CD133+CD13+ CSC, which have a high tumor-initiating property, isolated from a human hepatoma cell line, Huh7 cells. In addition, the availability of glycans on the cell surface as possible markers for CSC was examined.
Materials and Methods
Cell culture. Human liver cancer cell line Huh7 was purchased from American Type Culture Collection (ATCC, Manassas, VA, USA) and cultured in RPMI 1640 (Sigma, St. Louis, MO, USA) medium containing 10% FCS (Invitrogen, Carlsbad, CA, USA), supplemented with 100 units/mL penicillin G and 100 mg/mL streptomycin, followed by incubation at 37°C under a humidified atmosphere containing 5% CO2.
Flow cytometric analysis and fluorescence-activated cell sorting (FACS). For surface marker analysis with a flow cytometer, confluent cells in 10-cm dishes were washed once with phosphate-buffered saline (PBS) and then dissociated using PBS containing 1 mM EDTA. After centrifugation, 1 × 106 cells were suspended in 100 μL PBS containing 0.1% BSA, followed by incubation for 40 min at 4°C with the following antibodies: allophycocyanin (APC)-conjugated anti-human CD133 (eBioscience, San Diego, CA, USA), phycoerythrin (PE)-conjugated anti-human CD13 (BD Biosciences, San Jose, CA, USA), and FITC-conjugated SSA lectin (Seikagaku Corp., Tokyo, Japan), which recognizes α2,6-sialylated glycans. Isotype-matched mouse IgG was used as a control (BD Biosciences). Doublet cells were eliminated using FSC-A/FSC-H and SSC-A/SSC-H. Dead cells were eliminated with 7-amino-actinomycin D (BD Biosciences). For FACS cell sorting, 5–10 × 106 cells were stained as described above and then sorted using FACS BD Aria II (BD Biosciences).
Lectin microarray. The total patterns of oligosaccharide structures in CD133+CD13− and CD133+CD13+ cells were investigated by means of evanescent-field fluorescence-assisted lectin microarray. Forty-five kinds of lectin were immobilized on a glass slide in triplicate. The procedure was described in detail in a previous report.(26) Briefly, cellular proteins in PBS containing 1% Triton X-100 were labeled with Cy3-succimidyl ester (GE Healthcare, Chalfont St. Giles, UK) at room temperature for 1 h in the dark. Excess reagent was removed by gel filtration chromatography. The resultant Cy3-labeled protein solution (100 μL, 250 ng/mL or 2 μg/mL) was applied to a lectin microarray. After incubation at 20°C for 15 h, the glass slide was scanned with an evanescent-field fluorescence scanner, GlycoStation (GP Biosciences Ltd, Kanagawa, Japan). All of the data were analyzed with array pro analyzer version 4.5 (Media Cybernetics, Inc., Bethesda, MD, USA). The net intensity value for each spot was calculated by subtracting the background value. The signal intensity value for each lectin was expressed as the average of the net intensity values for three spots. The WGA signals were used to normalize the signal intensity of each lectin, because binding to WGA lectin was relatively stable and almost the same in many kinds of cells.
Multiple quantitative PCR array for glycogenes. Multiple quantitative PCR array was performed as previously described.(27,28) Briefly, total RNA was extracted from CD133+CD13− and CD133+CD13+ cells. cDNA was synthesized from 1 μg of total RNA and then subjected to a multiple quantitative PCR array carrying 186 glycogenes and three housekeeping genes. Absolute copy numbers of transcripts in 7.5 ng of total RNA were plotted on a log scale. Genes with a number of transcripts smaller than 100 were omitted from our results due to quite low expression.
In vivo tumorigenicity analysis. Various numbers of cells, ranging from 100 to 5000, were prepared and suspended in 200 μL of RPMI1640 medium and Matrigel (BD Biosciences) mixture (1:1 volume) for injection. These cells were subcutaneously injected into the abdominal and dorsal regions of 6-week-old female non-obese diabetic/severe combined immunodeficiency disease (NOD/SCID) mice (Charles River, Wilmington, MA, USA). Tumor formation was monitored weekly for up to 30 days. An independent special pathologist confirmed tumor formation by examining the histology of HE-stained tissues. The data from two independent experiments are summarized in Table 1.
Table 1. The tumor-initiating ability of CD133+CD13−SSA− and CD133+CD13−SSA+ cells among Huh7 cells
1 × 102 cells
5 × 102 cells
1 × 103 cells
5 × 103 cells
Sphere formation assay. Cells were seeded on non-coated 6-cm culture dishes (AGC Techno Glass, Chiba, Japan) at 4 × 104 cells per dish in serum-free DMEM/F-12 medium (Invitrogen) containing 20 ng/mL epithelial growth factor (Sigma-Aldrich, St. Louis, MO, USA), 10 ng/mL fibroblast growth factor-2 (Collaborative Biomedical Products, Bedford, MA, USA), 0.4% BSA, B27 NeuroMix (PAA Laboratories, Pasching, Austria), 100 units/mL penicillin G and 100 mg/mL streptomycin. B27 NeuroMix contains vitamins, hormones and other growth factors including insulin, transferrin, catalase, antioxidants and fatty acids. The culture medium was changed every 3 days. After 10 days, each dish was examined under a light microscope and spheroid colonies were measured. The ratio of the area of each spheroid colony to the visual field was calculated. Data are represented as the means of two independent experiments and relative values on the basis of the proportion of CD133+CD13−SSA− cells.
Glycan profiling of CD133+CD13+ CSC from Huh7 cells using a lectin microarray. We recently reported that CD133+CD13+ cells exhibit the characteristics of CSC including self-renewal and high tumorigenic abilities.(17) To determine glycan structures unique to hepatic CSC, CD133+CD13+ cells isolated from Huh7 cells were used as model cells of CSC. CD133+CD13+ and CD133+ CD13− cells were isolated from Huh7 cells using FACS, and then 25 ng of Cy3-labeled proteins derived from these cells was subjected to lectin microarray analysis (Fig. 1A). Interestingly, the intensities of five sialylated glycan-recognizing lectins (SNA, SSA, TJA-1, ACG and ABA) among 45 lectins were significantly higher in CD133+CD13+ cells than CD133+CD13− ones. The intensities of several lectins were under the detection limit in the used assay conditions. Thus, next we performed the same experiment using 200 ng of the Cy3-labeled proteins. As shown in Figure 1B, MAL-I and MAH, which recognize sialylated glycans, exhibited higher affinity to CD133+CD13+ cells than CD133+CD13− ones. The intensities of seven sialylated glycan-recognizing lectins are summarized in Figure 2. The results indicate that sialylated glycans are structures unique to CD133+CD13+ CSC.
Expression profile of glycogenes in CD133+CD13− and CD133+CD13+ cells isolated from Huh7. Next, we searched for genes responsible for the enhanced expression of sialylated glycans in CD133+CD13+ CSC. The human genome contains approximately several hundred glycogenes. We previously established multiplex quantitative PCR (qPCR) array format for comprehensive profiling of an expression pattern of glycogenes. The qPCR array consists of probes and primer sets for measuring mRNA of 186 glycogenes, and enables the determination of expression profiles for these glycogenes in a single assay. An expression profile of 186 glycogenes were analyzed and compared between CD133+CD13− and CD133+CD13+ cells (Table S1). As shown in Figure 3, eight genes were differentially expressed more than twofold between both cells. Seven genes, HS6ST1, ST6GALNAC1, B4GALT5, GALNT13, GCNT3, GCNT4 and B4GALNT2, were expressed at an elevated level in CD133+CD13+ cells. In contrast, only DPAGT1 is highly expressed in CD133+CD13− cells. Among the eight genes, ST6GALNAC1 exhibited the biggest difference (4.385-fold higher in CD133+CD13+ cells) and was the only glycogene involved in sialylation. Thus, this result suggests that ST6GALNAC1 is a gene responsible for high expression of sialylated glycans in CD133+CD13+ cells.
CD133+CD13+ CSC were present in the SSA-positive population. To confirm the sialylation level in CD133+CD13+ CSC, we performed flow cytometric analysis using anti-CD133 and CD13 antibodies, and several sialylated glycan-recognizing lectin. As previously reported, CD133+CD13+ CSC comprised a very minor population of Huh7 cells (Fig. 4A). These cells were invariably present in the SSA-positive population (Fig. 4B). Other lectins showed almost equal binding to both CD133+CD13− and CD133+CD13+ cells (data not shown). The SSA lectin recognizes α2,6-sialylated glycans. These results confirmed that CD133+CD13+ CSC highly express α2,6-sialylated glycans recognized by SSA lectin on their surface.
CD133+CD13−SSA+ cells efficiently generated tumors in NOD/SCID mice. Next, we investigated whether SSA lectin can be used for the isolation of hepatic CSC. As shown in Figure 4B, the CD133+SSA+ population, but not the CD133+SSA− one, definitely includes CD133+CD13+ CSC. Because CD133+CD13+ CSC have already been revealed to exhibit strong CSC ability,(17) the CD133+CD13− population was divided into SSA-positive and negative cells to eliminate an effect of CD133+CD13+ CSC. CD133+CD13−SSA+ and CD133+CD13−SSA− cells were isolated from Huh7 cells (Fig. 5A), and then various numbers of isolated cells, ranging from 100 to 5000, were subcutaneously inoculated into NOD/SCID mice. A significant difference in tumorigenicity was observed between these two subpopulations (Fig. 5B). Tumor formation was observed in four of eight mice on injection of as few as 1 × 102 CD133+CD13−SSA+ cells. In contrast, only one mouse injected with 1 × 102 CD133+CD13−SSA− cells showed generation of a tumor. In addition, the injection of CD133+CD13−SSA+ cells at a higher dose (dose range: 5 × 102–5 × 103) generated a much larger tumor than that of the CD133+CD13−SSA− ones. The data are summarized in Table 1. These results indicate that CD133+CD13−SSA+ cells have a significantly higher tumor-initiating property than the CD133+CD13−SSA− ones.
CD133+CD13−SSA+ cells efficiently formed sphere in vitro. Because numerous previous studies demonstrated that the formation of spheres in serum-free medium is one of the characteristics of CSC, CD133+CD13−SSA+ and CD133+CD13−SSA− cells were cultured under stem cell-selective conditions. CD133+CD13−SSA+ cells had aggregated into floating spheroid clusters after 10 days when the cells were cultured on non-coated dishes in serum-free Ham F12 supplemented with EGF, FGF-2 and B27 NeuroMix (Fig. 6A). In contrast, we observed little formation of spheres by CD133+CD13−SSA− cells cultured under the same conditions. The quantitative analysis showed a significant increase of sphere formation in CD133+CD13−SSA+ cells compared with the CD133+CD13−SSA− ones (Fig. 6B). This finding demonstrates that CD133+CD13−SSA+ cells have a higher CSC property than the CD133+CD13−SSA− ones.
In order to identify the characteristic glycan structures in CSC, we used the lectin microarray system, which is an emerging technique for the profiling of entire cellular glycan structures. Due to its extremely high sensitivity, this method is the best tool for analyzing glycan structures in a small number of cells, like CSC. In the present study, a lectin microarray showed that all of seven sialic acid-recognizing lectins, MAL-I, SNA, SSA, TJA-1, ACG, ABA and MAH, more strongly bound to CD133+CD13+ CSC from Huh7 cells than CD133+CD13− cells. A similar result was obtained in another hepatoma cell line, Hep3B (Figs S1,S2). MAL-I and MAH recognize α2,3-sialic acid residues. ACG exhibits a higher affinity to α2,3-sialylated and high branched N-glycans than non-sialylated ones. ABA exhibits an affinity to sialylated T-antigen (NeuAcα2,3Galβ1,3GalNAc). In contrast, SNA, SSA and TJA-1 recognize α2,6-sialic acid residues. These findings strongly suggest that sialylated glycans are the characteristic structures of CSC. In future studies, it would be required to reveal the structures of the sialylated glycans expressed on CD133+CD13+ CSC in detail. Moreover, in the present study, it was revealed that SSA lectin could be used as an additional marker for isolating CSC. Future studies should also examine whether other lectins, such as α2,3-sialic acid-recognizing ones, could be possible markers for isolating CSC.
When a difference of more than twofold was defined to be significant, the lectin microarray also exhibited a high affinity of TxLC-I, ACA, GSL-I-A4 to CD133+CD13+ CSC. TxLC-I recognizes a relatively broad range of structures, including galactosylated and agalactosylated N-glycan. ACA also recognizes agalactosyl N-glycan. Interestingly, TxLC-I and ACA can also bind to sialylated glycans, especially α2,6 sialic acid and sTn antigen, respectively. Thus, a high affinity of both lectins might be caused by the increased expression of sialylated glycans in CD133+CD13+ CSC. GSL-I-A4 recognizes α-linked GalNAc on O-glycan. While the meaning of high affinity of GSL-I-A4 to CD133+CD13+ CSC remains unclear, the change of sialylated glycans was the most common and drastic in CD133+CD13+ CSC.
Glycosylation is regulated by the complicated mechanisms involving many kinds of glycogenes, such as glycosyltraneferases, sugar nucleotide-synthetic enzymes, and sugar nucleotide transporters. Thus, comprehensive analysis is required for a detailed understanding of the regulatory mechanisms of glycosylation. A DNA microarray has been used as a comprehensive transcription level analysis. However, it is difficult to detect expressions of most glycogenes by a DNA microarray due to their low expressions. Therefore, in the present study, multiple qPCR array analysis that we previously established was performed to search genes responsible for increased expression of sialylated glycans in CD133+CD13+ CSC. ST6GALNAC1 stands for α-N-acetyl galactosaminide α2,6-sialyltransferase 1 and transfers a sialic acid via an α2,6 linkage to O-linked GalNAc residue. The cancer-associated sialyl-Tn (sTn; Neu5Acα2,6GalNAc-O-Ser/Thr) antigen is formed by ST6GALNAC1-catalyzed sialylation of GalNAc residues on mucins.(29) Several investigators have reported that expression of the sTn antigen is a poor prognostic factor in patients with adenocarcinoma of the stomach,(30–32) ductal cell carcinoma of the mammary glands(33,34) and epithelial ovarian cancer.(35) Thus, sTn antigen synthesized by ST6GALNAC1 might be associated with the characteristics of CSC.
The lectin microarray showed high affinity of CD133+CD13+ CSC to both α2,3- and α2,6-sialic acid-recognizing lectins. The increased binding to α2,3-sialic acid-recognizing lectins is not explained by the high expression of ST6GALNAC1. CMP-sialic acid is involved in all types of sialylation as a donor substrate for all sialyltransferases. Because genes of synthetic enzymes of CMP-sialic acid are not carried on the array, these genes and their product, CMP-sialic acid, might be responsible for the increased α2,3-sialylation. In the case of fucosylated glycans, synthesis and transport of GDP-fucose, a donor substrate for all fucosyltransferases, mainly regulates cellular fucosylation in HCC.(36,37)
Several previous studies have shown that sialic acids were associated with normal stem cells. Satomaa et al.(38) reported that the surface of human ES cells were clearly labeled by MAA lectin, which recognizes structures containing α2,3-linked sialic acids, preferably α2,3-sialylated N-acetyllactosamine, indicating that such sialylated glycans were abundant on the surface of human ES cells. In addition, the heavily glycosylated sialomucin molecule CD34 is known to be expressed on virtually all hematopoietic precursor cells, including multipotent stem cells.(39) Moreover, podocalyxin, which is a sialoglycoprotein structurally related to CD34, is highly expressed on embryonic stem cells,(40) and has been suggested to be a cell surface marker for hemangioblasts, the common precursors of hematopoietic and endothelial cells.(41) Because normal stem cells and CSC are considered to share similar self-renewal programs, these data suggest the involvement of sialylated glycans in the stem/progenitor cell property of CSC. It has remained unclear how sialylated glycans regulate CSC’s ability. Identification of molecules carrying sialylated glycans in CD133+CD13+ CSC would facilitate the understanding of the function of sialylated glycans in the characteristics of CSC.
In the present study, we demonstrated that CD133+CD13−SSA+ cells from liver cancer cell lines, Huh7 and Hep3B, exhibited high CSC ability compared with CD133+CD13−SSA− cells (Figs 5,6 and Table S2). These results indicate that CD133+CD13−SSA+ cells as well as CD133+CD13+ cells have sufficient ability to give rise to new tumors. These results suggest the existence of CD13-negative CSC among liver cancer. Recently, several studies revealed that EpCAM+,(42) CD133+CD44+(43) and BMI1+ cells(44) are hepatic tumor-initiating cells with stem/progenitor cell features, and that these molecules can be used as CSC markers to isolate CSC from HCC cell lines.(45) Thus, the CD133+CD13−SSA+ population might include these other types of hepatic CSC. Because CD13+ cells were definitely enriched in the SSA-positive population, as shown in Figure 3, the CD133+SSA+ population includes both CD133+CD13+ CSC and other types of CSC such as EpCAM+, CD133+CD44+ and/or BMI1+ cells. Therefore, CD133+SSA+ cells might be a more suitable and attractive target for the development of more effective therapies for HCC. In addition, this finding indicates that glycan-recognizing lectins are valuable tools for isolating small subsets of cells, such as stem cells. Actually, we previously reported that the isolation of hepatic progenitor cells from liver tissues of rats with fulminant hepatitis was achieved by the use of E4PHA lectin.(46) Thus, SSA lectin would be a useful tool for isolating CSC from tumor tissues.
The present study was supported by the innovation promotion program of the New Energy and Industrial Technology Development Organization and the Global-Center of Excellence (G-COE) Program of Osaka University School of Medicine.
Shunsaku Takeishi is an employee of GP Bioscience Ltd. Other authors have no conflict of interest.