Transformation of Human Epithelial Cells via Modulation of Unique Sets of Genes

Galectin-3 is a multifunctional beta-galactoside-binding protein implicated in apoptosis, malignant transformation, and tumor progression. The mechanisms by which galectin-3 contributes to malignant progression are not fully understood. In this study, we found that the introduction of wild-type galectin-3 into nontumorigenic, galectin-3-null BT549 human breast epithelial cells conferred tumorigenicity and metastatic potential in nude mice, and that galectin-3 expressed by the cells was phosphorylated. In contrast, BT549 cells expressing galectin-3 incapable of being phosphorylated (Ser6-->Glu Ser6-->Ala) were nontumorigenic. A microarray analysis of 10,000 human genes, comparing BT549 transfectants expressing wild-type and those expressing phosphomutant galectin-3, identified 188 genes that were differentially expressed (>2.5-fold). Genes affected by introduction of wild-type phosphorylated but not phosphomutant galectin-3 included those involved in oxidative stress, a novel noncaspase lysosomal apoptotic pathway, cell cycle regulation, transcriptional activation, cytoskeleton remodeling, cell adhesion, and tumor invasion. The reliability of the microarray data was validated by real-time reverse transcription-PCR (RT-PCR) and by Western blot analysis, and clinical relevance was evaluated by real-time RT-PCR screening of a panel of matched pairs of breast tumors. Differentially regulated genes in breast cancers that are also predicted to be associated with phospho-galectin-3 in transformed BT549 cells include C-type lectin 2, insulin-like growth factor-binding protein 5, cathepsins L2, and cyclin D1. These data show the functional diversity of galectin-3 and suggest that phosphorylation of the protein is necessary for regulation (directly or indirectly) of unique sets of genes that play a role in malignant transformation.


Introduction
Galectins are a family of carbohydrate-binding proteins that have been well conserved through evolution and are abundant in epithelial and immune cells of animals (1,2).These proteins have sequence similarities in the carbohydrate recognition domain and specificity for h-galactoside moieties found on both N-and Olinked glycans (3,4).To date, 14 members of the galectin family have been identified and classified into proto, chimera, and tandem-repeat types according to their structure (5).Galectin-3 is a 31 kDa chimeric gene product consisting of three distinct structural domains: a short NH 2 -terminal domain of 12 amino acids, which controls its cellular targeting; a repetitive collagen-like sequence rich in glycine, tyrosine, and proline, which serves as a substrate for matrix metalloproteinases; and a carboxy-terminal domain with a globular structure encompassing the carbohydratebinding site (6)(7)(8).
Galectin-3 undergoes phosphorylation at Ser 6 by casein kinase 1 and dephosphorylation by protein phosphatase 1 (24,25).This posttranslational modification signaling acts as an ''on/off '' switch for its carbohydrate-binding capability (25).Recent studies have shown that phosphorylation of nuclear galectin-3 at Ser 6 is critical for its transport from the nucleus and for its antiapoptotic functions (26,27).Here, we present evidence that phosphorylation of galectin-3 also plays a key role in malignant transformation and is associated with the activation of specific sets of genes and pathways.This work provides new insights into the molecular mechanisms underlying the role of galectin-3 in tumorigenesis and metastasis.
Mice.Female athymic NCr-nu mice were purchased from the National Cancer Institute-Frederick Cancer Research Facility (Frederick, MD).The mice were housed in the approved specific pathogen-free facility of our institution and used at 7 to 8 weeks of age.The care and use of laboratory animals was approved by The University of Texas M.D. Anderson Cancer Center Institutional Animal Care and Use Committee.
Immunofluorescence staining and flow cytometry.Cells were fixed, permeabilized, and blocked in flow cytometry buffers (Santa Cruz Biotechnology, Santa Cruz, CA) following the instructions of the manufacturer.Cells were double stained with a 1:100 dilution of rat antigalectin-3 monoclonal antibody (TIB166, 0.26 mg/mL; American Type Culture Collection, Manassas, VA) and a 1:500 dilution of rabbit antiphospho galectin-3 (Ser 6 , 0.88 mg/mL) for 1 hour at room temperature.Anti-phosphogalectin-3 antibody was established in conjunction with Zymed Laboratories (South San Francisco, CA) by immunizing rabbits against a synthetic phosphopeptide comprising of 14 NH 2 -terminal amino acids of galectin-3 phosphorylated at Ser 6 [DNF (pS) LHDALSGSGN (C)].The antibody was affinity purified on an immobilized ''hot'' peptide followed by affinity absorption on an immobilized ''cold'' peptide.The specificity of the antibody was determined by a differential ELISA using hot and cold peptides.Western blot analysis of human recombinant galectin-3 and a phosphorylated protein also confirmed specificity of the antibody for phosphorylated galectin-3.Secondary antibodies, FITC-conjugated goat anti-rat (Chemicon, Temecula, CA), and phycoerythrin-conjugated goat anti-rabbit (Molecular Probes, Eugene, OR) were added at 1:200 dilution and incubated for 30 minutes at room temperature.In each experiment, 10 6 cells were processed for flow cytometric analysis on a FACSCalibur apparatus (Becton Dickinson, Franklin Lakes, NJ).The dual labeling was analyzed by Cell Quest PRO software (Becton Dickinson).Single colorstained cells were used for instrument compensation.
Protein extraction and Western blot analysis.Total cell lysate extractions and Western blots were done as previously described (25).The blots were then probed with anti-galectin-3 or anti-cyclin D1 (Calbiochem, La Jolla, CA); anti-MUC2, anti-MUC5AC, or anti-MUC1 (Vector Laboratories, Burlingame, CA); anti-insulin-like growth factor binding protein 5 (IGFBP5), anti-PRSS3, anti-dual specificity phosphatase 6 (DUSP6), and anti-CTSL2 (Santa Cruz Biotechnology): or anti-h-actin (Sigma, St. Louis, MO) as a controls for loading in each lane, followed by the appropriate horseradish peroxidase-conjugated secondary antibodies (Chemicon).The antigen was detected using the enhanced chemiluminescence detection system (Amersham, Piscataway, NJ) according to the instructions of the manufacturer.
Tumorigenicity assay.Cells were harvested from cultures of BT549/V, BT549 gal 25A, BT549 gal 25B, BT549SE51, BT549SE56, BT549SA7, and BT549SA9 using 0.02% EDTA and resuspended in Ca 2+ -and Mg 2+ -free PBS.Nude mice were injected with 2 Â 10 6 cells mixed in 0.1 mL Matrigel (Becton Dickinson) into the mammary fat pad as described previously (28).Tumor diameters were measured weekly using calipers.Results were pooled from two separate experiments and all experiments were done with the approval of the Institutional Animal Core and Ethics Committee.Statistical analysis was done using StatView 4.02.All tests were two sided, with P = 0.05 regarded as significant.
Spontaneous lung micrometastasis assay.BT549 transfectant cells (1 Â 10 6 ) mixed with 2 mL of Matrigel were injected into the mammary fat pad of 6 to 8-week-old immunodeficient nude mice (NCR-NV, Taconic Farms, Germantown, NY).Mice were sacrificed by CO 2 inhalation 4 to 5 weeks after injection.At autopsy, the lungs were removed and examined for micrometastases.The lungs were washed with Hank's buffer, minced, and incubated with 5 mL DMEM containing 5 mg collagenase type IV and 36 units elastase on ice for 1 hour.Tissue was placed into a Stomachers bag with fresh medium for 45 seconds.This procedure was repeated twice with fresh medium.Medium were combined and spun at 500 Â g for 5 minutes.Cells were resuspended in 2 mL DMEM and plated at different dilutions in the presence of selection medium (800 Ag/mL G418).After 1 week, colonies were stained with methylene blue and counted.
Analysis of cell cycle distribution.Cells were plated at 2 Â 10 5 per well in six-well plates for 2 hours and then incubated in serum-free medium for 48 hours followed by 24-hour incubation in complete medium.Cells were trypsinized and aliquots of 1 Â 10 6 cells were fixed in 70% ethanol at À20jC for 24 hours.Fixed cells were then washed and suspended in 1 mL of fluorochrome solution (PI/RNASE staining buffer, BD Biosciences Clontech, Franklin Lakes, NJ) for 30 minutes in the dark at room temperature.Cell cycle analysis was done using FACSCalibur flow cytometer (BD Biosciences Clontech).Quantitation of cell cycle distribution was done using Mod Fit LT software (Verity Software House, Topsham, ME).
Microarray hybridization.Total RNA was extracted from 80% to 90% confluent BT549/V, BT549 gal 25A, and BT549SE51 clones using Trizol reagent (Invitrogen, Carlsbad, CA) and further purified on RNeasy spin columns (Qiagen, Valencia, CA) according to the manufacturer's directions.Thirty micrograms of total RNA was labeled with Cy3 (BT549/V) or Cy5 (BT549gal 25A and BT549SE51) using the 3DNA Array 350 Expression Array Detection kit (Genisphere, Inc., Hatfield, PA) following the instructions of the manufacturer.The labeled cDNA was purified using a QIAquick PCR Purification kit (Qiagen), dissolved in 40 AL ExpressHyb solution (BD Biosciences Clontech), and heated to 95jC for 10 minutes.The denatured probes were applied to a high-density 10-microarray slide (Unigene), which contains 10,000 randomly spotted oligonucleotides.(Unfortunately, galectin-3 is not included in the microchip gene library of the cancer genomic core laboratory at M.D. Anderson.)Hybridization was done at 60jC for 12 to 16 hours in a humid incubator.After hybridization, the slide was washed at 37jC, once each in 1Â SSC, 0.01% SDS, 0.2Â SSC, and 0.01% SDS, and then twice in 0.1Â SSC.Hybridized arrays were scanned at a 10 Am resolution on a microarray scanner (GeneTaq LSIV; Genomic Solutions, Ann Arbor, MI).
Microarray quantification.Scanned images were quantified in ArrayVision (Imaging Research, Inc., St. Catherine's, Ontario, Canada).Measurements were recorded for spot intensity, local background intensity, and signal-to-noise ratio.Spot intensity was computed as the integrated absorbance or volume in a fixed-size circle.Background intensity was computed as the median pixel value in four diamond-shaped regions at the corners of each spot.The signal-to-noise ratio was computed by dividing the background-corrected intensity by the SD of the background pixels.
Microarray data analysis.Quantified array data were imported into S-Plus software (Insightful Corp., Seattle, WA) for analysis.Backgroundcorrected intensities were globally rescaled to set the 75th percentile in each channel equal to 1,024.Rescaled intensities of <128 were replaced by the threshold value; this threshold was chosen to lie just below the smallest intensity of any spot with a signal-to-noise ratio >1.Next, intensities were transformed by computing the base 2 logarithm.Finally, the logtransformed spot intensities were normalized using robust local regression (29).A spot was identified as differentially expressed if the mean intensity in the two channels exceeded 512 and the estimated change exceeded 2.5-fold or if the mean intensity in the two channels exceeded 256 and the estimated change exceeded 4.5-fold.
Real-time quantitative PCR.Total RNA was isolated from BT549 gal25A, BT549V, BT549SE51, and primary breast tumor specimens as described above.A total of five galectin-3-positive and five grade-matched galectin-3-negative [as determined by real-time quantitative reverse transcription-PCR (RT-PCR)]

Cancer Research
Cancer Res 2005; 65: (23).December 1, 2005 invasive ductal carcinomas were obtained from surgical pathology specimens at M.D. Anderson Cancer Center under a protocol approved by the Committee on Human Experimentation and used in this study.Each sample pair was obtained from matched tumor and corresponding normal tissues from the same patient.The BD human reference total RNA is a mixture of total RNAs from multiple adult human tissues chosen to represent a broad range of expressed genes (BD Clontech, Palo Alto, CA).Initial experiments were done to determine the valid range of RNA concentrations and to show the similarity of PCR efficiencies for each gene of interest compared with the endogenous control gene cyclophilin.To determine fold changes in each gene, real-time RT-PCR was done on the ABI Prism 7900 using the commercially available gene expression assay for CCND1, CLECSF2, CTSL2, DUSP6, galectin-3, GFRA1, IGFBP5, and PRSS3 (Hs00277039, Hs00192860, Hs00822401, Hs00169257, Hs00173587, Hs00237133, Hs00181213, and Hs00605637, respectively), and the cyclophilin Vic-labeled Pre-Developed Assay Reagent (Applied Biosystems, Foster City, CA) without multiplexing.A 50 AL final reaction volume containing 1Â TaqMan Universal PCR Master Mix without AmpErase UNG (Applied Biosystems), 1Â Multiscribe with RNase inhibitors, and 1Â gene expression assay was used to amplify 50 ng total RNA with the following cycling conditions: 30 minutes at 48jC, 10 minutes at 95jC, then 50 cycles of 95jC for 15 seconds and 60jC for 1 minute.The 7900 Sequence Detection System 2.2 software automatically determined fold change for each gene in each sample relative to the Becton Dickinson Human Reference Pool using the yyCt method.Calculations were also done for each gene in tumors relative to their corresponding matched normal tissue.

Results
Galectin-3 expression and phosphorylation in stable BT549 transfectants.To assess the role of the phosphorylation of galectin-3 at Ser 6 in malignant transformation, we used BT549 stable transfectants previously established in our laboratory (7,27) as an experimental cellular model.Parental BT549 cells are nontumorigenic galectin-3-null cells.We transfected these cells with either wild-type galectin-3 (capable of undergoing phosphorylation) or Ser 6 mutants incapable of undergoing phosphorylation (Ser 6 !Glu or Ser 6 !Ala).After selection with G418, several clones were selected for further analysis: a control clone (BT549/V) expressing a vehicle plasmid, two wild-type galectin-3 clones (BT549gal 25A and gal 25B), two Ser 6 !Glu mutant galectin-3 clones (BT549SE51 and SE56), and two Ser 6 !Ala mutant galectin-3 clones (BT549SA7 and SA9).Of note is that mutation of human recombinant galectin-3 at Ser 6 did not alter ligand binding but did render it resistant to phosphorylation by casein kinase 1 in vitro (25).The expression of galectin-3 in these transfectants was determined by Western blot analysis (Fig. 1A) and flow cytometry (Fig. 1B).Almost all (>90%) of the cells in the transfectant clones expressed galectin-3 whereas the parental cells and the BT549/V clone control cells did not express detectable levels of galectin-3.The levels of galectin-3 expression varied up to 4-fold between the clones as measured by mean FITC fluorescence intensity (Fig. 1B).
A phospho-galectin-3 (Ser 6 )-specific antibody was used to monitor the phosphorylation of galectin-3 in vivo.Flow cytometric analysis of the BT549 transfectants revealed that 97% of cells transfected with wild-type galectin-3 (gal 25A and gal 25B) expressed high levels of phosphorylated galectin-3, whereas phosphorylated galectin-3 was completely absent in the phosphomutant transfected cells (Fig. 1B) The amount of phospho-galectin-3 in gal 25A clone was 1.3-fold higher than in gal 25B as determined by the mean phycoerythrin fluorescence intensity.These data agree with our previous findings using metabolic labeling with 32 P and Western blotting with a P-Ser antibody, which show that only the wild-type galectin-3 transfectants undergo phosphorylation in vivo (7,26,27).
Phosphorylation of galectin-3 is a prerequisite for malignant transformation.To investigate whether the phosphorylation status of galectin-3 correlates with tumorigenicity and malignant transformation, we explored the effect of phosphorylation status on tumor development in athymic nude mice.All mice (n = 10) injected with BT549gal 25A and BT549gal 25B cells (containing wild-type phosphorylated galectin-3) developed progressively growing tumors (Fig. 2A).In contrast, all mice (n = 25) injected with BT549 phosphomutant clones or BT549/V cells remained tumor-free for 40 days.In a separate experiment, the effect of galectin-3 phosphorylation status on spontaneous metastasis was examined.Autopsy did not reveal macroscopic metastases in the liver, lung, or lymph nodes of the tumor-bearing mice.However, three of four mice injected with BT549gal 25A developed lung micrometastases by 4 weeks after inoculation, with number of colonies ranging between one and six (Fig. 2B).When mice were sacrificed 5 weeks after inoculation, four of four and five of five mice injected with BT549gal 25Aand BT549gal 25B, respectively, showed lung micrometastases with higher numbers of colonies ranging from 8 to 38 for BT549gal 25A and 6 to 15 for BT549gal 25B (Fig. 2B).In contrast, none of the mice (0 of 20) injected with the phosphomutant galectin-3-expressing clones or with BT549/V control (0 of 5) developed lung micrometastasis (Fig. 2B).Tumorigenicity and metastasis of the galectin-3 transfectants correlated with the level of phospho-galectin-3 expression, but not with total galectin-3.The levels of total galectin-3 in gal 25A and gal 25B are similar as determined by Western blot and flow cytometry.The median FITC fluorescence intensity (total galectin-3) for gal 25A and gal 25B was 361.2 and 353.7, respectively.However, phospho-galectin-3 expression is 30% higher in gal 25A than in gal 25B (470.8 versus 370.4,median phycoerythrin fluorescence intensity; Fig. 1B).Tumors derived from BT549gal 25A had a significantly higher growth rate compared with BT549gal 25B tumors (P < 0.0005; Fig. 2A) and more aggressive metastatic behavior.The mean number of lung micrometastases colonies derived from gal 25A was 23 compared with 8.5 for gal 25B.
Effect of phospho-galectin-3 on cell cycle progression and proliferation.Cell-cycle progression analysis done by flow cytometry showed that stable expression of wild-type galectin-3 in BT549 cells resulted in a significant accumulation in G 2 -M phase of the cell cycle compared with vector control (Fig. 3A).The number of cells in G 2 -M phase was increased from 18.8% in BT549V to 27% and 24.1% in BT549gal 2A and gal 2B, respectively, and  decreased significantly in the phosphomutant clones (range 8.2-13.5%,P < 0.005).A concomitant decrease in the number of cells in G 1 from 56.1% of BT549V to 37.0% for BT549gal 25A and 39.3% for BT549gal 25B (P < 0.005 and P < 0.002, respectively) and a G 1 arrest for the phosphomutant clones were also observed (range 60.6-67.6%;Fig. 3A, table).Accordingly, enhanced growth rates were observed for BT549gal 25A and gal 25B relative to BT549V control (P < 0.003) as shown in Fig. 3B.Interestingly, the phosphomutant clones were significantly growth inhibited over the 8-day growth period compared with vector controls (P < 0.003; Fig. 3B).These results suggest a dominant negative activity for phosphomutant galectin-3.The transition from G 1 to G 2 -M phase and enhanced proliferation observed in BT549gal 25A and gal 25B was not correlated with the expression of total galectin-3 but rather with the levels of phospho-galectin-3.
Identification of genes differentially expressed in cells containing phosphorylated and unphosphorylated galectin-3.To identify genes and pathways that may be regulated (directly or indirectly) by phospho-galectin-3, we did global differential gene expression analysis between BT549/gal 25A (contain wild-type galectin-3) and BT549/V control cells, and between BT549SE51 (contain phosphomutant galectin-3) and BT549/V cells.We identified 210 genes that were differentially expressed in BT549gal 25A relative to BT549V control using a 2.5-fold change as the cutoff.Of these genes, 22 were common to both BT549gal 25A and BT549SE51 (Table 1).Among the 188 genes that were unique to BT549/gal 25A (not differentially expressed in cells containing the phosphomutant galectin-3), 73 were highly up-regulated and 115 were significantly down-regulated.(A partial list is shown in Table 2 and the complete list is given in Supplementary Table S2) Of the 73 up-regulated genes, 17 either encode hypothetical proteins or have currently unknown functions.Among the 115 genes that were down-regulated in BT549/gal 25A, 25 encode hypothetical proteins or have unknown functions.Glutathione S-transferase and retinoblastoma 1 were selected as strong feature genes supporting previously reported findings (17,30).Glutathione is an important cofactor for the enzymatic decomposition of intracellular peroxidases by the glutathione peroxidase enzyme (31).Glutathione S-transferase was up-regulated >17-fold in BT549gal 25A, suggesting that overexpression of phosphogalectin-3 could cause a shift in oxidative metabolism, resulting in an increased demand for glutathione synthesis and a more oxidizing intracellular environment.Another antioxidant protein that was also up-regulated is the reversion-induced LIM protein.
Some of the genes identified in our screen have previously been associated with transformation and proliferation in breast and other tissues, whereas others represent novel markers of phospho-galectin-3 function.Examples of known differentially regulated genes include cyclin D1, retinoblastoma 1, and extracellular matrix proteins.Cyclin D 1 regulates the progression of the early to mid-G 1 phase of the cell cycle and is strongly implicated in breast tumorigenesis (32), in part explaining some of the growth-promoting and oncogenic action of phospho-galectin-3.Reorganization of the extracellular matrix proteins is an essential step for migration and invasion.Members of the keratin family are associated with metastasis in non-small cell lung cancer (33) and collagen has been reported to inhibit cell motility and migration (34).We have shown up-regulation of keratins 18 and 19, as well as down-regulation of several collagen genes, IIIa1, Va2, and IVa1 and integrin a10 and integrin a6.Among the newly identified up-regulated genes involved in cell cycle regulation are those encoding cyclin B 2 , serum deprivation response protein, and epiregulin.Novel down-regulated genes include fibroblast growth factor 2 and midkine.
The most highly expressed up regulated genes, not previously reported to be associated with breast carcinoma, were C-type lectin member 2 (CLECSF2), an activation-induced Ca 2+ -dependent carbohydrate-binding integral membrane protein expressed in various cell types of hematopoietic origin (35); DUSP6, which blocks both the phosphorylation and enzymatic activity of the mitogen-activated protein kinase (MAPK) extracellular signalregulated kinase (ERK; ref. 36); and two trypsinogens (PRSS3 and PRSS1).Serine protease trypsin and its precursor trypsinogen have been linked to tumor progression by activation of the matrix metalloproteinases in pancreatic, gastric, or colorectal cancer (37).Among the most significantly down-regulated group of genes in BT549gal 25A were members of the cathepsin family (L2, F, H, L-like 1, and L-like 3).Cathepsins are papain family cysteine proteases that are a major component of the lysosomal proteolytic system.Recently, a novel noncaspase lysosomal apoptotic pathway in which the execution of apoptosis relies solely on the activation of cathepsins has been documented (38).Thus, our data suggest a potential protective role for phospho-galectin-3 in this newly discovered noncaspase apoptotic pathway.
Confirmation of microarray data by real-time quantitative reverse transcription-PCR and Western analysis.Real-time quantitative RT-PCR analysis of BT549gal 25A, BT549SE51, and BT549/V RNA was done to confirm the relative levels of expression in the transfectant clones for a subset of differentially expressed genes generated from the oligonucleotide microarray experiments.All (12 of 12) genes analyzed showed a pattern of transcript abundance consistent with the microarray data (Table 3).
The expression of selected proteins in BT549gal 25A, BT549SE51, and BT549/V was also compared by Western blotting.As shown in Fig. 4, the expression of IGBP5, PR553, DUSP6, cyclin D 1 , and MUC5AC was up-regulated in BT549gal 25A cells but not in BT549SE51 or BT549/V cells.In contrast, CTSL2 and MUC1 expression was down-regulated in BT549gal 25A (data is shown for one representative clone of each transfectant group).These results agree well with the microarray predictions and the RT-PCR data and further confirm the reliability of the microarray analysis.
To evaluate the clinical relevance of genes predicted to be associated with phospho-galectin-3, we compared their expression in matched pairs of galectin-3 positive and negative breast tumor groups (n = 10).As a representative example, the relative quantity of four transcripts in the two tumor groups is shown in Fig. 5. CLECSF2 and IGFBP5 were up-regulated in all galectin-3-positive tumors, but not in galectin-3-negative tumors with a median of 3.3 F 0.4 and 8.8 F 0.9 fold increase, respectively (P < 0.05).Cyclin D 1 was up-regulated and cathepsins L2 was down-regulated in both groups, which is in accordance with the pattern observed in BT549gal25A.

Discussion
We report here that phosphorylation of galectin-3 is associated with alterations in expression of unique sets of genes and is a prerequisite for the tumorigenicity of human breast epithelial cells.We confirmed that overexpression of wild-type galectin-3 in human breast epithelial cells promotes tumorigenicity in athymic mice and development of spontaneous metastasis in the lung.These data is in accordance with previously published reports and support the hypothesis that nuclear galectin-3 is associated with tumor invasion and metastasis of breast cancer cells (39)(40)(41).Parallel experiments in which a mutant form of galectin-3 incapable of phosphorylation was introduced into galectin-3 null cells strongly suggest that phosphorylation of galectin-3 is necessary for tumor formation.Microarray analysis comparing transfectants expressing wild-type galectin-3 and those expression phosphomutant galectin-3 yielded further evidence that wild-type phosphorylated galectin-3 is associated with expression of a variety of key genes in several functional categories.This is further supported by additional data using RT-PCR and Western analysis to confirm microarray findings.To identify potential mechanisms for the oncogenic property of phospho-galectin-3, we classified the differentially regulated genes into six functional categories: those involved in oxidative stress, apoptosis, cell cycle, signal transduction and transcriptional activation, cytoskeleton remodeling, and cell adhesion.The activation of diverse pathways by phospho-galectin-3 may reflect the functional diversity of galectin-3.
A possible role for phospho-galectin-3 in growth stimulation and tumorigenicity is supported by the up-regulation of cyclin D1 and members of the MAPK family, such as MAP3K10 and MAP2K1, suggesting that the c-Jun-NH 2 -kinase pathway is activated, which is in accordance with previous reports (26).Cyclin D1 is overexpressed in f50% of all breast cancers and its role in tumor progression is well established (42).The functional significance of cyclin D1 in breast cancer has been shown by gene-targeted silencing experiments.Cyclin D1 knockout mice are resistant to breast cancers induced by the neu and ras oncogenes, but remain fully sensitive to tumor driven by c-MYC or Wnt-1 (42).Phosphogalectin-3 may induce the expression of cyclin D1 either directly by binding/stabilization of nuclear protein-DNA complex formation at the Sp1 and cAMP-responsive element sites of the cyclin D1 promoter (20) or indirectly by activating the Wnt signaling pathway through binding upstream molecules like h-catenin (43,44).Other signal transduction genes that seem to be regulated by phospho-galectin-3 include protein kinase C (PKC) ~, IP3 receptor, tumor necrosis factor superfamily member 15, IGFBP5, and vascular endothelial growth factor h, which is down-regulated 3fold.Overexpression of the IP3 receptor may alter Ca 2+ release from internal pools by the second messenger IP3, resulting in activation of Ca 2+ -dependent signaling pathways, such as calmodulin and PKC (45).It is of interest that two PKC substrates, the pleckstrin 2 homologue and serum deprivation response proteins, were also upregulated by wild-type phosphorylated galectin-3.
Several genes in the cytoskeletal reorganization and cell adhesion pathways are also regulated in BT549gal 25A, including keratins 18 and 19.Keratins constitute the major intermediate filaments in several simple epithelial tissues and their presence has been used extensively in the diagnosis of tumors of epithelial and nonepithelial origin.For example, keratin 18 is commonly associated with both well and poorly differentiated carcinoma cells (46).In addition, g-filamin is an actin-binding protein that is up-regulated by phospho-galectin-3 and actin filaments are key molecules in the regulation of cell adhesion, spreading, and  configuration.Integrin a10 and integrin a6, which are downregulated, are regarded as invasion suppressor genes in many types of cancers (47).Suppressed expression of these genes in the transformed BT549gal 25A is in agreement with the metastatic potential of these cells.Among the genes most strongly downregulated by phospho-galectin-3 are several collagen genes: collagen III aI, collagen V a2, and collagen IV a1.
It is of interest that phospho-galectin-3 suppressed expression of the transmembrane mucin MUC1 and up-regulated the secreted gel forming mucin MUC5AC.Mucins are large extracellular proteins that are heavily glycosylated with broad functions that include protection and lubrication.Altered expression of mucin gene products or their glycosylation pattern accompanies the development of cancer and influences adhesion and invasion (48).MUC1 is normally expressed ubiquitously on the ducts and glands of simple secretary epithelial tissues.Overexpression of MUC1 by breast cancers has been reported and correlates with poor survival (49).MUC5AC is highly expressed in the goblet cells of the human airway and gastric epithelium and is a tumor-associated marker in the colon (48).The cytoplasmic moiety of MUC1 participates in intracellular signaling by interacting with the erb B receptors resulting in epidermal growth factor-dependent activation of RAS and ERK 1/2 MAPK pathway.The extracellular portion of MUC1, like MUC5AC, interacts with ICM-1 on endothelium, aiding in the metastatic spread of tumor cells (49).Our observation that MUC1 is replaced by MUC5AC in the model system we used requires further study to determine its functional significance.A functional link between mucins, galectin-3, and metastasis has been shown in our laboratory.Silencing of galectin-3 in highly metastatic human cancer cells resulted in reduced expression of MUC2 and inhibition of tumor growth and liver metastases in athymic mice (16).
IGFBP5 and CLECSF2 were found to be up-regulated in both BT549gal 25A, BT549gal 25B, and in galectin-3-positive breast tumors.The IGFBPs are a family of proteins that bind with high affinity to insulin growth factors and modulate their mitogenic actions by regulating their ability to interact with their signaling receptor (50).The role of IGFBP5 in human breast cancer cell growth is complex, with some reports that it may be growth inhibitory and associated with induction of apoptosis, whereas others show a survival role in response to apoptotic stimuli (51).CLECSF2 is an integral membrane protein that recognizes carbohydrate chains in a Ca 2+ -dependent manner and is involved in host defense mechanisms (35).It can trigger protein tyrosine phosphorylation pathways, thus providing an alternative route to growth hormones signaling C-type lectin, which is expressed preferentially in lymphoid tissues and in most hematopoietic cell types (52).Transcription of CLECSF2 was transiently upregulated during lymphocyte activation by phorbol ester.A functional role for CLECSF2 is breast cancer remains to be established.
The series of experiments described in this paper provide evidence that sustained phosphorylation of galectin-3 is associated with the tumorigenic potential of human breast epithelial cells.The data show the functional diversity of galectin-3 and suggest that phosphorylation of the protein is necessary for regulation of genes in a variety of functional categories that may play a role in malignant transformation.The costs of publication of this article were defrayed in part by the payment of page charges.This article must therefore be hereby marked advertisement in accordance with 18 U.S.C.Section 1734 solely to indicate this fact.

Figure 2 .
Figure 2. Phosphorylation of galectin-3 is required for tumor growth and formation of pulmonary micrometastases in athymic mice.A, athymic nude mice were injected with 2 Â 10 6 cells mixed in 0.1 mL of Matrigel into the mammary fat pad.Tumor diameters were measured weekly.Tumor grew in all mice injected with cells expressing wild-type galectin-3 but no tumors formed in the mice injected with cell expressing phosphomutant galectin-3 or vector alone.B, a colony of BT549gal 25A cells derived from the lungs of nude mice sacrificed 5 weeks after inoculation of BT549gal 25A cells in the fat pad.Table, spontaneous lung micrometastasis experiments showing the number of colonies formed in each mouse 4 to 5 weeks after injection with the different transfectants.

Figure 4 .
Figure 4. Protein expression profiles of a subset of differentially expressed genes from the microarray data in BT549 parental (P ), BT549/V control, BT549gal95A, and BT549SE51.

Table 1 .
Differentially expressed genes found in both BT549ga1 25 A and BT549SE51

Table 2 .
Partial list of differentially expressed genes in BT549gal 25A but not BT549SE51

Table 3 .
Validation of microarray data by real-time quantitative RT-PCR