Introduction

Galectins represent a family of lactose-binding lectins with paramount biological activities (Hughes, 2001). Galectin-3 (Gal-3), a small (30 kDa) lectin, has been implicated during tumor progression (van den Brûle and Castronovo, 2000; Danguy et al., 2002; Califice et al., in press). This monomeric protein is composed of three distinct domains: a COOH-terminal carbohydrate-binding domain, a 20-amino-acid NH₂-terminal domain and a domain consisting of repeating elements rich in Pro, Gly and Tyr (R-domain) (Herrmann et al., 1993). Gal-3 participates to several physiological and pathological events. The lectin is detected in specific cell compartments where it appears to play various functions (van den Brûle et al., 2000). Associated to the plasma membrane and secreted in the extracellular matrix (Frigeri and Liu, 1992; Sato and Hughes, 1994), Gal-3 mediates cell–cell and cell–matrix interactions through its ability to bind to a variety of glycoconjugates (reviewed in van den Brûle and Castronovo, 2000; Califice et al., in press). Gal-3 is also found in the cytoplasm (Gritzmacher et al., 1988) and perinuclear mitochondrial membranes (Yu et al., 2002), where it is involved in the control of apoptosis, possibly through an interaction with the Bcl-2 protein (Yang et al., 1996; Akahani et al., 1997). In vitro experiments have shown that nuclear Gal-3 can bind nucleic acids (Wang et al., 1995). Proliferating fibroblasts present with increased nuclear expression of Gal-3, compared to quiescent cells (Moutsatsos et al., 1987). Both phosphorylated and unphosphorylated Gal-3 forms are found in the nucleus of these fibroblasts while only the phosphorylated form is found in the cytosol (Cowles et al., 1990). Gal-3 does not contain a nuclear localization domain (Gaudin et al., 2000). When we began this study, the mechanisms underlying Gal-3 nuclear import were not known. Nuclear export of Gal-3 is mediated through a leucin-rich nuclear export signal (Tsay et al., 1999). Nuclear Gal-3 is involved in pre-mRNA splicing (Dagher et al., 1995), as a member of the survival of motor neuron (SMN) complex (Park et al., 2001). Moreover, Gal-3 might regulate gene transcription since it induces cyclin D(1) promoter activity (Lin et al., 2002).

Expression and also cellular localization of Gal-3 are important for the prognostic evaluation of a variety of cancers. Sanjuan et al. (1997) demonstrated downregulation of Gal-3 expression in colorectal cancer, with increased cytoplasmic expression of Gal-3 at more advanced stages. Lotz et al. (1993) reported that exclusion of the lectin from the nucleus correlates with transition from normal mucosa to adenoma and carcinoma. Endometrial cancer cells are characterized by downregulation of Gal-3 expression compared with normal mucosa; interestingly, cytoplasmic expression of Gal-3 in the cancer cells is associated with deeper invasion of the myometrium, compared to the lesions where Gal-3 was located in the nucleus (van den Brûle et al., 1996). Downregulation of Gal-3 was observed in prostate carcinoma (van den Brûle et al., 2000). Nuclear exclusion and cytoplasmic localization of Gal-3 are correlated with disease progression (van den Brûle et al., 2000). In tongue cancers, nuclear Gal-3 decreased, and cytoplasmic Gal-3 increased during progression from normal to cancer and increased cytoplasmic Gal-3 was associated to reduced disease-free survival (Honjo et al., 2000).

We explored the specific functions of Gal-3 in human prostate cancer cells when located in the nucleus or cytosol of LNCaP, a Gal-3-negative prostate cancer cell line. We performed stable transfection of expression vectors that allow targeting of Gal-3 in specific cell compartments, that is, by fusion with nuclear localization sequences (NLSs) for nuclear targeting. The resulting stable clones have been tested for the effect of nuclear or cytoplasmic Gal-3 on the in vitro and in vivo expression of the cancer phenotype.

Results

Characterization of the cells

Stable LNCaP transfected clones expressing cytoplasmic or nuclear Gal-3, or nuclear GFP, were established. We selected four LNCaP clones with nuclear Gal-3 (transfected with pCMV/myc/nuc/G3, N1–N4) and four clones with cytoplasmic Gal-3 (transfected with pEF1-galec3-neo or pRC/CMV/G3, C1–C4; Figure 1). We present here the data obtained using nuclear clones N1 and N2 and cytoplasmic clones C1 and C2 (transfected with pEF1-galec3-neo). Western blotting experiments demonstrated a 30 kDa band in the C clones and a 35 kDa band in the N clones, corresponding to the Gal-3-(NLS)₃ fusion protein (Figure 1b). We performed nuclear and cytoplasmic-enriched fractions from the clones and in DU145 and PC-3 prostate cancer cell lines (Figure 1c). Comparison of Gal-3 protein levels confirmed the nuclear localization of the Gal-3 fusion protein in the N clones, with minor cytoplasmic localization, probably related to the enrichment process and to the fact that the Gal-3 fusion protein is synthesized in the cytoplasm. The N clones presented with weaker Gal-3 levels compared to those of endogenous nuclear Gal-3 from other prostate cancer cell lines, DU145 and PC-3. The cytoplasmic clones C1 and C2 are characterized by prominent localization of Gal-3 in the cytoplasm, with a minor fraction in the nucleus (Figure 1c). This latter observation is not unexpected as there is no reason to think that LNCaP cells are not able to perform shuttling of Gal-3 to the nucleus. Moreover, the nuclear Gal-3-(NLS)₃ fusion protein, like native Gal-3, could be exported from the nucleus to the cytoplasm by interaction between the leucin-rich nuclear export domain and the CRM1 protein (Tsay et al., 1999).

Figure 1.

Characterization of LNCaP stable transfectants. No staining is shown in LNCaP cells (a1) and M4 control clone (not shown); Gal-3 is localized in the nuclei of N clones (a2) and in the cytoplasm of C clones (a3; initial magnification: 400). (b) Western blotting of total cell extracts shows expression of the Gal-3-(NLS)₃ fusion protein (35 kDa) in the N clones, and of Gal-3 (30 kDa) in the C clones and A2058 melanoma cell line. Equal loading of the samples was demonstrated by similar detection of -tubulin (Tub, 50 kDa). (c) The Gal-3-(NLS)₃ fusion protein was detected mostly in the nuclear fractions of the N clones, with levels inferior to those observed in the DU145 (DU) and PC-3 (PC) cells. Conversely, Gal-3 was mostly detected in the cytoplasmic fractions of the C clones, with little amounts in the nuclear fractions. Adequate cell partitioning was confirmed by -tubulin localization in the cytoplasmic but not in the nuclear fractions. M4, negative control. (d) The Gal-3-(NLS)₃ fusion protein expressed in the N clones retains its lactose binding activity as demonstrated by Western blotting analysis of the elution fractions from lactose-Sepharose affinity chromatography, compared to the C clones expressing Gal-3. GFP was detected in the nuclei of the F1 clone by fluorescence microscopy (e; initial magnification: 400)

Full figure and legend (139K)

We further verified that the Gal-3-(NLS)₃ fusion protein expressed in the N clones conserved its ability to bind lactose, using affinity chromatography (Figure 1d). We selected a nuclear GFP clone (transfected with pCMV/myc/nuc/GFP; F1; Figure 1e), and clone M4 (wild-type pRC/CMV expression vector) as controls. We did not detect any differences in morphology, proliferation rates or adhesion to laminin-1 in all the clones examined (not shown). We did not find any difference in 6 and 1 integrin expression by flow cytometry in the clones (not shown), as Gal-3 increases 61 integrin cell surface expression in breast cancer cells (Warfield et al., 1997).

Gal-3 expression does not modify the androgen dependence status of LNCaP cells

Treatment of the cells for 24 h with 0, 5 and 500 nM dihydrotestosterone (DHT) did not demonstrate any differences in androgen receptor (AR) and prostate-specific antigen (PSA) levels between the various clones. PSA expression was null in all cells incubated without DHT. We observed a DHT-dependent expression of PSA and AR (data not shown). Thus, the expression of Gal-3, either in the cytoplasm or in the nucleus, did not modify the androgen dependence status of LNCaP cells.

Cytoplasmic Gal-3 promotes and nuclear Gal-3 decreases Matrigel invasion but not chick heart invasion

We found that cytoplasmic Gal-3 is associated to Matrigel invasion rates increased by at least 200%, compared to the LNCaP cell lines and control clones M4 and F1 (Figure 2). The number of invaded cells of the N clones was decreased by 60% compared to the controls. This effect is specific to nuclear Gal-3 since clone F1 (nuclear GFP) presents with similar levels of invasion compared to the LNCaP cells and the M4 control clone.

Figure 2.

Cytoplasmic Gal-3 (clones C1 and C2) promotes and nuclear Gal-3 (clones N1 and N2) inhibits Matrigel invasion, compared to parental LNCaP cells (LN), wild-type vector-transfected cells (M4) and a nuclear GFP clone (F1). Matrigel invasion is determined by counting the cells that invaded through Matrigel. The histogram depicts the results of a representative experiment. Three independent experiments were performed and four fields were considered for each clone in each experiment. ^*P<0.05 vs M4 (ANOVA-1 and Scheffe's test)

Full figure and legend (10K)

The various clones were also tested using the chick heart invasion assay. All clones were noninvasive, and no differences were observed using this model.

Cytoplasmic Gal-3 promotes and nuclear Gal-3 inhibits anchorage-independent growth

Gal-3 localization affects colony number but not colony size. Cytoplasmic Gal-3 is associated to increased formation of colonies in soft agar compared to the control clones (29619 and 28012 colonies for C1 and C2, respectively, vs 18810 colonies for M4; Figure 3), whereas lower numbers of colonies were observed in clones with nuclear Gal-3 (988 colonies; P<0.05). The number of colonies formed by nuclear GFP-transfected cells (1587) was not significantly different from that obtained with the mock-transfected control cells (P=0.15). The data demonstrate again a dual role for Gal-3, increasing the in vitro colony formation by LNCaP cells when expressed in the cytoplasm and decreasing it when expressed in the nucleus. This is not related to modified proliferation, as it is not affected by the transfections.

Figure 3.

Cytoplasmic Gal-3 (clones C1 and C2) increases and nuclear Gal-3 (clones N1 and N2) decreases anchorage-independent growth measured as the number of colonies formed in soft agar after 3 weeks, compared to wild-type vector-transfected cells (M4) and a nuclear GFP clone (F1). The histogram depicts the results of a representative experiment. Three independent experiments were performed and the colonies were counted in five dishes for each clone in each experiment. ^*P<0.05 vs M4 (ANOVA-1 and Scheffe's test)

Full figure and legend (7K)

Nuclear Gal-3 is proapoptotic while cytoplasmic Gal-3 is antiapoptotic

LNCaP clones presented with a similar baseline apoptosis level (data not shown), not modified by serum starvation. Apoptosis was induced by overnight incubation with 2 g/ml actinomycin D. Cytoplasmic Gal-3 protected LNCaP prostate cancer cells from apoptosis, decreasing actinomycin D-induced apoptosis by approximately 50% (P<0.05 for C1 and C2 clones compared to M4; Figure 4). These results confirm the antiapoptotic activity of Gal-3 (Yang et al., 1996; Akahani et al., 1997; Matarrese et al., 2000a; Moon et al., 2001; Song et al., 2002; Yoshii et al., 2002; Yu et al., 2002). Conversely, nuclear Gal-3 increased apoptosis in LNCaP cells by about 100% (P<0.0001 for N1 and N2, vs M4).

Figure 4.

Cytoplasmic Gal-3 (clones C1 and C2) decreases and nuclear Gal-3 (clones N1 and N2) promotes apoptosis, as demonstrated by TUNEL staining (a–e, g, left graph) and by annexin V–propidium iodide method in flow cytometry (f, g, right graph). Apoptosis was induced by either 2 g/ml actinomycin D (a–f), 10 ng/ml TNF- (g, left graph) or 8 Gy X-ray irradiation (g, right graph). LN, LNCaP; M4, control clone; F1, nuclear GFP clone. (a, b) Fluorescence microphotographs showing apoptotic cells detected by the TUNEL method for a nuclear Gal-3 clone (N1, panel a) and a cytoplasmic Gal-3 clone (C1, panel b) treated by actinomycin D (initial magnification: 100). (c, d) Corresponding DAPI counterstaining. Panel e is a quantification of the experiments shown in (a–d), and panel g (left chart) shows a similar experiment where apoptosis was induced by TNF-. Panels f and g (right chart) depict the percentage of early apoptotic cells quantified by annexin V–propidium iodide method in flow cytometry. Each experiment was performed at least three times. Each histogram depicts the results of a representative experiment. For each condition, five fields were considered for the TUNEL assay and samples of 10⁴ cells were acquired for the annexin V–propidium iodide method in flow cytometry. ^*P<0.05 vs M4 (ANOVA-1 and Scheffe's test)

Full figure and legend (172K)

Figure 4f shows similar results with the annexin V/propidium iodide method. Cytoplasmic Gal-3 also decreased actinomycin D-induced apoptosis, while nuclear Gal-3 increased it. Nuclear GFP expression did not modify induced apoptosis compared to the controls.

Induction of apoptosis by treatment with TNF- and by X-ray irradiation (Figure 4g) resulted in similar data.

Actinomycin D-induced apoptosis was associated with poly(ADP-ribose) polymerase (PARP) cleavage in cells expressing nuclear Gal-3, and to a lesser extent in mock- and GFP-transfected control cells, but not in cytoplasmic Gal-3 cells (Figure 5a). These data demonstrate that cytoplasmic Gal-3 decreases and nuclear Gal-3 increases caspase-dependent apoptosis.

Figure 5.

Cytoplasmic Gal-3 (clones C1 and C2) decreases and nuclear Gal-3 (clones N1 and N2) promotes PARP cleavage (a), caspase-8 activation (a, casp8) and caspase-9 activity (b). F1, nuclear GFP clone; M4, control clone; LN, LNCaP. Apoptosis is induced (i) or not (ni) by a 20 h, 2 g/ml actinomycin D treatment. (a) Western blot analysis showing the cleavage of PARP (116 kDa, cleaved into 85 and 25 kDa (not detected) fragments) and caspase-8 (proform, doublet 55/50 kDa, doublet of the cleaved forms 40/36 and 23 kDa (not shown)), and bcl-2 expression in 20 g of proteins from the clones. (b) Graphic representation of caspase-9 activity measured by colorimetric assay, showing the ratios of the absorbance at 405 nm (A405) in induced and noninduced conditions. The histogram depicts the results of a representative experiment. Three independent experiments were performed

Full figure and legend (61K)

The level of expression of the caspase-8 proform (55/50 kDa) was similar in all clones (Figure 5a). Only actinomycin D-induced nuclear Gal-3 clones presented with an active cleaved caspase-8 (Figure 5a). We did not find differences in caspase-8 activity between the clones, suggesting a lack of sensitivity of the assay (data not shown).

The 47 kDa proform of caspase-9 was lower after actinomycin D treatment in all clones (data not shown). No proteolysis was detected. Caspase-9 does not require proteolysis for activation, but binding to APAF-1 (Rodriguez and Lazebnik, 1999; Stennicke et al., 1999). Induction of caspase-9 activity by actinomycin D treatment was significantly increased in the nuclear Gal-3 clones, compared to controls (Figure 5b).

Decreased Bcl-2 expression was observed in all clones after actinomycin D (Figure 5a). All cells presented with a similarly increased level of Bax after actinomycin D (data not shown).

Gal-3 cellular localization affects in vivo tumor formation, apoptosis and vascular density

Two independent xenograft experiments in athymic mice gave similar results. The first experiment is detailed here, including M4, C2 and N1. In a third experiment, mice were injected with either N1, N2 and F1 in order to test the influence of nuclear Gal-3 and nuclear GFP (Figure 6).

Figure 6.

Cytoplasmic Gal-3 (clone C2) promotes and nuclear Gal-3 (clone N1) decreases primary tumor growth in nude mice, compared to the M4 control clone, by modulation of apoptosis and induced angiogenesis (see text for details). Left charts represent volumes (mm³) of tumors developed in nude mice injected subcutaneously with 1 10⁶ cells at day 0. Upper left insets in the graphs represent Gal-3 immunostaining (G3) in the primary tumors (initial magnification: 400). Center pictures represent, fluorescence microphotographs showing apoptotic cells at the border of the primary tumors as detected by the TUNEL method (initial magnification: 100). Right microphotographs show Factor VIII immunostaining in the tumors (initial magnification: 100). Histogram: Decreased tumor volumes observed in N clones are specific to nuclear Gal-3 as a nuclear GFP clone (F1) is characterized by tumor growth significantly higher than N clones, and similar to clone M4. The histogram depicts tumor volumes (means.d., mm³) observed 70 days after injection with clones N1, N2 and F1. n, number of tumors evaluated for each clone; ^*P<0.05 vs F1 (ANOVA-1 and Scheffe's test). Tumor volumes observed with clone F1 are not significantly different from those observed with control clone M4 (P>0.05)

Full figure and legend (330K)

The tumor take was similar for each clone (778% of injected sites). Cytoplasmic Gal-3 tumors were larger and nuclear Gal-3 tumors were smaller, compared to M4. The mean latency to reach 250 mm³ (Table 1) was significantly higher for N1 and N2 than that for C1 and C2 (P<0.02). We observed a nonsignificant trend for longer latency for the nuclear clones compared to controls (P=0.0561). The latency of the C clones was not significantly lower than controls (P=0.27). This could be due to the fact that tumors started to grow after variable times. After 70 days, the tumors observed in N1 and N2 mice appeared later and were smaller than those from F1 (43.220.9 and 41.223.2, vs 148.990.3, respectively; P<0.05; Figure 6). As expected, the tumor volumes obtained with F1 are similar to those observed with M4 (P>0.05).

Table 1 - Dynamics of tumor formation.

Full table (19K)

Gal-3 expression was maintained in all tumors (Figure 6), but some cells lost their expression, as expected from the absence of the G418 selection.

Higher levels of apoptosis were detected at the borders of tumors developed in mice injected with nuclear Gal-3 clones (8219 apoptotic cells by microscope field for N1), compared to other cells (21 and 11 for M4 and C2; P<0.0001; Figure 6). This could contribute to the smaller tumor sizes in the N clones.

Gal-3 is reported to be angiogenic (Nangia-Makker et al., 2000).

Vascularization (detected by Factor VIII) was highest in cytoplasmic Gal-3 tumors (C1: 2.870.35; C2: 2.930.25) compared to controls (LNCaP: 1.830.46; M4: 1.700.60), whereas nuclear Gal-3 tumors were characterized by barely detectable vessels (N1: 0.430.57; N2: 0.570.63; P<0.05 for all clones but LNCaP, vs M4). The same observation was made with a subset of tumors of similar size (not shown). Prominent macroscopic vascularization was observed around the cytoplasmic Gal-3 (data not shown).

Finally, in the second experiment, four cytoplasmic Gal-3 tumors did locally invade the vertebrae of the mice. No metastases were observed neither macroscopically nor microscopically for any clone.

Discussion

Gal-3 expression is altered in cancer cells and appears to participate in the acquisition of the malignant phenotype. We demonstrated that Gal-3 exerts opposite activities whether it is expressed in the cytoplasm or nucleus of the LNCaP prostate cancer cell line. Cytoplasmic Gal-3 transfectants exhibit a significant enhancement in their in vitro and in vivo aggressiveness, as reported (Raz et al., 1990; Nangia-Makker et al., 1995; Akahani et al., 1997; Warfield et al., 1997; Matarrese et al., 2000a). We demonstrate, for the first time, that nuclear Gal-3 has a major negative impact on the malignant capacities of the cancer cells. This is in agreement with the observation that prostate cancer cells expressing Gal-3 in the cytosol are characterized by poor prognosis, compared to cells expressing lectin in both nucleus and cytoplasm (van den Brûle et al., 2000). The cytoplasmic clones (C) are characterized by minor amounts of Gal-3 in their nucleus, probably because LNCaP cells can import Gal-3 in the nucleus. This phenomenon could contribute to somewhat attenuate the phenotypes observed in these clones. Yet, none of the tested C clones (C1–C4), either grown in vitro or in vivo, neither showed a conspicuous nuclear immunostaining for Gal-3 nor behaved like clones with essentially nuclear Gal-3 (N) in any of our experiments. Conversely, the nuclear clones (N) presented with minor amounts of cytoplasmic Gal-3-[NLS]₃, because of cytoplasmic synthesis and maybe exit from the nucleus to the cytoplasm (Tsay et al., 1999).

Increased colony formation by cytoplasmic Gal-3 transfectants is in agreement with other studies (Honjo et al., 2001; Yoshii et al., 2001). This could be explained by decreased anoïkis (Kim et al., 1999) and increased surface expression of 61 integrin (Warfield et al., 1997), the latter observation not shown here. Cytoplasmic Gal-3 promotes, like exogenous Gal-3 (Le Marer and Hughes, 1996), while nuclear Gal-3 inhibits in vitro Matrigel invasion. Although 61 integrin is associated with invasive prostate carcinoma in vivo (Cress et al., 1995), the increased invasive phenotype of cytosolic Gal-3 transfectants cannot be related to modified cell surface expression of 61, as our clones present with similar cell surface levels of 6 and 1 and similar adhesion rates to laminin-1 (data not shown). Chick heart invasion was not affected by Gal-3 localization. It is not the first time that cancer cells invade Matrigel but are considered as noninvasive in the chick heart invasion assay (Le Marer and Bruyneel, 1996). The Matrigel invasion assay and the chick heart invasion assay are two different models that provide specific results. The Matrigel invasion assay is sensitive to the intrinsic invasiveness of the cells; the chick heart invasion assay also takes into account the influence of factors from living host cells.

Inhibition of apoptosis rather than enhanced cellular proliferation is the critical pathophysiological factor that contributes to the development of prostatic adenocarcinoma (Gurumurthy et al., 2001). Cytoplasmic Gal-3 is associated with decreased susceptibility to apoptosis, while nuclear Gal-3 leads to increased induced apoptosis. The latter observation is compatible with the recent report that leptomycin B increased the cisplatin-induced apoptosis of Gal-3-expressing BT-549 breast cancer cells (Takenaka et al., 2004). This suggests a dual role for Gal-3 in apoptosis regulation, depending on its localization. The antiapoptotic activity of cytoplasmic Gal-3 (Yang et al., 1996; Akahani et al., 1997; Matarrese et al., 2000a; Moon et al., 2001; Song et al., 2002; Yoshii et al., 2002; Yu et al., 2002) could be mediated by the NWGR motif shared with Bcl-2, as shown by mutation (Akahani et al., 1997). In our study, all transfectants, when treated with actinomycin D, underwent similar modulations of Bcl-2 and Bax expression, as expected (Akahani et al., 1997). Cytoplasmic Gal-3 abolished PARP cleavage, probably by inhibition of caspase activation. However, neither significant differences in caspase-8 and -9 cleavages nor in their specific activities were observed between LNCaP cells expressing cytoplasmic Gal-3 and controls. In BT-549 breast cancer cells, Gal-3 inhibits the intrinsic apoptotic pathway (Akahani et al., 1997; Kim et al., 1999; Matarrese et al., 2000a; Moon et al., 2001) but not the extrinsic pathway (unpublished data cited in Moon et al., 2001). However, Gal-3 counteracts TNF--induced apoptosis, involving the extrinsic pathway, in another breast cancer cell line (Matarrese et al., 2000b). In LNCaP cells, cytoplasmic Gal-3 inhibits both intrinsic (actinomycin D or X-ray irradiation) and extrinsic (TNF-) apoptotic pathways. Nuclear Gal-3 significantly increased PARP cleavage, caspase-8 activation and caspase-9 activity. Caspase-8 cleavage was not expected since actinomycin D should involve the caspase-8-independent intrinsic pathway. However, the intrinsic and extrinsic pathways appear interdependent in LNCaP cells (Nesterov et al., 2001).

Cytoplasmic Gal-3 transfectants were associated with development of larger and earlier tumors than controls, as expected (Nangia-Makker et al., 1995; Honjo et al., 2001). Conversely, nuclear Gal-3 tumors were smaller and grew more slowly. This effect was specific to nuclear Gal-3, as nuclear Gal-3 clones induced smaller tumors than a nuclear GFP clone (Figure 6). A recent study demonstrated that Gal-3-transfected LNCaP formed tumors at a slower rate than control lines (Ellerhorst et al., 2002). These apparently conflicting results could likely be due to low Gal-3 expression in the examined clones, which is even lost after injection to the animal. The promoting effect of cytoplasmic Gal-3 on LNCaP tumor development could be related to increased angiogenesis (Nangia-Makker et al., 2000) and decreased apoptosis. Conversely, nuclear Gal-3 could decrease tumor formation by decreasing angiogenesis and increasing apoptosis.

Besides confirming the tumorigenic role of cytoplasmic Gal-3, our data indicate that nuclear Gal-3 decreases the invasive phenotype in prostate adenocarcinoma cells. A few nuclear functions have already been reported for Gal-3. Gal-3 binds to nucleic acids (Wang et al., 1995) and is involved in mRNA splicing (Dagher et al., 1995). It interacts with Gemin4 and both are components of a macromolecular complex designated as the survival of motor neuron (SMN) complex (Park et al., 2001). Thus, Gal-3 might associate with Gemin4 in the cytoplasm, possibly during the course of snRNP biogenesis, and be imported to the nucleus. Gal-3 may be taken along by way of its interaction with Gemin4. Once in the nucleus, Gal-3 may participate with the SMN complex in the assembly of the spliceosome and, later, be exported from the nucleus, in association with Gemin4 and possibly other members of the SMN complex, to repeat the cycle of snRNP biogenesis and delivery (Davidson et al., 2002).

A novel function of nuclear Gal-3 in the regulation of gene transcription has been suggested, based on the observation that Gal-3 enhances cyclin D₁ promoter activity through SP1 and a cAMP-responsive element in human breast epithelial cells (Lin et al., 2002). In transformed thyroid cells, Gal-3 upregulates the transcription factor TTF-1 transcriptional activity, most probably through the interaction between Gal-3 and TTF-1 homeodomain (Paron et al., 2003). Since TTF-1 is a member of the NK class of homeodomain-containing transcription factors (Guazzi et al., 1990) and these proteins show a high conservation of the homeodomains, Gal-3 could be able to interact with other members of the Nkx homeodomain family. Another member of the Nkx family, Nkx3.1, has been found in the nucleus (Bowen et al., 2000; Korkmaz et al., 2000) of both normal (Bieberich et al., 1996; Gelmann et al., 2003) and prostate cancer cells (Bowen et al., 2000; Gelmann et al., 2003), including the LNCaP cell line (Bowen et al., 2000). As Nkx3.1 is considered as a prostate tumor suppressor (Bhatia-Gaur et al., 1999; Bowen et al., 2000; Abdulkadir et al., 2002; Kim et al., 2002), we postulated that, in the nucleus, Gal-3 could decrease the expression of the cancer phenotype by interacting with Nkx3.1, but we were unable to demonstrate an interaction between these proteins by co-immunoprecipitation (data not shown). Anyhow, a deeper investigation of the nuclear components that interact with Gal-3 is necessary to understand better the mechanisms of action of nuclear Gal-3.

Our experimental results based on the LNCaP prostate cancer cell line are in perfect agreement with our previous observations that cancer cells of primary tumors from endometrium, colon and prostate are characterized by progression patterns depending on Gal-3 localization (van den Brûle et al., 1996; 2000; Sanjuan et al., 1997). However, primary tumors are heterogeneous, and emergence of a subset of more aggressive cancer cells provides an explanation for tumor progression (Hanahan and Weinberg, 2000). Our above-mentioned studies classify tumors using the most prominent pattern of Gal-3 expression, but do not allow to examine the value of the heterogeneity of Gal-3 cellular localization for prediction of prognosis. Hence, it is possible that the overall behavior of a primary tumor depends on the prominent pattern of Gal-3 localization, and also on a minority of subclones that could account for some aspects of tumor progression. Moreover, the present study does not rule out that some still unknown stimuli could affect Gal-3 localization in the clones, and modify the phenotype of a subset of cancer cells.

In conclusion, this study, together with the previous observation that Gal-3 was consistently excluded from the nucleus in prostate cancer cells but not in nontumoral prostatic glands (van den Brûle et al., 2000), indicates that Gal-3 can play antitumor activities when present in the nucleus, whereas it can favor tumor progression when expressed in the cytoplasm. It is generally admitted that primary cancer are heterogenous and that only a subset of malignant cells are able to successfully invade and disseminate. Our study suggests that nuclear exclusion of Gal-3 is a key factor for the acquisition of the invasive and metastatic phenotype while cells from the tumor with nuclear Gal-3 expression would be less aggressive. The molecular mechanisms underlying this observation remain to be investigated, and could constitute the basis of future anticancer therapeutic strategies.

Material and methods

Cell culture and reagents

Cell lines were obtained from the American Type Culture Collection, Rockville, MD, USA, and the culture reagents from GibcoBRL, Merelbeke, Belgium. The LNCaP-FG human prostate cancer cell line was grown in RPMI 1640 supplemented with 10% heat-inactivated fetal bovine serum (FBS), 10 mM Hepes and 1 mM sodium pyruvate. The DU145 prostate carcinoma cell line was grown in MEM supplemented with 10% FBS and 2 mM glutamine, and the PC-3 cell line was grown in Ham's F12 K medium supplemented with 7% FBS. All cell lines were maintained at 37°C in a water-saturated 95% air–5% carbon dioxide atmosphere.

Plasmid contructions and transfections

We used several expression vectors in our experiments to target Gal-3 to the cell nucleus (pCMV/myc/nuc/G3, based on pCMV/myc/nuc, Invitrogen, Merelbeke, Belgium) or cytoplasm (pRC/CMV/G3 (van den Brûle et al., 1997) and pEF1-galec3-neo (Yang et al., 1998), a gift from Dr Fu-Tong Liu, UC Davis, Sacramento, CA, USA). Transfection controls were performed with pRC/CMV (Invitrogen); resulting clones were checked for genomic incorporation of the intact CMV promoter by PCR using primers 5'-ATAGCGGTTTGACTCACG and CCAGAATAGAATGACACC-3'. The pCMV/myc/nuc/GFP vector (Invitrogen) provided a control for targeting GFP, a Gal-3-unrelated protein, to the nucleus.

The full-length cDNA encoding the human Gal-3 was subcloned from plasmid 2.2 (Robertson et al., 1990) into the pCMV/myc/nuc vector, which contains three in-frame NLSs downstream the polylinker for C-terminal fusion to the recombinant Gal-3. Briefly, a human 750-bp Gal-3 cDNA was PCR-amplified using specific primers (forward: 5'-AACTGCAGATGGCAGACAATTTTTCGCTC-3'; reverse: 5'-TTCTGCAGTATCATGGTATATGAAGCACT-3'), subcloned in pCR^®2.1-TOPO (Invitrogen), ligated to pCMV/myc/nuc and sequenced.

Half confluent LNCaP monolayers in 100 mm dishes were transfected using linearized plasmid and the DOTAP reagent (Roche, Vilvoorde, Belgium). Selection was performed by addition of 500 g/ml G418 (Roche).

Western blotting

Expression of Gal-3, -tubulin, PSA, AR, caspase-8, caspase-9, Bcl-2, Bax and cleavage of PARP in the LNCaP clones were analysed by Western blotting, using protein lysates in 1% SDS and SDS–PAGE. We used monoclonal antibodies against Gal-3 (Ho and Springer, 1982), prostate-specific antigen (BioGenex, San Ramon, CA, USA), -tubulin (Sigma, Bornem, Belgium), PARP, caspase-9 (BD Pharmingen, Erembodegem, Belgium), Bcl-2 (Santa Cruz, Tebu-Bio, Boeckout, Belgium) and Bax (Oncogene, VWR, Leuven, Belgium), and polyclonal antisera raised against Gal-3 (Frigeri and Liu, 1992), human androgen receptor (Novocastra Laboratories Ltd, Newcastle upon Tyne, UK) and caspase-8 (BD Pharmingen). The membranes were developed by chemiluminescence (ECL, Amersham Pharmacia, Rosendaal, The Netherlands). After transfer, membranes and gels were stained respectively with Ponceau red and Coomassie blue to verify adequate loading. Densitometric analysis was performed using the public domain NIH Image program (http://rsb.info.nih.gov/nih-im
age/).

Immunocytochemistry and immunohistochemistry

Immunohistochemical staining was performed with an anti-Gal-3 (Frigeri and Liu, 1992; van den Brûle et al., 2000) and anti-Factor VIII polyclonal antiserum (DakoCytomation, Heverlee, Belgium) as described (van den Brûle et al., 1996). Preconfluent cultures of the LNCaP clones in Permanox chamberslides were fixed in methanol:acetone 50 : 50 for 30 min at 4°C.

Preparation of nuclear and cytoplasmic protein extracts

Pellets of trypsinized cells were lysed in 10 mM Hepes, 20 mM KCl, 2 mM MgCl₂, 0.1 mM. EDTA, 1 mM DTT, 10 mM KCl, 0.2% NP-40 and protease inhibitors, pH 7.9. After centrifugation, the cytoplasmic supernatant was stored at -80°C. The nuclear pellet was resuspended in extraction buffer (20 mM Hepes, 1.5 mM MgCl₂, 0.2 mM EDTA, 630 mM NaCl, 25% glycerol, 0.5 mM DTT and protease inhibitors, pH 7.9). After centrifugation, the nuclear supernatant was stored at –80°C before Western blotting. Presence of cytoplasmic proteins was verified in all samples by detection of -tubulin.

Affinity chromatography

Cell lysates (in 25 mM Tris, 150 mM NaCl, 0.5%. NP-40 and 0.1 mM PMSF, pH 7.4) were incubated with lactose-Sepharose 4B (Amersham Pharmacia) under agitation at 4°C. After three washes with the same buffer without NP-40, the matrix was resuspended in Laemmli sample buffer before Western blotting.

Cell proliferation

T25 culture flasks seeded at 5000 cells/cm² were counted after 96 h. Trypan blue exclusion showed that cell viability was consistently above 90%. The doubling time T_c of the various clones was calculated using the formula T_c=0.3 T/log A_T/A₀, where A is the number of cells at time T and A₀ is the initial number of cells (Wieder, 1999).

Attachment to laminin-1

Attachment assays were performed as described (van den Brûle et al., 1995).

Androgen dependence and expression of PSA and AR

Cells grown for 48 h in phenol red-free medium containing charcoal-stripped serum were treated with DHT (Sigma) 0, 5 and 500 nM for 24 h.

Anchorage-independent. growth

Anchorage-independent growth assays were based on the ability of cancer cells to form colonies in soft agar as previously described (Gilles et al., 1993). The colonies larger than 15 cells after 3 weeks were counted under optical microscope. We also evaluated the size of the colonies obtained with the different clones.

Matrigel invasion

Matrigel invasion assays were performed in Biocoat Matrigel Invasion Chambers (Becton Dickinson, Erembodegem, Belgium) as recommended by the manufacturer. All experiments were conducted in duplicate and reproduced at least three times.

Chick heart invasion assay

This assay is based on the confrontation of precultured chick heart fragments with aggregates of cells to be tested (Bracke et al., 2001).

Apoptosis

Apoptosis was quantified by the TUNEL method (In Situ Cell Death Detection Kit, Fluorescein Roche) and annexin V–propidium iodide staining (Locigno et al., 2002). Apoptosis was induced by overnight treatment with 2 g/ml actinomycin D (Sigma) (Kleeff et al., 2000). We also used a 24 h treatment with 10 ng/ml TNF- (Sigma) (Muenchen et al., 2000) or 8 Gy X-ray irradiation (Kyprianou et al., 1997), 48 h before apoptosis detection.

Caspase activity assays

Caspase-8 and -9 activities were measured by colorimetric assays based on spectrophotometric detection of the chromophore p-nitroanilide (pNA) after cleavage from the labeled substrates IETD-pNA and LEHD-pNA, respectively (BioVision, Gentaur, Brussels, Belgium), as recommended.

In vivo xenograft tumor formation in mice

Male 6–7 weeks Balb/c athymic nude mice (Charles River Laboratories, Brussels, Belgium) were housed in specific pathogen-free conditions (Ethical Committee of the University of Liège, F2001/8). Three independent experiments were performed, using five to six mice for each LNCaP clone. Each mouse was injected subcutaneously at the posterolateral body wall, on each side, with 10⁶ cells in 100 l RPMI 1640 medium containing 50% Matrigel (Kleinman et al., 1986). Mice were weighed and tumors measured with a caliper twice a week. Tumor volumes were calculated using the formula 0.4 length width² (Noel et al., 1992). Animals were euthanized using a CO₂ chamber, if any tumor reached the size of 1.8 cm for the long diameter and/or moribund appearance, and the tumors were dissected for further analysis.

Statistical analysis

Statistical analysis of the results was performed with the StatView 5.0. software (SAS Institute Inc., Cary, NC, USA).

References

1. Abdulkadir SA, Magee JA, Peters TJ, Kaleem Z, Naughton CK, Humphrey PA and Milbrandt J. (2002). Mol. Cell. Biol., 22, 1495–1503. | Article | PubMed | ISI | ChemPort |
2. Akahani S, Nangia-Makker P, Inohara H, Kim HR and Raz A. (1997). Cancer Res., 57, 5272–5276. | PubMed | ISI | ChemPort |
3. Bhatia-Gaur R, Donjacour AA, Sciavolino PJ, Kim M, Desai N, Young P, Norton CR, Gridley T, Cardiff RD, Cunha GR, Abate-Shen C and Shen MM. (1999). Genes Dev., 13, 966–977. | PubMed | ISI | ChemPort |
4. Bieberich CJ, Fujita K, He WW and Jay G. (1996). J. Biol. Chem., 271, 31779–31782. | Article | PubMed | ISI | ChemPort |
5. Bowen C, Bubendorf L, Voeller HJ, Slack R, Willi N, Sauter G, Gasser TC, Koivisto P, Lack EE, Kononen J, Kallioniemi OP and Gelmann EP. (2000). Cancer Res., 60, 6111–6115. | PubMed | ISI | ChemPort |
6. Bracke ME, Boterberg T and Mareel MM. (2001). Methods in Molecular Medicine Vol. 58. Schumacher SABaU (ed). Humana Press Inc.: Totowa, NJ, pp. 91–102.
7. Califice S, Castronovo V and van den Brule FA. Int. J. Oncol. (in press).
8. Cowles EA, Agrwal N, Anderson RL and Wang JL. (1990). J. Biol. Chem., 265, 17706–17712. | PubMed | ISI | ChemPort |
9. Cress AE, Rabinovitz I, Zhu W and Nagle RB. (1995). Cancer Metast. Rev., 14, 219–228.
10. Dagher SF, Wang JL and Patterson RJ. (1995). Proc. Natl. Acad. Sci. USA, 92, 1213–1217. | Article | PubMed | ChemPort |
11. Danguy A, Camby I and Kiss R. (2002). Biochim. Biophys. Acta., 1572, 285–293. | Article | PubMed | ISI | ChemPort |
12. Davidson PJ, Davis MJ, Patterson RJ, Ripoche MA, Poirier F and Wang JL. (2002). Glycobiology, 12, 329–337. | Article | PubMed | ISI | ChemPort |
13. Ellerhorst JA, Stephens LC, Nguyen T and Xu XC. (2002). Prostate, 50, 64–70. | Article | PubMed | ISI | ChemPort |
14. Frigeri LG and Liu FT. (1992). J. Immunol., 148, 861–867. | PubMed | ISI | ChemPort |
15. Gaudin JC, Mehul B and Hughes RC. (2000). Biol. Cell., 92, 49–58. | Article | PubMed | ChemPort |
16. Gelmann EP, Bowen C and Bubendorf L. (2003). Prostate, 55, 111–117. | Article | PubMed | ChemPort |
17. Gilles C, Piette J, Rombouts S, Laurent C and Foidart JM. (1993). Int. J. Cancer., 53, 872–879. | PubMed |
18. Gritzmacher CA, Robertson MW and Liu FT. (1988). J. Immunol., 141, 2801–2806. | PubMed |
19. Guazzi S, Price M, De Felice M, Damante G, Mattei MG and Di Lauro R. (1990). EMBO J., 9, 3631–3639. | PubMed | ISI | ChemPort |
20. Gurumurthy S, Vasudevan KM and Rangnekar VM. (2001). Cancer Metast. Rev., 20, 225–243. | Article |
21. Hanahan D and Weinberg RA. (2000). Cell, 100, 57–70. | Article | PubMed | ISI | ChemPort |
22. Herrmann J, Turck CW, Atchison RE, Huflejt ME, Poulter L, Gitt MA, Burlingame AL, Barondes SH and Leffler H. (1993). J. Biol. Chem., 268, 26704–26711. | PubMed |
23. Ho MK and Springer TA. (1982). J. Immunol., 128, 1221–1228. | PubMed | ISI | ChemPort |
24. Honjo Y, Inohara H, Akahani S, Yoshii T, Takenaka Y, Yoshida J, Hattori K, Tomiyama Y, Raz A and Kubo T. (2000). Clin. Cancer Res., 6, 4635–4640. | PubMed | ISI | ChemPort |
25. Honjo Y, Nangia-Makker P, Inohara H and Raz A. (2001). Clin. Cancer Res., 7, 661–668. | PubMed | ISI | ChemPort |
26. Hughes RC. (2001). Biochimie, 83, 667–676. | Article | PubMed | ISI | ChemPort |
27. Kim HR, Lin HM, Biliran H and Raz A. (1999). Cancer Res., 59, 4148–4154. | PubMed | ISI | ChemPort |
28. Kim MJ, Bhatia-Gaur R, Banach-Petrosky WA, Desai N, Wang Y, Hayward SW, Cunha GR, Cardiff RD, Shen MM and Abate-Shen C. (2002). Cancer Res., 62, 2999–3004. | PubMed | ISI | ChemPort |
29. Kleeff J, Kornmann M, Sawhney H and Korc M. (2000). Int. J. Cancer, 86, 399–407. | Article | PubMed | ISI | ChemPort |
30. Kleinman HK, McGarvey ML, Hassell JR, Star VL, Cannon FB, Laurie GW and Martin GR. (1986). Biochemistry, 25, 312–318. | Article | PubMed | ISI | ChemPort |
31. Korkmaz KS, Korkmaz CG, Ragnhildstveit E, Kizildag S, Pretlow TG and Saatcioglu F. (2000). Gene, 260, 25–36. | Article | PubMed |
32. Kyprianou N, King ED, Bradbury D and Rhee JG. (1997). Int. J. Cancer, 70, 341–348. | Article | PubMed | ISI | ChemPort |
33. Le Marer N and Bruyneel E. (1996). Anticancer Res., 16, 2767–2772. | PubMed |
34. Le Marer N and Hughes RC. (1996). J. Cell. Physiol., 168, 51–58. | Article | PubMed | ChemPort |
35. Lin HM, Pestell RG, Raz A and Kim HR. (2002). Oncogene, 21, 8001–8010. | Article | PubMed | ISI | ChemPort |
36. Locigno R, Pincemail J, Henno A, Treusch G and Castronovo V. (2002). Int. J. Oncol., 20, 69–75. | PubMed |
37. Lotz MM, Andrews Jr CW, Korzelius CA, Lee EC, Steele Jr GD, Clarke A and Mercurio AM. (1993). Proc. Natl. Acad. Sci. USA, 90, 3466–3470. | PubMed | ChemPort |
38. Matarrese P, Fusco O, Tinari N, Natoli C, Liu FT, Semeraro ML, Malorni W and Iacobelli S. (2000a). Int. J. Cancer, 85, 545–554. | Article | PubMed | ISI | ChemPort |
39. Matarrese P, Tinari N, Semeraroa ML, Natolib C, Iacobelli S and Malorni W. (2000b). FEBS Lett., 473, 311–315. | Article | PubMed | ISI | ChemPort |
40. Moon BK, Lee YJ, Battle P, Jessup JM, Raz A and Kim HR. (2001). Am. J. Pathol., 159, 1055–1060. | PubMed | ISI | ChemPort |
41. Moutsatsos IK, Wade M, Schindler M and Wang JL. (1987). Proc. Natl. Acad. Sci. USA, 84, 6452–6456. | Article | PubMed | ChemPort |
42. Muenchen HJ, Lin DL, Walsh MA, Keller ET and Pienta KJ. (2000). Clin. Cancer Res., 6, 1969–1977. | PubMed | ISI | ChemPort |
43. Nangia-Makker P, Honjo Y, Sarvis R, Akahani S, Hogan V, Pienta KJ and Raz A. (2000). Am. J. Pathol., 156, 899–909. | PubMed | ISI | ChemPort |
44. Nangia-Makker P, Thompson E, Hogan C, Ochieng J and Raz A. (1995). Int. J. Oncol., 7, 1079–1087.
45. Nesterov A, Lu X, Johnson M, Miller GJ, Ivashchenko Y and Kraft AS. (2001). J. Biol. Chem., 276, 10767–10774. | Article | PubMed | ISI | ChemPort |
46. Noel A, Simon N, Raus J and Foidart JM. (1992). Biochem. Pharmacol., 43, 1263–1267. | Article | PubMed | ISI |
47. Park JW, Voss PG, Grabski S, Wang JL and Patterson RJ. (2001). Nucleic. Acids Res., 29, 3595–3602. | Article | PubMed | ISI | ChemPort |
48. Paron I, Scaloni A, Pines A, Bachi A, Liu FT, Puppin C, Pandolfi M, Ledda L, Di Loreto C, Damante G and Tell G. (2003). Biochem. Biophys. Res. Commun., 302, 545–553. | Article | PubMed | ISI | ChemPort |
49. Raz A, Zhu DG, Hogan V, Shah N, Raz T, Karkash R, Pazerini G and Carmi P. (1990). Int. J. Cancer, 46, 871–877. | PubMed |
50. Robertson MW, Albrandt K, Keller D and Liu FT. (1990). Biochemistry, 29, 8093–8100. | Article | PubMed | ChemPort |
51. Rodriguez J and Lazebnik Y. (1999). Genes Dev., 13, 3179–3184. | Article | PubMed | ISI | ChemPort |
52. Sanjuan X, Fernandez PL, Castells A, Castronovo V, van den Brule F, Liu FT, Cardesa A and Campo E. (1997). Gastroenterology, 113, 1906–1915. | PubMed | ISI | ChemPort |
53. Sato S and Hughes RC. (1994). J. Biol. Chem., 269, 4424–4430. | PubMed | ISI | ChemPort |
54. Song YK, Billiar TR and Lee YJ. (2002). Am. J. Pathol., 160, 1069–1075. | PubMed | ISI | ChemPort |
55. Stennicke HR, Deveraux QL, Humke EW, Reed JC, Dixit VM and Salvesen GS. (1999). J. Biol. Chem., 274, 8359–8362. | Article | PubMed | ISI | ChemPort |
56. Takenaka Y, Fukumori T, Yoshii T, Oka N, Inohara H, Kim HR, Bresalier RS and Raz A. (2004). Mol. Cell Biol., 24, 4395–4406. | Article | PubMed | ISI | ChemPort |
57. Tsay YG, Lin NY, Voss PG, Patterson RJ and Wang JL. (1999). Exp. Cell Res., 252, 250–261. | Article | PubMed |
58. van den Brûle FA, Bellahcène A, Jackers P, Liu F-T, Sobel ME and Castronovo V. (1997). Int. J. Oncol., 11, 261–264.
59. van den Brûle FA, Buicu C, Baldet M, Sobel ME, Cooper DN, Marschal P and Castronovo V. (1995). Biochem. Biophys. Res. Commun., 209, 760–767. | Article | PubMed |
60. van den Brûle FA, Buicu C, Berchuck A, Bast RC, Deprez M, Liu FT, Cooper DN, Pieters C, Sobel ME and Castronovo V. (1996). Hum. Pathol., 27, 1185–1191. | PubMed |
61. van den Brûle F and Castronovo V. (2000). Lectins and Pathology Caron M and Sève A-P (eds). Harwood Academic Publishers: Australia.
62. van den Brûle FA, Waltregny D, Liu FT and Castronovo V. (2000). Int. J. Cancer, 89, 361–367. | Article | PubMed | ISI | ChemPort |
63. Wang L, Inohara H, Pienta KJ and Raz A. (1995). Biochem. Biophys. Res. Commun., 217, 292–303. | Article | PubMed | ISI | ChemPort |
64. Warfield PR, Makker PN, Raz A and Ochieng J. (1997). Invasion. Metast., 17, 101–112.
65. Wieder R. (1999). Cell Growth, Differentiation and Senescence: A Practical Approach Studzinski G (ed). Oxford University Press: New York, pp. 1–32.
66. Yang RY, Hill PN, Hsu DK and Liu FT. (1998). Biochemistry, 37, 4086–4092. | Article | PubMed | ISI | ChemPort |
67. Yang RY, Hsu DK and Liu FT. (1996). Proc. Natl. Acad. Sci. USA, 93, 6737–6742. | Article | PubMed | ChemPort |
68. Yoshii T, Fukumori T, Honjo Y, Inohara H, Kim HR and Raz A. (2002). J. Biol. Chem., 277, 6852–6857. | Article | PubMed | ISI | ChemPort |
69. Yoshii T, Inohara H, Takenaka Y, Honjo Y, Akahani S, Nomura T, Raz A and Kubo T. (2001). Int. J. Oncol., 18, 787–792. | PubMed | ISI | ChemPort |
70. Yu F, Finley Jr RL, Raz A and Kim HR. (2002). J. Biol. Chem., 277, 15819–15827. | Article | PubMed | ISI | ChemPort |

Acknowledgements

We thank Fu-Tong Liu and Ri-Yao Yang (UC Davis, Sacramento, CA, USA) for the pEF1-galec3-neo plasmid, Françoise Rentier for expert advices in cloning strategy, Rosita Winkler and David Waltregny, for fruitful discussions, Rita Colman and Daan Vandekerckhove for performing the chick heart invasion assays, and Pascale Heneaux, Naïma Maloujahmoum, Sabine Thonard, Jacques Foguenne and Guy Roland for expert technical assistance. Stéphane Califice is the recipient of a Télévie Grant, and Frédéric van den Brûle is a Senior Research Associate of the National Fund for Scientific Research, Belgium. This work has been partially supported by the National Fund for Scientific Research, the Interuniversity Attraction Pole Program, the Association Sportive contre le Cancer, Télévie and the Léon Frédéricq Foundation (Belgium).