K. M. Detjen, Charité-Universitätsmedizin Berlin, Campus Virchow-Klinikum, Med. Klinik m.S. Hepatologie und Gastroenterologie, Augustenburger Platz 1, D-13353 Berlin, Germany Fax: +49 30 450 559939 Tel: +49 30 450 559679 E-mail: email@example.com
The tumor suppressor p16INK4a has functions beyond cell-cycle control via cyclin-dependent kinases. A coordinated remodeling of N- and O-glycosylation, and an increase in the presentation of the endogenous lectin galectin-1 sensing these changes on the surface of p16INK4a-expressing pancreatic carcinoma cells (Capan-1), lead to potent pro-anoikis signals. We show that the p16INK4a-dependent impact on growth-regulatory lectins is not limited to galectin-1, but also concerns galectin-3. By monitoring its expression in relation to p16INK4a status, as well as running anoikis assays with galectin-3 and cell transfectants with up- or downregulated lectin expression, a negative correlation between anoikis and the presence of this lectin was established. Nuclear run-off and northern blotting experiments revealed an effect of the presence of p16INK4a on steady-state levels of galectin-3-specific mRNA that differed from decreasing the transcriptional rate. On the cell surface, galectin-3 interferes with galectin-1, which initiates signaling toward its pro-anoikis activity via caspase-8 activation. The detected opposite effects of p16INK4a at the levels of growth-regulatory galectins-1 and -3 shift the status markedly towards the galectin-1-dependent pro-anoikis activity. A previously undescribed orchestrated fine-tuning of this effector system by a tumor suppressor is discovered.
The tumor suppressor p16INK4a, a frequent target for deletion mutations underlying carcinogenesis, is known as a binding partner of cyclin-dependent kinases CDK4 and CDK6, interfering with their association with D-type cyclins [1,2]. Emerging evidence broadens the spectrum of p16INK4a functionality beyond cell-cycle control. In fact, recent experiments have unraveled an effect on distinct aspects of gene expression. One functional consequence is to restore the susceptibility of tumor cells to anoikis (the category of apoptosis caused by inadequate or inappropriate cell–matrix contacts). In detail, enhanced production of the α5-integrin subunit and also an increased cell-surface presence of α5β1-integrin (the fibronectin receptor) were detected in p16INK4a-restituted Capan-1 pancreatic carcinoma cells . Routing of this glycoprotein and the levels of its cell-surface presentation, binding activity and capacity for downstream signaling may, in principle, depend not only on the protein part, but also on its glycosylation. Because the fibronectin receptor is heavily glycosylated with 26 sites for N-glycosylation  and variations in the structures of glycan chains, such as status of sialylation, have a bearing on protein functions in general and on integrins in particular [5–8], glycan remodeling affords an attractive level of regulation of integrin functionality. The assumption of a modulatory impact of p16INK4a at the level of glycosylation was a reasonable and testable hypothesis. This hypothesis was verified by combining glycogene microarray analysis, chromatographic glycan profiling and lectin binding . Influencing N- and O-glycan galactosylation and sialylation, salient biochemical signals in protein–carbohydrate/protein interplay [10–12], have thus become a new aspect of p16INK4a functionality. Hereby, the tumor suppressor can affect integrin processing and routing, as well as its affinity for protein ligands.
In addition, modified glycan chains can be engaged in recognitive interactions at the cell surface with endogenous lectins, which translate sugar-encoded messages into cellular responses [13–15]. Even seemingly small modifications can act as potent switches of affinity. In fact, altering branch-end positions, the status of core substitutions and glycan density strongly affects lectin reactivity, as measured for example for members of the family of adhesion/growth-regulatory galectins [16–19]. Fittingly for a functional implication, microarray analysis on a set of 1996 cancer-associated genes, proteomic profiling and cytofluorometry and anoikis assays revealed the upregulation of homodimeric galectin-1 (Gal-1) as a carbohydrate-binding effector for anoikis induction under the control of p16INK4a . Thus, this tumor suppressor sensitizes cells for the onset of anoikis by increasing the complementary sides of pro-anoikis protein–carbohydrate interactions.
As the term Gal-1 implies, the functionality of this protein might be embedded in a network with other family members . This raises the possibility of additive or antagonistic effects. Exploring this issue, model studies on SK-N-MC neuroblastoma cells illustrated the potential of other galectins, especially galectin-3 (Gal-3), to factor in growth inhibition by Gal-1 at the level of cell-surface ligand binding [21,22]. Having recently delineated a direct connection between p16INK4a and Gal-1 , the question arises as to whether the influence of p16INK4a is restricted to this family member. The following four lines of evidence direct the focus of attention on Gal-3. First, it is functionally antagonistic to Gal-1 in the neuroblastoma system by blocking access to ganglioside GM1, the common galectin ligand on the cell surface in these cells . Second, activating K-ras mutations are common in pancreatic cancer, and Gal-3 (but not Gal-1) interacts with oncogenic K-ras–GTP to promote Raf-1 and phosphoinositide 3-kinase activities, as well as a signal attenuating extracellular signal-regulated kinase . This is in accordance with the lack of aberrantly enhanced extracellular signal-regulated kinase signaling in pancreatic tumors which harbor the mentioned gene defect . Third, Gal-3 protects BT549 breast cancer cells from anoikis . Fourth, Capan-1 wild-type cells are known to express Gal-3 . On this basis, we investigated the influence of the presence of p16INK4a on the level of Gal-3 production and possible functional competition between galectins-1 and -3. The reported results document the relevance of Gal-3 in interfering with anoikis induction and, more importantly, shed light on the intriguing capacity of this tumor suppressor to coordinately exploit this aspect of the galectin network.
Expression of p16INK4a decreases the level of Gal-3
In order to delineate any impact of the tumor suppressor p16INK4a on Gal-3 expression we worked with Capan-1 pancreatic carcinoma wild-type and vector-transfected cells (mock) as well as three independently generated clones stably transfected with p16INK4a tumor suppressor cDNA (p16 1-3). Of note, these clones had been studied previously, revealing increased expression and cell-surface presentation of the α5-integrin subunit and Gal-1 [3,9]. Western blot analyses first resulted in the expected pattern for the p16INK4a protein, then confirmed the reported presence of Gal-3 in Capan-1 wild-type cells  and excluded an effect of the control transfection on Gal-3 and general protein synthesis (Fig. 1A). In contrast to the mock process, the presence of p16INK4a protein had an effect. A clear decrease in the intensity of signals for Gal-3 was invariably observed, although proliferating cell nuclear antigen (PCNA) staining as a control for loading remained rather constant (Fig. 1A,B). To examine whether such a negative correlation between Gal-3 and p16INK4a can also be seen in vivo, we processed routinely fixed sections from normal pancreas tissue and pancreatic cancer immunohistochemically. The resulting staining profiles, shown in Fig. S1, confirmed an inverse expression pattern in accordance with the in vitro data. In principle, Gal-3 can act as an anti-anoikis effector in the cell and/or at the cell surface, for example, blocking Gal-1 functionally as seen previously in a neuroblastoma model . In view of the proven pro-anoikis cell-surface activity of Gal-1, the cultured cells provided the opportunity to test the hypothesis of reduced cell-surface expression of Gal-3. Indeed, cytofluorimetric monitoring of mock-transfected and p16INK4a-restituted Capan-1 cells revealed that cell-surface Gal-3 was downregulated (Fig. 1C). As reported previously , Gal-1 cell-surface presentation determined in the control increased (not shown). The presence of the tumor suppressor thus reduced the level of Gal-3 and its cell-surface presentation. To further test whether Gal-3 protein abundance was regulated by prolonged culture in suspension, we maintained cells for up to 24 h under this condition, with the percentage of cells undergoing anoikis increasing as expected in p16INK4a-expresssing cells (Fig. 2A). Loss of anchorage did not notably change the level of Gal-3 detected in western blots (Fig. 2B). These results document a significant decrease in the presence of Gal-3 in tumor-suppressor-positive cells, which is not further enhanced in suspension culture. This suggests regulation at the level of transcription, as previously detected for Gal-1 . To test this, we determined the steady-state mRNA concentration by northern blotting and the de novo transcription rate by run-off assays.
Northern blotting using samples from p16INK4a-positive cells revealed a conspicuous decrease in the availability of Gal-3-specific mRNA, when compared with mock-transfected controls, whereas similar mRNA levels for the housekeeping gene GAPDH were present irrespective of cell status (Fig. 2C, left). This diminished steady-state amount may be because of alterations in de novo production or availability. Nuclear run-off assays, which report on de novo synthesis, revealed rather similar signal intensities in samples from mock- and p16INK4a-transfected cells for the two tested housekeeping genes and for Gal-3 (Fig. 2C, right). De novo Gal-1 gene transcription, by contrast, increased significantly in the presence of the tumor suppressor (not shown). Thus, the transcriptional rate is a major factor in increasing the production of Gal-1, whereas post-transcriptional processes decrease Gal-3 production. We propose a functional relevance for this alteration. In order to prove an effect of Gal-3 on anoikis in this cell system, especially regarding a functional antagonism with Gal-1 at the level of the cell surface, we followed three routes of investigation to test the validity of this assumption.
Gal-3 is an inhibitor of anoikis
In the first set of experiments, we tested the capacity of Gal-3 to block anoikis in p16INK4a-expressing cells when added to the medium. As shown in Fig. 3, exposure of cells to Gal-3 was inhibitory in this experimental setting. The percentage of p16INK4a-positive cells undergoing anoikis after 20 h in suspension was reduced when kept in the presence of suitable concentrations of Gal-3. To further strengthen the link between Gal-3 and anoikis, the availability of this lectin was upregulated deliberately by generating p16INK4-positive cells, which additionally express Gal-3 at a high level via a second transfection. Thus, the presence of Gal-3 was increased, counteracting the p16INK4a-dependent downregulation. A series of five clones was obtained with notably enhanced Gal-3 concentration, an unspecific influence of the second transfection step rendered unlikely by a mock control (Fig. 4A). The level of anoikis in these control cells was used as a reference, and four Gal-3-overexpressing clones revealed a trend towards reduced anoikis susceptibility. This reached statistical significance in clone Gal-3/2 (Fig. 4B). Results showing that exogenous addition and vector-directed overexpression of Gal-3 can reduce the level of anoikis shaped the notion that Gal-3 physically hampers the induction of this cell death program. In this case, even p16INK4a-negative cells should become sensitized toward anoikis if their Gal-3 production is downregulated. This reasoning led to the third set of experiments.
To test the given hypothesis, we generated clones harboring an antisense vector for Gal-3. Because of this engineering, the clones contained a lower level of Gal-3 than wild-type and mock-transfected cell populations (Fig. 4C). Measurements of the corresponding cell-cycle profiles and percentages of cells undergoing anoikis revealed a Gal-3-dependent increase in this parameter (Fig. 4D,E). Evidently, the level of Gal-3 is important to protect wild-type cells from anoikis induction, and the forced expression of Gal-3 in p16INK4a-positive cells reduces their susceptibility to anoikis.
Because this pattern is inverse to the Gal-1-induced effects reported previously , it is reasonable to propose functional competition between galectins-1 and -3 at the cell surface. To probe for such an effect, we stimulated anoikis by addition of Gal-1 to the culture medium. Should Gal-3 be an inhibitor, the extent of Gal-1-dependent anoikis induction via glycan binding will decrease. Stepwise increases in the Gal-3 concentration progressively diminished the Gal-1-dependent effect (Fig. 5A), and galectin binding, shown to be carbohydrate dependent , was reduced in cross-competition assays (Fig. 5B,C). The presence of Gal-3 can thus impair the pro-anoikis effect of Gal-1 at the level of the cell surface, here most likely targeting the fibronectin receptor as had been shown in this and other carcinoma cell systems [9,27]. If Gal-3 interferes with Gal-1 binding to and cross-linking the α5-subunit, then Gal-3 should also negatively affect post-binding signaling by the integrin. We tested and established the involvement of caspase-8 activation for Gal-1-dependent anoikis induction (Fig. 6A) and revealed a negative impact of Gal-3 at this level (Fig. 6B,C).
The term ‘tumor suppressor’ summarizes an obvious function of a protein on malignancy. Naturally, a suppressor engages secondary effectors at different levels, which turn its presence as a master organizer into renormalization of phenotypic characteristics. Because of growing insights into the role of glycosylation in cellular communication, including growth control [6,28], we hypothesized that p16INK4a is capable of taking advantage of this network, explicitly by modulating the lectin reactivity of glycans and/or lectin expression. As a consequence, further study provided a novel explanation for why this tumor suppressor restores susceptibility to anoikis in Capan-1 pancreatic carcinoma cells . Having detected coordinated upregulation of both Gal-1 and suitable cell-surface ligands, the results provoked a question regarding orchestration of effects on expression in the lectin family beyond Gal-1. Evidently, the chimera-type Gal-3, known as a functional antagonist of Gal-1 in a tumor model, offered a prime target for study.
Cumulatively, the experiments reported herein revealed p16INK4a-dependent regulation of the presence of Gal-3 in this cell system. They also delineated a strong influence of Gal-3 on anoikis induction. Of special note, we detected functional competition at the cell surface with the recently proven pro-anoikis activity of Gal-1, which involves caspase-8 activation. Combining previous results on Gal-1/α5β1-integrin co-immunoprecipitation, and the effects of p16INK4a on α5β1-integrin, cell-surface glycosylation and Gal-1 [3,9] with the presented data enabled us to set up a scheme of p16INK4a-orchestrated changes that favor Gal-1-dependent anoikis (Fig. 7). Whether and how a reduction in the intracellular activities of Gal-3 documented in other tumor cell types (e.g. interaction with oncogenic K-ras and transcriptional modulation of cell-cycle regulators such as cyclins A, D1 and E as well as p21/p27 [23,25,29]) will cooperate with the functional interference of cell-surface binding of homodimeric Gal-1, is not clear. At the cell surface, the particular profile of the chimera-type galectin for ligand cross-linking shown recently  will definitely not elicit pro-anoikis signaling of Gal-1 in this cell system. This functional competition between galectins, described initially for SK-N-MC neuroblastoma cells and ganglioside GM1 , may well play a more general role in tumor growth control. Its detection gives a clear direction to further research. In this respect, such a study is warranted on the p27-dependent downregulation of carcinoma cell growth triggered by Gal-1 and most likely the α5β1-integrin , as well as the Gal-1/GM1-dependent communication between effector T and regulatory T cells, which involves α4/α5β1-integrins and Ca2+ influx via TRPC5 channels . Eventually, these investigations will unravel the mostly unexplored intricacies of the galectin network monitored in tumors by RT-PCR and by immunohistochemistry [32–34]. Such cell biological studies will then shed light on additive/synergistic, as opposed to antagonistic, activities in malignancy and immune regulation.
Equally important, our study broadens the basis for galectin involvement in tumor suppressor activity. Here, the key reference point to date had been p53. In detail, SAGE screening on DLD-1 colon carcinoma cells expressing p53 identified galectin-7 as a p53-induced gene 1 in a group of 14 genes markedly upregulated from a total of 7202 tested transcripts . Along this line, genotoxic stress by UVB irradiation of human keratinocytes afforded a second system in which a connection between p53 and galectin-7 was made likely . When expressed in HeLa cell transfectants, the intracellular function of galectin-7 appeared to relate to affecting gene expression profiles. Conspicuous changes were attributed to pro-apoptosis signaling upstream of c-Jun N-terminal kinase activation and cytochrome c release . When tested as an extracellular effector, carbohydrate-dependent binding led to growth inhibition of activated T cells or neuroblastoma cells via caspase-dependent or -independent pathways [38,39]. Of further relevance, the recently documented effect of compensation of loss of supressor genes in microsatellite instability on glycosylation, including α2,6-sialylation of N-glycans, broadens the scope for co-regulation . Further work to strengthen this connection between a tumor suppressor and lectin/glycan remodeling as an effector pathway, especially by examining clinical samples, is clearly justified.
Examining galectin expression in clinical samples of pancreatic cancer by gene-expression profiling uncovered a consistent, albeit quantitatively variable, upregulation of Gal-3 expression [41–45]. The similarly consistent upregulation of Gal-1 gene expression and protein production may at first seem puzzling. Looking at its immunohistochemical pattern, it could mostly be accounted for by a desmoplastic reaction around the tumor cells, with a Gal-1/tissue plasminogen activator interaction being involved [44–50]. Taking our data literally, an absence or low level of Gal-1, combined with an abundance of Gal-3 in the tumor cells, seems to reflect impairment of p16INK4a. As shown in the Supporting Information, preliminary immunohistochemical analysis on a limited number of cases focusing on the presence of p16INK4a and Gal-3 indicated a tendency for a negative correlation, in accordance with the in vitro data. These results encourage thorough investigation of this aspect to strengthen clinical relevance. By following this line of research, the orchestration of galectin expression described here may become instrumental in devising a novel therapeutic strategy to rationally shift the balance between resistance and susceptibility to favor the anoikis process.
Materials and methods
Human galectins were produced by recombinant expression, isolated by affinity chromatography on lactosylated Sepharose 4B as crucial step followed by gel filtration, and tested for purity using 1D and 2D gel electrophoresis and nano-electrospray ionization mass spectrometry, as well as for activity by hemagglutination [9,38,51,52]. Biotinylation labeling was performed under activity-preserving conditions using the commercial N-hydroxysuccinimide ester derivative of biotin (Sigma, Munich, Germany). Its incorporation into the galectins was determined using a proteomics protocol, and the activity of the labeled proteins was checked using carbohydrate-dependent solid-phase and cell binding assays [52,53]. Polyclonal antibodies were raised in rabbits and rigorously checked for lack of cross-reactivity using enzyme-linked immunosorbent assays and western blotting [54,55]. No experimental animals were used in this study. Monitoring the cell-surface presentation of galectins was carried out by flow cytofluorometry using fluorescent goat anti-rabbit IgG with 20 μg·mL−1 galectin-type-specific IgG fractions and FACS equipment (Becton-Dickinson, Heidelberg, Germany) .
Human Capan-1 pancreatic carcinoma cells and clones with stable vector-directed p16INK4a expression were established and cultured as described previously .
Antibodies to PCNA and p16INK4a were obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA) and NeoMarkers (Fremont, CA, USA), respectively.
Protein extraction and western blotting
Cells were lyzed in radioimmunoprecipitation assay buffer (50 mm Tris/HCl at pH 7.5, 0.15 m NaCl, 0.25% SDS, 0.05% sodium deoxycholate, 1% NP-40, 1 mm dithiothreitol, 1 μg·mL−1 aprotinin, 2 mm leupeptin, 1 mm Na3VO4, 1 mm NaF, 1 mm phenylmethylsulfonyl fluoride), and sonicated on ice. Aliquots (5–10 μg) were subjected to SDS/PAGE and electroblotted onto poly(vinylidene fluoride) membranes (NEN, Cologne, Germany). Blots were incubated overnight at 4 °C with the respective antibodies (diluted 1 : 1000 in 5% non-fat dried milk in phosphate buffered saline with 0.5% Tween-20). Immunoreactive bands were visualized by enhanced chemoluminescence (NEN). Cell processing, blotting and signal generation when using 2D gel electrophoresis followed the procedure previously used to detect Gal-1 .
Immunohistochemistry was performed on cryosections with antibodies controlled for specificity and lack of intergalectin cross-reactivity, as described previously [3,56,57]. Briefly, sections were fixed in 4% paraformaldehyde and primary antibodies were applied at a dilution of 10 μg·mL−1 (Gal-3) or 1.25 μg·mL−1 (p16INK4a). Immunoreactivity was detected with biotinylated secondary antibodies and avidin–biotin–peroxidase complex, with 3-amino-9-ethylcarbazole as the chromogenic substrate. Sections were counterstained with hemalaun.
Northern blot analysis
Total RNA from 108 cells was isolated using RNAzol™B (WAK-Chemie Medicals GmbH, Bad Homburg, Germany), and the poly-(A+) fraction was purified using the PolyATrack® System 1000 (Promega, Mannheim, Germany) according to the manufacturer’s protocol. Aliquots were separated on a 1% agarose/3-(N-morpholino)propanesulfonic acid/formaldehyde gel, blotted to Hybond N+ filters (Amersham Pharmacia, Freiburg, Germany) and linked to the membrane using UV light. Full-length cDNA preparations for human Gal-3 and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were labeled with [32P]dCTP[αP] by random priming (Megaprime DNA labeling kit; Amersham Pharmacia). Residual nucleotides were removed and hybridization was carried out in Quick-Hyb buffer (Stratagene, La Jolla, CA, USA) at 65 °C for 2 h. Following hybridization, membranes were washed at 65 °C to a stringency of 0.1× NaCl/Cit, 0.1% SDS and exposed to X-ray films at –70 °C.
Blots were prepared using isolated and denatured cDNA fragments immobilized on Hybond N+ nylon membranes (Amersham Pharmacia). The amount of denatured cDNA for Gal-1, GAPDH and β-actin blotted was 5, 2 and 1 μg·slot−1, respectively. Nuclear RNA preparation, labeling and hybridization were performed as described previously . Blots were washed twice for 10 min at 42 °C with 40 mm NaH2PO4/Na2HPO4 pH 7.2, 1% SDS. Labeled de novo mRNA transcripts were detected on autoradiographs.
Induction and detection of anoikis
For determination of anoikis, 2 × 105 cells were cultured as suspension cultures in plates coated with poly(2-hydroxyethyl methacrylate) (poly-HEMA) (Sigma, Deisenhofen, Germany) for the indicated times. Apoptotic cells were then quantitated from the pre-G1 fraction in cell cycle analyses as described .
Stable transfection of Gal-3 sense/antisense cDNA
Full-length cDNA obtained from amplification of human Gal-3-specific mRNA of human DLD-1 colon carcinoma cells in either the sense (pcDNA–Gal-3S) or antisense (pcDNA–Gal-3AS) orientation was subcloned into the pcDNA3.1 vector, and the Effectene™ Transfection Reagent (Quiagen, Hilden, Germany) was used to generate stably transfected cells following the manufacturer’s protocol.
Determination of binding of labeled galectins
Cells were incubated with 100 μg·mL−1 of biotinylated Gal-1 and -3, washed and surface-bound probe was detected by flow cytometry using an indocarbocyanine–streptavidin conjugate. Fluorescence intensity was recorded on a FACSCalibur™ (Becton Dickinson) and analyzed with cellquest™ software [9,58]. Cells incubated with the fluorescent indicator only were used to determine the background fluorescence.
Determination of caspase-8 activity
Cells with activated caspase-8 were detected using the carboxyfluorescein-labeled derivative of the caspase-8 inhibitor Z-LETD-FMK (FAM-LETD-FMK) (Biocarta, Hamburg, Germany), which irreversibly binds to activated caspase-8. Fluorescence intensity was evaluated by flow cytometry.
Unless indicated, unpaired Student’s t-test analyses (two-tailed distribution, two-sample unequal variance) were performed using prism software (Prism, San Diego, CA, USA). Data were considered significant at P-values < 0.05.
This work is dedicated to Prof. Dr Stefan Rosewicz (1960–2004), who was crucial to start this project line. We are grateful to Drs B. Friday, G. Ippans and S. Namirha for helpful comments, to L. Mantel for excellent technical assistance as well as to the Dr Mildred Scheel Stiftung, Sonnenfeld Stiftung, the Wilhelm-Sander-Stiftung, LMUexcellent program, the Verein zur Förderung des biologisch-technologischen Fortschritts in der Medizin e.V. and the EC program on Marie Curie Research Training Networks (contract no. MRTN-CT-2005-019561) for generous financial support.