Tumour suppressor p16INK4a – anoikis-favouring decrease in N/O-glycan/cell surface sialylation by down-regulation of enzymes in sialic acid biosynthesis in tandem in a pancreatic carcinoma model


  • Maho Amano,

    1. Field of Drug Discovery Research, Faculty of Advanced Life Science, Graduate School of Life Sciences, Hokkaido University, Sapporo, Japan
    Search for more papers by this author
    • These authors contributed equally to this work
  • Hanna Eriksson,

    1. Cancer Proteomics Mass Spectrometry, Science for Life Laboratory, Department of Molecular Medicine and Surgery, Karolinska Institutet, Stockholm, Sweden
    Search for more papers by this author
    • These authors contributed equally to this work
  • Joachim C. Manning,

    1. Institute of Physiological Chemistry, Faculty of Veterinary Medicine, Ludwig-Maximilians-University, Munich, Germany
    Search for more papers by this author
  • Katharina M. Detjen,

    1. Medizinische Klinik mit Schwerpunkt Hepatologie und Gastroenterologie, Charité Campus Virchow Klinikum, Berlin, Germany
    Search for more papers by this author
  • Sabine André,

    1. Institute of Physiological Chemistry, Faculty of Veterinary Medicine, Ludwig-Maximilians-University, Munich, Germany
    Search for more papers by this author
  • Shin-Ichiro Nishimura,

    Corresponding author
    • Field of Drug Discovery Research, Faculty of Advanced Life Science, Graduate School of Life Sciences, Hokkaido University, Sapporo, Japan
    Search for more papers by this author
  • Janne Lehtiö,

    Corresponding author
    1. Cancer Proteomics Mass Spectrometry, Science for Life Laboratory, Department of Oncology–Pathology, Karolinska Institutet, Stockholm, Sweden
    • Field of Drug Discovery Research, Faculty of Advanced Life Science, Graduate School of Life Sciences, Hokkaido University, Sapporo, Japan
    Search for more papers by this author
  • Hans-Joachim Gabius

    Corresponding author
    1. Institute of Physiological Chemistry, Faculty of Veterinary Medicine, Ludwig-Maximilians-University, Munich, Germany
    • Field of Drug Discovery Research, Faculty of Advanced Life Science, Graduate School of Life Sciences, Hokkaido University, Sapporo, Japan
    Search for more papers by this author


S.-I. Nishimura, J. Lehtiö or H.-J. Gabius, Field of Drug Discovery Research, Faculty of Advanced Life Science, Graduate School of Life Sciences, Hokkaido University, Kita-21, Nishi-11, Kita-ku, Sapporo, Hokkaido 001-0021, Japan;

SciLifeLab, Box 1031, 17121 Solna, (visiting address: Tomtebodavägen 23A); Sweden;

Institute of Physiological Chemistry, Faculty of Veterinary Medicine, Ludwig-Maximilians-University, Veterinärstr. 13, 80539 Munich, Germany

Fax: +81-11-7069042; +46-8-5252481425; +49-89-2180992290

Tel: +81-11-7069043; +46-8-52481416; +49-89-21803791

E-mail: shin@sci.hokudai.ac.jp; janne.lehtio@ki.se; gabius@tiph.vetmed.uni-muenchen.de or gabius@lectins.de


Tumour suppressor p16INK4a is known to exert cell-cycle control via cyclin-dependent kinases. An emerging aspect of its functionality is the orchestrated modulation of N/O-glycosylation and galectin expression to induce anoikis in human Capan-1 pancreatic carcinoma cells. Using chemoselective N/O-glycan enrichment technology (glycoblotting) and product characterization, we first verified a substantial decrease in sialylation. Tests combining genetic (i.e. transfection with α2,6-sialyltransferase-specific cDNA) or metabolic (i.e. medium supplementation with N-acetylmannosamine to track down a bottleneck in sialic acid biosynthesis) engineering with cytofluorometric analysis of lectin binding indicated a role of limited substrate availability, especially for α2,6-sialylation, which switches off reactivity for anoikis-triggering homodimeric galectin-1. Quantitative MS analysis of protein level changes confirmed an enhanced galectin-1 presence along with an influence on glycosyltransferases (β1,4-galactosyltransferase-IV, α2,3-sialyltransferase-I) and detected p16INK4a-dependent down-regulation of two enzymes in the biosynthesis pathway for sialic acid [i.e. the bifunctional UDP-N-acetylglucosamine 2-epimerase/N-acetylmannosamine kinase (GNE) and N-acetylneuraminic acid 9-phosphate synthase] (P < 0.001). By contrast, quantitative assessment for the presence of nuclear CMP-N-acetylneuraminic acid synthase (which is responsible for providing the donor for enzymatic sialylation that also acts as feedback inhibitor of the epimerase activity of GNE) revealed a trend for an increase. Partial restoration of sialylation in GNE-transfected cells supports the implied role of sialic acid availability for the glycophenotype. Fittingly, the extent of anoikis was reduced in double-transfected (p16INK4a/GNE) cells. Thus, a second means of modulating cell reactivity to the growth effector galectin-1 is established in addition to the common route of altering α2,6-sialyltransferase expression: regulating enzymes of the pathway for sialic acid biosynthesis.


aminooxy-functionalized tryptophanyl arginine


Chinese hamster ovary


UDP-N-acetylglucosamine 2-epimerase/N-acetylmannosamine kinase


immobilized pH gradient




Maackia amurensis agglutinin




N-acetylneuraminic acid 9-phosphate synthase


peanut agglutinin


sialic acid


Sambucus nigra agglutinin


The cell surface not only presents proteins, but also the glycan chains of glycoconjugates as determinants for biorecognition. Because carbohydrates are ‘ideal for generating compact units with explicit informational properties’ [1], protein/lipid-presented glycans are versatile signals as a result of their unsurpassed structural diversity and the possibility for swift remodelling [2, 3]. The phenotyping of tumour cells, by structural glycan analysis or by using lectins as a tool, has provided evidence for a wealth of deviations from the normal profile, directing special interest to α2,6-sialylation of glycoproteins [4-7]. The immediate regulation of this feature by the ras oncogene (up-regulation) [8-10] and four tested tumour suppressors (down-regulation) [11, 12] emphasizes its potential for affecting malignancy-relevant properties. Fittingly, up-regulation of α2,6-sialylation in N-glycans has been correlated with tumour invasiveness and migration [12-16], a feature that appears to qualify elevated sialylation in distinct serum glycoproteins as a potential glycobiomarker for malignancy [17, 18]. Moving to the level of individual glycoproteins, above-normal sialylation of the β1-integrin subunit, along with its α5-partner in the fibronectin receptor (α5β1-integrin), modulates interactions with matrix constituents such as collagen I or fibronectin (in this case lowering binding), whereas hyposialylation improves the reactivity of the integrin to fibronectin [10, 15, 16, 19]. Thus, a detailed study of sialylation and its regulation in tumour cells is definitely warranted, with the described integrin being an especially promising candidate as a result of its critical role for cell-matrix contact and the onset of anoikis. In this respect, the up-regulation of the α5-integrin subunit by restitution of expression of the tumour suppressor p16INK4a in a model of pancreatic cancer (i.e. the Capan-1 cell line) and its documented significance for anoikis induction [20] has provided incentive for a thorough analysis of an assumed connection with glycobiological parameters. Of note with respect to its potential clinical relevance, the genetic background of the Capan-1 cells shares key aberrations with human pancreatic cancer by expressing mutated DPC14 and p53, as well as activated K-ras [20].

The ensuing work disclosed indications for a tumour-suppressor-dependent decrease in cell surface sialylation and a role of α5β1-integrin cross-linking by the homodimeric galectin-1 (an adhesion/growth-regulatory endogenous lectin [21]) in anoikis induction via the caspase-8 pathway [12, 22]. These data have shaped the concept that the concomitant modulation of lectin counter-receptor (i.e. suitable glycosylation on the protein with downstream effector function) to enhance the capacity for forming galectin-integrin complexes is a salient aspect of the growth-limiting activity of this tumour suppressor. Because galectin binding can be precluded by α2,6-sialylation [23-26], the presence of this branch-end structure appears to act as a switch with respect to caspase activation. Indeed, transfectants for α2,6-sialyltransferase of human colonocytes (SW48) or a murine T cell line (PhaR 2.1) illustrate the effectiveness of this glycan remodelling in protecting the cells from galectin-induced cell death [27, 28]. To further examine the assumed link between suppressor presence and sialylation in this pancreas cancer model, we first applied chemoselective N- and O-glycan enrichment technology to accomplish structural analysis [29-31] and then tested the impact of different modalities to treat cells [i.e. genetically, by transfecting transferase-specific cDNA, and metabolically, by offering an excess of a substrate used for sialic acid (SA) biosynthesis] to track down the cause(s) for reduced sialylation. Identification of control points, first by metabolic supplementation and then by focused proteome analysis using narrow-range peptide isoelectric focusing with LC-MS/MS analysis [32], was backed by characterizing i) the glycophenotype of a transfectant, in which the p16INK4a-induced deficit on this aspect of glycome generation was partially compensated, ii) its cell surface reactivity for (ga)lectins and iii) its anoikis rate.

Results and Discussion

Profiling of N- and O-glycans

Quantitative glycan analysis was performed on aliquots of detergent extracts from cell pellets considering the entire glycomic complexity, including free glycans. Cell cultures were routinely expanded when obtaining starting material to ensure pellet generation from the same cell batch that was controlled by flow cytofluorimetric analysis for constant lectin reactivity. After enzymatic release of glycans from glycoproteins and bead capture, SAs were chemically protected to enable simultaneous MS analysis of neutral and acidic glycans. A total of 80 glycans (68 N-glycans and 12 free oligosaccharides) was detected. A complete listing of glycan composition and picomol content, including the internal standard (peak 66; set at 30 pmol), is provided in Table S1.

The detailed profile of N-glycans (Fig. 1A), along with a summary presented in six categories to make overall changes readily detectable (Fig. 1B), reveals a clear reduction of sialylation. When looking at the groups of fully sialylated N-glycans, N-glycans with mixed-type termini [SA + N-acetyllactosamine (LacNAc)] and LacNAc-terminated structures, the relative abundance in the N-glycan population is shifted from fully and partially sialylated species to LacNAc-terminated structures (P < 0.01; Fig. 1B). Considering the grading in branch preference of α2,6-sialyltransferase, first acting on α1,3-antenna, and then the α1,6-linked substrate at the level of biantennary glycans [33-36], decreases for peaks 49 and 72 can involve α2,6-sialylation in hybrid-type or biantennary core-fucosylated N-glycans, and for peak 68 in biantennary N-glycan without core substitution (Fig. 1A). Looking at integrin glycosylation, the described complex-type N-glycans are found to be represented in the N-glycome of the human α5β1-integrin, at a relative quantity of 1.5% and 4.6% for the two α2,6-substituted disialylated oligosaccharides [37]. In turn, the amount of LacNAc-terminated biantennary structures (peaks 35 and 45) increased, as present in the glycome of the human integrin [37]. Previous monitoring of cell surface lectin binding that indicated a reduced reactivity of cell surfaces for Sambucus nigra agglutinin (specific for Neu5Acα2,6Gal/GalNAc and clustered Tn antigen, i.e. for distinct N- and O-glycans) [12] is in agreement with this result. The increase in the extent of cell surface binding of the T-antigen-specific peanut agglutinin (PNA) reported in [12] let us expect down-regulation of sialyl/disialyl forms of this core 1 disaccharide of O-glycans and/or of the core 2 extension, which will also impair PNA reactivity; for O-glycan nomenclature, see Patsos and Corfield [38]. To settle this issue, we analyzed the O-glycan profile using the same technology.

Figure 1.

Quantitative N-glycan analysis (A) and glycotype survey focused on the categories of fully sialylated and partially sialylated (SA + LacNAc) as well as SA-free LacNAc-terminated compounds (B) for tumour suppressor p16INK4a-expressing cells versus the mock control of wild-type cells. ‘Others’ refers to GlcNAc-terminated structures. The individual N-glycan structures depicted (using the glycomod tool and the GlycoSuiteDB database) showed significantly different expression levels for the two cell populations (P < 0.01) (A). The survey on the different categories (representative structures for each type are given) illustrates the shift in the N-glycan profile (B). Yellow circle, galactose; blue square, N-acetylglucosamine; green circle, mannose; red triangle, fucose; purple diamond, SA. P-values for Students' t-test are given along with asteriks for entering information with respect to the level of statistical significance (*P < 0.05, **P < 0.01).

The detected di- to oligosaccharides are listed in Table S2. As shown in Fig. 2, for the most marked changes (P < 0.001), T-antigen sialylation decreased significantly; conversely, the presence of a tetrasaccharide [core 2 structure as depicted in Fig. 2 and/or an extended (linear) core 1 structure] increased. These data also explain why cell staining with jacalin, a plant lectin that tolerates α2,3-sialylation of the T-antigen, is less sensitive to p16INK4a restitution than PNA [12], providing a structural basis for the patterns of staining. Tetrasaccharide sialylation also decreased (peaks 10 and 12), whereas fucosylation at this stage was characteristic for cells with the presence of p16INK4a (Table S2). Overall, these results confirm that presence of the tumour suppressor p16INK4a is effective with respect to altering the extent of N- and O-glycan sialylation.

Figure 2.

Expression levels of distinct mucin-type O-glycans [i.e. disialylated core 1, monosialylated core 1 and a core 2 (or extended core 1) structure] on tumour suppressor p16INK4a-expressing cells versus the mock control of wild-type cells (**P < 0.001, for additional details, see Fig. 1).

Because α2,6-sialylation of N-glycans is a key factor for switching off galectin reactivity, for mono- and disialylated biantennary N-glycans shown at the level of cells [26], its regulation can either work at the level of expression of the responsible enzyme (i.e. α2,6-sialyltransferase) and/or the availability of the substrate(s) (i.e. the activated donor CMP-N-acetylneuraminic acid). Although surface sialylation patterns are considered to be ‘typically insensitive to changes in metabolic flux’ (based on the concentration of CMP-N-acetylneuraminic acid being higher than the KM values of sialyltransferases) [39, 40], supplementation of culture medium with N-acetylmannosamine (ManNAc), a metabolic precursor of SA but not other sugars such as GlcNAc or Man, was shown to increase surface sialylation in human B-lymphoma cells (BJA-B) [41], bringing the rate-limiting enzymatic step in SA biosynthesis [i.e. epimerization of UDP-GlcNAc to UDP-ManNAc by the bifunctional UDP-GlcNAc 2-epimerase/ManNAc kinase (GNE)] into play. Moreover, such treatment impaired the binding of galectin-1 to cells of a lung cancer line (U1752) [42], which is of potential relevance in our system. Both GNE (e.g. in rat hepatomas or tumour cells from leukaemia/lymphoma lines) [41, 43, 44] and α2,6-sialyltransferase (e.g. in colon cancer or B cells) [45, 46] are regulatable, such that it is necessary to test both possibilities. That a microarray on glycosyltransferase gene expression had detected no difference in signal intensities for α2,6-sialyltransferase [12] does not exclude its involvement as a result of post-transcriptional events that preclude the prediction of a direct correlation between gene expression and sialylation [45, 46]. Enforced gene expression by transfection represents one approach for delineating any enzyme involvement (genetic route), whereas serum supplementation comprises the metabolic route for pinpointing a site of control. In both cases, lectins were used as sensors to pick up changes in cell surface sialylation.

Lectin profiling to detect increase in sialylation

Several series of transfections to compensate for any deficit in the availability of the active transcript of the α2,6-sialyltransferase-I gene were performed, followed by monitoring with labelled Sambucus nigra agglutinin (SNA). As shown in Fig. 3A for a comparison between mock and sense-transfected cells, no convincing evidence for a consistent and significant increase was obtained. As a positive control, we performed experiments with Chinese hamster ovary (CHO) cells, which lack expression for α2,6-sialyltransferase [47], in parallel. For this line, cases with conspicuous SNA reactivity, which was sensitive to enzymatic desialylation, were seen (Fig. 3B). Thus, lectin reactivity apparently did not critically depend on gene expression for this enzyme, although these experiments do not unequivocally exclude operation at all possible sites of regulation, including post-translational control.

Figure 3.

Semilogarithmic illustration of surface binding of biotinylated lectins to Capan-1 pancreatic carcinoma cells expressing the tumour suppressor p16INK4a (A, C–F) and CHO cell transfectants stably expressing α2,6-sialyltransferase (B). The shaded area represents the 0% value obtained by processing in the absence of biotinylated lectin; quantitative data on the percentage of positive cells and mean fluorescence intensity are shown as appropriate. (A) SNA-dependent staining (1 μg·mL−1) of mock-treated (top) and a clone of α2,6-sialyltransferase-expressing Capan-1 cells (bottom). (B) SNA-dependent staining (5 μg·mL−1) of CHO cells expressing α2,6-sialyltransferase without (black line) and after (dashed line) neuraminidase treatment. (C–F) Staining of the tumour suppressor-expressing cells with SNA (5 μg·mL−1; C), MAA-I (2 μg·mL−1; D), PNA (1 μg·mL−1; E) and human galectin-1 (20 μg·mL−1; F) after mock treatment (black line) compared to growth in medium supplemented with ManNAc (grey line).

We next tested for an impact of metabolic flux. By making a substrate for hexokinases available with ManNAc, the rate-limiting ‘bottleneck’ at the mentioned control enzyme for SA synthesis (GNE) can be bypassed [41, 48]. Consequently, the addition of sugars that enter the biosynthetic route to SA upstream of the ‘bottleneck’ (i.e. GlcNAc, glucosamine, glucose or mannose) did not affect sialylation status [41]. This assumption was verified by rescue with ManNAc in the mouse model for hereditary inclusion body myopathy [49, 50] and GNE-deficient stem cells [51]. The presence of this compound in the culture medium (and not other sugars upstream of the rate-limiting step) led to increased SNA binding to treated cells, although less so for the Maackia amurensis agglutinin (MAA-1), targeting α2,3-sialylated type II N-glycans (Fig. 3C,D). Cell surface presentation of unsubstituted (PNA-reactive) T-antigen was also markedly decreased (Fig. 3E), fully in line with the O-glycan analysis data shown in Fig. 2. Relevant for the increased anoikis susceptibility of the cells, we next tested for reactivity to galectin-1. This was shown to be lowered (Fig. 3F), suggesting a functional correlation (via galectin-1) with the reduction in the extent of anoikis observed after culturing cells with ManNAc [52]. Because microarray and quantitative PCR had indicated a p16INK4a-reduced activity of GNE but not α2,6-sialyltransferase [12, 52], we aimed to strengthen this case using proteomic analysis, focusing on glycoenzymes (glycosyltransferases and enzymes of SA biosynthesis) flanked by the internal controls with the p16 protein and galectin-1.

Proteome analysis focused on glycoenzymes

To demonstrate the sensitivity of the applied technique, we first determined the relative levels of tumour suppressor p16INK4a protein and galectin-1, as well as of two galactosyltransferases, with the latter belonging to the set of low-abundant proteins represented in a previous microarray analysis [12]. Experimentally, four replicates of each cell line were digested into peptides before labelling with isobaric tags for relative and absolute quantification (iTRAQ labelling). This processing facilitates up to eight different samples for comparison in a single run [53, 54]. Each peptide sample was mixed with the isobaric tag for saturation labelling, in which primary amines were the reactants, predominantly at the N-terminus of the peptides. This tag is cleaved off in the fragmentation step of the MS analysis, generating a reporter ion whose intensity corresponds to the abundance of that peptide derived from each sample. The iTRAQ-labelled samples were pooled, and the peptides were separated using in-gel isoelectric focusing on a narrow pH range (3.7–4.9). This peptide IEF has been shown to be compatible with iTRAQ-based quantification [55]. The given pH range is optimal for tryptic peptides because approximately 30% of human protein-derived peptides fall into this range, representing approximately 95% of the entire human proteome [32]. Thus, the complexity of the sample is dramatically reduced at the same time as maintaining high proteome coverage. The peptides were eluted from the focusing gel, generating 72 fractions. These fractions were further separated using reversed-phase LC, before conducting the Orbitrap MS analysis, where peptides were fragmented to obtain the amino acid sequence. Each sequence was then matched with the entries of a database of known protein sequences, and the source of the peptide(s) (i.e. the proteins, together with quantitative information) could be identified (a schematic workflow is provided in Fig. 4, peptide coverage in each case is shown in Fig. S1).

Figure 4.

Scheme of workflow in quantitative proteomic analysis applying narrow-range peptide isoelectric focusing combined with LC-MS/MS-based protein identification and quantification.

As expected based on western blotting [12, 20], the p16INK4a level was increased by 7.5-fold, and galectin-1 by 1.6-fold (Fig. 5A,B). Microarray and proteomic data for β1,4-galactosyltransferase-IV matched well with an increase of 4.21 for the microarray [12] and 5.2 for the proteomics (Fig. 5C). By contrast, a decreased level of β1,4-galactosyltransferase-I for the microarray (0.33) [12] was not seen for the proteomics (Fig. 5D), suggesting a similar case of multifactorial control as discussed above for α2,6-sialyltransferase. Accordingly, we also checked β1,3-N-acetylglucosaminyltransferases, which are involved in the synthesis of LacNAc repeats in concert with β1,4-galactosyltransferases. Among them, the type I enzyme demonstrated similar levels irrespective of tumour suppressor status, with the type III enzyme (acting on O-glycans to extend the core 1 structure) down-regulated to 0.23 (P = 0.0005).

Figure 5.

Protein level determination between mock controls and the tumour suppressor p16INK4a-expressing Capan-1 cells for (A) tumour suppressor protein (p16), (B) galectin-1 (Gal-1), (C) β1,4-galactosyltransferase-IV (β4GalT4), (D) β1,4-galactosyltransferase-I (β4GalT1), (E) GNE and (F) NANS. Relative intensities of quantitative signals for each protein based on multiple peptide determination are shown. Values are given as the mean ± SD along with asterisks for information with respect to the level of statistical significance (**P < 0.01, ***P < 0.001).

For GNE, with five peptides covered that map to isoforms-1 and -3 [56, 57], a drastic reduction of protein level was detected (Fig. 5E). Of particular note, the level of the enzyme following GNE in the pathway of SA biosynthesis (i.e. N-acetylneuraminic acid 9-phosphate synthase; NANS) [58, 59] was significantly lowered (Fig. 5F). By contrast, a trend for up-regulation was seen for the nuclear enzyme producing the activated CMP-SA conjugate needed for the sialyltransferase reactions (not shown). This sugar donor for sialyltransferases additionally serves as feedback inhibitor of the 2′-epimerase activity of GNE [60-62]. In rat liver/hepatomas, the latter two enzymatic activities were found at comparable levels, with GNE alone being the control point [43]. Considering the drastic decrease in sialylation of the T-antigen (Fig. 2), which is also reflected by the extent of the increase in reactivity to PNA (Fig. 2E) [12] and likely to galectins binding the disaccharide when clustered [63-65], we can add the discovery of a highly significant decrease (P < 0.001) for α2,3-sialyltransferase-I to approximately 0.6 of the mock control (not shown). Because the β1-integrin subunit contains potential O-glycosylation sites in its I-like domain, where the RGD peptide binds [66], the sialylation of O-glycans may have a bearing on fibronectin binding. To add support for the idea of a major role of SA biosynthesis in tumour suppressor-dependent glycan changes, we generated stable transfectants for GNE in p16INK4a-expressing cells.

Glycophenotyping of GNE transfectant

The selection of clones yielded a broad spectrum of reactivity with SNA used as sensor, indicating the effectiveness of manipulating the level of GNE for cell surface sialylation. An example with strong enhancement is shown in Fig. 6, together with the control by sialidase treatment. Using the Polyporus squamosus agglutinin, a fungal lectin highly specific for α2,6-sialylated N-glycans (without notable cross-reactivity to sialylated Tn-antigen) [67-69], the p16INK4a-dependent reduction in this cell surface parameter (Fig. 7A,B) and its partial reversion by GNE restitution could unequivocally be demonstrated (Fig. 7C,D). As expected, PNA reactivity was diminished as well (not shown). Of note, the presence of core 1 β1,3-galactosyltransferase was found by proteomics to be independent of tumour suppressor status, excluding its impact on T-antigen production (not shown), which is fully in line with the Tn-status measured by the Dolichos biflorus agglutinin [12]. These lectin-binding data were corroborated by glycan analysis. Both for the N- and O-glycan profiles, the elevated GNE expression led to increases in sialylation, with a reduction in the relative quantity of galectin-reactive LacNAc termini, although not yet reaching the level of the wild-type cells (Figs 8 and 9). For example, the degree of both disialylated biantennary N-glycans without/with core fucosylation (peaks 68 and 72) increased markedly (Fig. 8A), and this trend is obvious from the survey provided in Fig. 8B. Looking at the O-glycans, a similar resialylation occurred. The expression level of disialyl T-antigen in the transfectants was closer to the wild-type cells than that of sialyl T-antigen, and the occurrence of the core 1/2 tetrasaccharide was conspicuously low (Figs 2 and 9). Sialylation of this tetrasaccharide, however, was not affected (Table S2).

Figure 6.

Semilogarithmic illustration of surface binding of biotinylated SNA (2.5 μg·mL−1) to Capan-1 cells transfected with vectors carrying cDNA for the tumour suppressor p16INK4a and for GNE, respectively, without neuraminidase treatment (black line) and after enzymatic α2,3/6-desialylation (dotted line).

Figure 7.

Semilogarithmic illustration of surface binding of biotinylated fungal lectin (1 μg·mL−1), which is specific for α2,6-sialylation of type II N-glycans, for the mock control (A) and the tumour suppressor p16INK4a-expressing cells (B), as well as for the mock control of the double-transfected cells (C) and the cells transfected with vectors for p16INK4a- and GNE-specific cDNAs (D).

Figure 8.

Quantitative N-glycan analysis (A) and glycotype survey (B) for the tumour suppressor p16INK4a-expressing cells without and after transfection of GNE-specific cDNA (for additional details, see Fig. 1).

Figure 9.

Expression levels of mucin-type O-glycans selected for sensitivity to tumour suppressor presence in p16INK4a-restituted cells without and after transfection of GNE-specific cDNA (for additional details, see Fig. 2).

The known strongly negative impact of the α2,6-sialyl-determinant in biantennary N-glycans on galectin-1 binding [26] makes it likely to observe a reduction in the extent of anoikis in the case of the GNE transfectant, which was indeed the case (Fig. 10). As the demonstrated partial nature of resialylation by GNE transfection allowed us to surmise, the sensitivity to galectin-1 appeared to be lowered, although a high concentration of galectin (125 μg·mL) could still push cells toward anoikis (not shown). Thus, GNE ‘overexpression’ can contribute to re-establishing certain aspects of the glycan pattern of the tumour-suppressor-negative cells. Together with the several glycosyltransferases previously detected by microarray [12], two enzymes of SA biosynthesis acting in tandem (i.e. GNE and NANS) are controlled by the tumour suppressor. These enzymes eventually influence N- and O-glycan α2,6-sialylation. This new aspect supplements the previously reported survey scheme [22] on the glycobiology of p16INK4a functionality in Capan-1 pancreatic carcinoma cells in vitro (Fig. 11).

Figure 10.

Determination of the extent of anoikis in tumour suppressor p16INK4a-expressing cells without (left) and after (right) transfection to increase the down-regulated level of GNE expression. P = 0.0421.

Figure 11.

Schematic illustration of the glycobiology of p16INK4a functionality in Capan-1 pancreatic carcinoma cells. The scheme presented previously [22] has been extended by the reported contribution of the regulation of SA synthesis (GNE and NANS).


The status of sialylation plays a remarkable role in switching on/off galectin-1-mediated growth control. Although the increased availability of ganglioside GM1 after desialylation is responsible for inhibiting the proliferation of neuroblastoma (SK-N-MC) and T effector cells [70-72] and α2,6-sialyltransferase can serve to mask binding sites on N-glycans (as described in the Introduction), the case of the tumour suppressor p16INK4a appears to place an emphasis on SA biosynthesis for regulating the extent of α2,6-sialylation. Apparently, the α5β1-integrin is a counter-receptor with respect to the up-regulated galectin-1 eliciting anoikis via caspase-8 after cross-linking the glycoprotein (Fig. 11) [12, 22]. The detected tandem down-regulation for GNE and NANS is remarkable. Because carcinoma cells are rather commonly reactive with this lectin, with an indication for a positive correlation of susceptibility to galectin-1-induced anoikis with the presence of the α5-integrin subunit [42, 73], the results of the present study have relevance beyond this special system. Hypothetically, the illustrated bypass with ManNAc might provide a therapeutical option in tumour cases, where galectin-1 reactivity is correlated with tumour progression such as glioblastoma, provided that α2,6-sialyltransferase or any other relevant glycosyltransferase for masking counter-receptors is expressed [74-76]. Intriguingly, stable transfectants for α2,6-sialyltransferase in U373 MG cells, a glioblastoma line lacking the expression of this enzyme, showed a reduced invasivity and intracranial tumour formation after xenotransplantation [77-80]. In this respect, it should be noted that α2,6-sialylation of N-glycans can be selective for distinct target glycoproteins (e.g. CD45 for a murine T cell line, the β1-integrin in human colonocytes or the α3β1-integrin for glioblastoma cells) [27, 28, 45, 80]. Equally intriguing, the case of Klotho-dependent α2,6-desialylation of the Ca2+-channel TRPV5 and the engendered cell surface retention as a result of lattice formation with galectin-1 [81, 82] demonstrates that this type of glycoform switching appears to have enormous potential for impacting upon cell physiology. If these target glycoproteins belong to the set of functional counter-receptors for galectin (as they do), a small modulation in the level of α2,6-sialylation can drastically affect cell responses. To round off the emerging picture, the concomitant regulation of core 1 O-glycan sialylation should not be overlooked as an additional means of regulating galectin-1 reactivity, as shown for LNCaP prostate cancer cells [83], and likely also for other aspects of integrin functionality.

In addition to a role along this metabolic pathway with respect to making the substrate available for switching off galectin reactivity, the down-regulation of GNE may also be involved in other responses for re-normalizing p16INK4a-expression. The recent discovery of the impact of the enzyme on reducing gene expression of α2,3-sialyltransferase-V in human HEK293 epithelial kidney cells (GM3 synthase) [84], as also seen by microarray in p16INK4a-restituted Capan-1 cells [12], broadens the view to functions beyond SA biosynthesis. Distinct gangliosides, affected by such alteration(s), not only make their presence felt by ERK1/2 phosphorylation and apoptosis/proliferation regulation, as shown for GM3/GD3 and HEK293 cells assigning a meaning to such changes [84], but also they may inhibit α5β1-integrin binding to fibronectin (GT1b in SCC12 keratinocyte-derived cells) [85] or directly bind galectin-1 when in functional association with the fibronectin receptor (GM1 on T effector cells) [71, 72, 86]; biosynthetic relationships between the gangliosides are detailed elsewhere [87, 88]. In this respect (i.e. a role of GNE in the regulation of gene expression), controls with ManNAc supplementation and cell culture with serum/serum-free medium (to avoid SA recycling) are essential for separating GNE-sensitive from SA-dependent readouts [51, 84, 89].

Beyond p16INK4a and its capacity to up-regulate galectin-1 expression [12], the tumour suppressor p53 is known to exhibit an inverse relationship with galectin-1 in glioblastoma cells [90, 91]. However, it can strongly act the same as p16INK4a for another homodimeric galectin (i.e. galectin-7; thus referred to as p53-induced gene 1) in human colon cancer (DLD-1) cells [92]. Interestingly, included among the set of genes up-regulated by more than 10-fold by galectin-7 in HeLa cells is the gene for α2,6-sialyltransferase [93], indicating a different mode of action of galectin in this system. The separation of intra- and extracellular galectin activities and a role for α2,6-sialylation in switching off the latter warrants further study. In aggregate, the results of the present study add the contribution of metabolic flux toward SA at two enzymes in tandem to the regulation of cell surface α2,6-sialylation as controlled by the tumour suppressor p16INK4a in this tumour model.

Materials and methods

Reagents and cells

Biotinylated plant lectins (MAA-I, PNA, SNA) were purchased from Vector Laboratories (distributed by Alexis Germany, Grünberg, Germany); the fungal lectin was obtained by recombinant production and purified by affinity chromatography as described previously [67, 69], and then labelled under activity-preserving conditions, as similarly performed for human galectin-1 [12]. The extent of labelling was determined by MS [94] and the labelled lectins were rigorously checked for activity and specificity using (neo)glycoproteins in solid-phase assays and control cells expressing or lacking distinct carbohydrate epitopes (e.g. CHO wild-type cells and glycosylation mutants) to exclude carbohydrate-independent binding [95-97]. Cells of the human pancreatic carcinoma line Capan-1 (HTB 79; American Type Culture Collection, Rockville, MD, USA), stably transfected with full-length cDNA for the human tumour suppressor p16INK4a inserted into a pRC/CMV vector or with control vector (mock control), were grown as described previously [12, 20]. Transfections with cDNA for the human GNE [98] (kindly provided by S. Hinderlich, Department of Life Sciences and Technology, Beuth Hochschule für Technik - University of Applied Sciences, Berlin, Germany) or α2,6-sialyltransferase (kindly provided by W. Kemmner, Research Group Surgical Oncology, Max Delbrueck Center for Molecular Medicine, Berlin, Germany), respectively, were performed with the neomycin-resistant p16INK4a-expressing cells and the SuperFect® reagent in accordance with the manufacturer's protocol (Qiagen, Hilden, Germany) and controls described previously [99], followed by selection in hygromycin-containing medium to obtain stable transfectants and then by profiling vector-positive cells by lectin binding.

Glycan profiling by cell-surface lectin staining

Quantitative determination of carbohydrate-dependent lectin binding, using the fluorescent streptavidin/R-phycoerythrin conjugate (dilution 1 : 40; Sigma, Munich, Germany) as reporter, was performed by flow cytofluorometry in a FACScan instrument (Becton Dickinson, Heidelberg, Germany). Routinely, solutions with 4 × 105 cells per sample were processed. Cell suspensions were first carefully washed to remove any serum components, which may interfere with lectin binding, and carbohydrate-independent binding was minimized by an incubation step with Dulbecco's NaCl/Pi containing 0.1 mg·mL−1 of carbohydrate-free BSA. The incubation with lectin-containing solution was performed at 4 °C for 30 min to keep uptake low, as described previously [12]. Control experiments, in which labelled lectin was omitted from the solution or the lectin was incubated in the presence of a cognate glycocompound [free glycan or (neo)glycoconjugate], were routinely carried out to determine the 0% value. To enhance metabolic flux in SA synthesis, the culture medium was supplemented with 20 mm ManNAc and the cells were kept for 24 h. Thereafter, fresh medium was added, and culture continued at 10 mm ManNAc for an additional 24 h, with control cells kept in parallel. Treatment with neuraminidase from Clostridium perfringens (specific for α2,3/6-sialylation; New England Biolabs, Frankfurt, Germany) was performed for 60 min at 37 °C with 1 μL of enzyme-containing solution per assay in NaCl/Pi.

Glycan profiling by glycoblotting and MS

Release of total cellular N-glycans was carried out as described previously with minor modifications [100]. In brief, cell pellets were scraped in NaCl/Pi containing 10 mm EDTA and washed with NaCl/Pi followed by lyophilization. After being dissolved with NaCl/Pi, cell material was treated with NaCl/Pi containing 1% Triton X-100 for 60 min on ice. The lysate was centrifuged at 204 000 g for 10 min at 4 °C, and the supernatant obtained was aliquoted. A part of the lysate was applied to determine the protein content and cold acetone was added to the other part (final ratio: 80%) to precipitate (glyco)proteins. After overnight incubation at −20 °C, the precipitate was collected by centrifugation at 130 000 g for 15 min at 4 °C followed by serial washes with acetonitrile. The resulting precipitate was dissolved with 50 μL of 80 mm ammonium bicarbonate solution containing 0.02% 1-propanesulfonic acid, 2-hydroxyl-3-myristamido and incubated at 60 °C for 10 min, and then fully reduced by 10 mm dithiothreitol at 60 °C for 30 min, followed by alkylation with 20 mm iodoacetamide in the dark at room temperature for 30 min. The mixture was then treated with 400 U of trypsin (Sigma) at 37 °C overnight, followed by heat-inactivation of the enzyme at 90 °C for 10 min. After cooling to room temperature, N-glycans were released from glycopeptides of the trypsin-digested sample by treatment with 2 U of peptide-N-glycosidase F (Roche Applied Science, Basel, Switzerland) at 37 °C, overnight. Then, the sample mixture was dried up by a SpeedVac (Thermo Fisher Scientific, San Jose, CA, USA) and stored at −20 °C until use.

For mucin-type O-glycans, a protocol described previously was followed [29]. Cell pellets were dissolved with 100 μL of 20 mm Tris-HCl (pH 9) containing 2% SDS and the solution was maintained for 60 min at 100 °C to fully denature glycoproteins, and then the resulting solutions were then centrifuged at 130 000 g for 15 min at 4 °C. The supernatants obtained were aliquoted as described above, and the source for the O-glycans was prepared for subsequent reaction with the addition of a dry powder of ammonium carbamate (final concentration ~ 1 mg·μL−1). The mixture was incubated for 40 h at 60 °C followed by the addition of 500 μL of H2O to the SpeedVac treatment at 60 °C. The residual materials were dissolved with aqueous acetic acid and adjusted to pH 4–5.

The batches of material from preparing N- and O-glycans were next separately subjected to glycoblotting for chemoselective glycan enrichment before comprehensive analysis [30, 101] using BlotGlyco H beads (Sumitomo Bakelite Co., Tokyo, Japan) as described previously [29, 100] with minor modifications. A 250-μL suspension of BlotGlyco H beads (10 mg·mL−1 suspension) was aliquoted onto a well of a MultiScreen Solvinert filter plate (Millipore, Billerica, MA, USA). Released glycans were dissolved with water and the solution was adjusted to a concentration of 100 μg·20 μL−1, and then 20 μL of each was applied to the well followed by the addition of 180 μL of 2% acetic acid in acetonitrile. The plate was incubated at 80 °C for 45 min for covalently conjugating total glycans from sample mixtures onto beads by hydrazone bonds. Each well was washed with 200 μL of a solution of 2 m guanidine-HCl in ammonium bicarbonate followed by washing with the same volume of water and 1% triethylamine in methanol. Each washing step was performed twice. Residual reactive groups on the beads were blocked by incubation with 10% acetic anhydride in methanol for 30 min at room temperature. The solution was removed by vacuum and then beads were serially washed twice by 200 μL of 10 mm HCl, methanol and dioxane, respectively. On-bead methyl esterification of carboxyl groups in SAs was carried out by incubation with 150 mm 3-methyl-1-p-tolyltriazene in dioxane at 60 °C to dryness. This usually took 90 min in a conventional oven. Beads were then consecutively washed with 200 μL of dioxane, methanol and water. The glycans ‘blotted’ onto the beads were subjected to the trans-iminization reaction with aminooxy-functionalized tryptophanyl arginine (ao-WR) or benzyloxiamine fluoride for 45 min at 80 °C. The ao-WR/benzyloxiamine fluoride-tagged glycans were eluted by adding 100 μL of water, and ao-WR-tagged oligosaccharides were purified by Mass PREP HILIC μElution Plate (Waters Corp, Milford, MA, USA) in accordance with the manufacturer's instructions.

These preparations of tagged glycans were concentrated 10-fold using a SpeedVac. The concentrated solutions were mixed with 2,5-dihydroxybenzoic acid [10 mg·mL−1 in 30% acetonitrile (1 : 2) or a mixture of this compound and monosodium salt] (9 : 1), respectively, and kept for crystal formation. Then the analytes were subjected to MALDI-TOF MS analysis using an Ultraflex III time-of-flight mass spectrometer (Brucker Daltonics, Bremen, Germany) in reflector, positive-ion mode, typically performing 1000–10 000 runs. Peaks in the MALDI-TOF MS spectra were fitted to isotopic patterns of matching N-glycans using flexanalysis, version 3.0 (Brucker Daltonics). To quantitate the data obtained, the intensity of the isotopic peaks of each glycan was normalized to the indicated concentration of the internal standard (A2 amide glycan or chitotetraose for N- or O-glycomics, respectively). Using these data as an input into the glycomod tool (http://web.expasy.org/glycomod/) and the GlycoSuiteDB database (http://glycosuitedb.expasy.org/glycosuite/glycodb), the respective structures for experimentally determined masses and glycan composition were derived.

Proteomic analysis

Eight cell pellets (four from the p16INK4a-transfected cells and four from mock-transfected cells) of 1–2 × 106 cells each were lysed in 300 μL of SDS-containing lysis buffer [4% SDS (w/v), 25 mm Hepes pH 7.6, 1 mm dithiothreitol]. The samples were heated at 95 °C for 5 min with shaking, after which they were sonicated for 5 min. The lysates were then cleared by centrifugation at 14 000 g for 15 min. The protein concentrations were measured using the Bio-Rad DC protein assay (Bio-Rad Laboratories, Hercules, CA, USA). For digestion, the filter aided sample preparation protocol was used [102]. In short, 200 μL of MQ water (Millipore) was added to the filter units (Nanosep Centrifugal Devices, catalogue number: OD010C34 10K; Pall, Port Washington, NY, USA) followed by centrifugation at 14 000 g for 15 min. Then, 200 μL of urea-containing buffer 1 (8 m urea, 1 mm dithiothreitol in 25 mm Hepes, pH 7.6) was mixed with 150 μg of protein from each sample. The samples were then added to the filter units and centrifuged at 14 000 g for 15 min. Next, 200 μL of urea-containing buffer 1 was pipetted to the solution followed by centrifugation at 14 000 g for 15 min. Thereafter, 200 μL of urea-containing buffer 2 (4 m urea, 25 mm iodoacetamide in 25 mm Hepes, pH 7.6) was added and mixed for 10 min in a thermomixer at room temperature. The units were then centrifuged at 14 000 g for 15 min. Then, 200 μL of urea-containing buffer 3 (4 m urea in 25 mm Hepes, pH 7.6) was added followed by centrifugation at 14 000 g for 15 min (twice). The flow-through was discarded. Next, 100 μL of urea-containing buffer 4 (0.25 m urea in 100 mm Hepes, pH 7.6), in which trypsin had been dissolved (modified sequencing grade; Promega, Madison, WI, USA) was added (dilution 1 : 50, trypsin : protein) and the samples were incubated overnight at 37 °C on a thermomixer with rotary mixing at 300 r.p.m. The filter units were then centrifuged at 14 000 g for 15 min. Then, 50 μL of MQ water was added, and the units were centrifuged again. The flow-through was collected and the peptide concentration was determined with the Bio-Rad DC protein assay. Subsequently, 50 μg of each peptide sample were labelled and pooled using the 8-plex iTRAQ kit (Applied Biosystems, Foster City, CA, USA) in accordance with the manufacturer's instructions. Excess reagent was removed from the pooled sample using an SCX-cartridge (StrataSCX; Phenomenex, Torrence, CA, USA). The eluate was dried in a speed-vac and the peptides processed by isoelectric focusing.

iTRAQ-labelled tryptic peptide samples were dissolved in 200 μL of solution containing 8 m urea. A narrow-range immobilized pH gradient (IPG)-strip for peptide focusing (pH 3.7–4.9, length 24 cm) together with dry sample application gel (33 × 3 × 2 mm) was kindly supplied by GE Healthcare Biosciences AB (Uppsala, Sweden). The application gel was rehydrated in sample overnight, whereas the strip was rehydrated overnight in solution containing 8 m urea and 1% Pharmalyte 2.5–5 (GE Healthcare Biosciences AB). The IPG strip was put in the focusing tray, and the application gel containing the sample was placed on the anodic end of the IPG strip with filter paper between the strip and the electrodes. The strip was covered with mineral oil, and the focusing was performed on an Ettan IPGphor (GE Healthcare Biosciences AB) until reaching 100 kVh. After focusing, peptides were extracted from the strip by a prototype liquid-handling robot. A plastic device with 72 wells was put onto the strip and 50 μL of MQ water was added to each well. After 30 min of incubation, the liquid was transferred to a 96-well plate and the extraction was repeated twice. Samples were then freeze dried in SpeedVac and maintained at −20 °C. Before the analysis, each fraction was resuspended in 8 μL of 3% acetonitrile and 0.1% formic acid.

LC-ESI-LTQ-Orbitrap analysis

LC-MS was performed on a hybrid LTQ-Orbitrap Velos mass spectrometer (Thermo Fisher Scientific). An Agilent HPLC 1200 system (Agilent Technologies, Santa Clara, CA, USA) was used for online reversed-phase nano-LC at a flow of 0.4 μL·min−1. Solvent A comprised 97% water, 3% acetonitrile and 0.1% formic acid; solvent B comprised 5% water, 95% acetonitrile and 0.1% formic acid. The curved gradient went from 2% B up to 40% B in 45 min, followed by a steep increase to 100% B in 5 min. Samples [3 (of 8) μL from each IPG fraction] were trapped on Zorbax 300SB-C18, 5 μm, 5 × 0.3 mm (Agilent Technologies) and separated on a NTCC-360/100-5-153 C18 column (Nikkyo Technos Co., Tokyo, Japan) installed onto the nanoelectrospray ionization source of the Orbitrap Velos instrument. Acquisition proceeded in approximately 3.5-s scan cycles, starting by a single full-scan MS at 30 000 resolution (profile mode), followed by two stages of data-dependent tandem MS (centroid mode): the top five ions from the full-scan MS were selected first for collision-induced dissociation (at 35% energy) with MS/MS detection in the ion trap and, finally, for high-energy collision dissociation (at 37.5% energy) with MS/MS detection in the orbitrap. Precursors were isolated with a width of 2 m/z, and dynamic exclusion was used with a duration of 60 s.

Data analysis

The MS/MS spectra were searched using proteome discoverer, version 1.3 (Thermo Fischer Scientific) against the Swiss-Prot human protein sequence database (update 6 July 2011). A precursor mass tolerance of 10 p.p.m., along with product mass tolerances of 0.02 Da for high-energy collision dissociation-Fourier transform MS and 0.8 Da for collision-induced dissociation-ion trap MS, were used. Additional settings were: trypsin with one missed cleavage; carbamidomethylation on cysteine and iTRAQ-8plex on lysine and N-terminus as fixed modifications; oxidation of methionine as variable modification. Quantitation of iTRAQ-8plex reporter ions was performed using proteome discoverer on high-energy collision dissociation-FTMS tandem mass spectra with an integration window tolerance of 20 p.p.m. Results were limited to ≥ 1 high-confident peptide for quantification using a false discovery rate of < 1%.

Determination of anoikis

For induction of anoikis, 2–3 × 105 cells were cultured for 22 h as suspension cultures in plates coated with poly(2-hydroxyethyl methacrylate) (PolyHEMA; Sigma). Cells were harvested by centrifugation, fixed with ice-cold ethanol (70%), resuspended in 300 μL of propidium iodide solution (20 μg·mL−1 in NaCl/Pi) containing 20 μg·mL−1 of RNase (Roche Applied Science) and incubated for 30 min at 37 °C. At least 1 × 105 cells were analyzed on FACSCalibur equipment utilizing cellquest™ software (Becton Dickinson). Apoptotic cells were then identified from the pre-G1 fraction of cell cycle analyses as described previously [12]. Anoikis is expressed as the percentage of cells undergoing apoptosis under these conditions (n = 4; independent experiments). Treatment with human galectin-1 (125 μg·mL−1) was started when cells were transferred to PolyHEMA-coated plates and continued until cells were harvested.

Statistical analysis

Student's t-test (http://translationalmedicine.stanford.edu/Mass-Conductor/FDR.html) was performed on sets of data with respect to the levels of expression of the selected proteins obtained from the mock-treated or tumour suppressor p16INK4a-expressing cells. The test used included correction for multiple testing.


This work was generously supported by the EC Seventh Framework Program (FP7/2007-2013) under contract agreement no. 26060 (GlycoHIT), as well as by grants for ‘Development of Systems and Technology for Advanced Measurement and Analysis (SENTAN)’ and ‘The Matching Program for Innovations in Future Drug Discovery and Medical Care’ from the Japan Science and Technology Agency (JST); the Ministry of Education, Culture, Science, and Technology, Japan; the Swedish Cancer Society; the Swedish Research Council; and the Verein zur Förderung des biologisch-technologischen Fortschritts in der Medizin e.V. (Heidelberg, Germany). We are grateful to Drs B. Friday, G. Ippans, Y. Nekcic and W. Notelecs for inspiring discussions.