Tumor Necrosis Factor-α Up-regulates the Expression of β1,4-Galactosyltransferase I in Primary Human Endothelial Cells by mRNA Stabilization*

During the course of an inflammatory response, the pro-inflammatory cytokine tumor necrosis factor-α (TNFα) triggers endothelial cells to increase the expression levels of adhesion molecules that are pivotal for the rolling, adhesion, and transmigration of leukocytes over the endothelial cell wall. Here we show that TNFα, in addition, has a regulatory function in the biosynthesis of proper carbohydrate molecules on endothelial cells that constitute ligands for adhesion molecules on leukocytes. Our data show that TNFα induced an increase in the expression of β1,4-galactosyltransferase-1 (β4GalT-1) in primary human umbilical vein endothelial cells in a time- and concentration-dependent manner. The β4GalT-1 mRNA up-regulation correlated with an increase in the Golgi expression and catalytic activity of the enzyme. Furthermore, an enhanced incorporation of galactose was observed in newly synthesized glycoproteins. Analysis of the molecular mechanism behind the up-regulation of β4GalT-1 showed that the increase in mRNA levels is due to an enhanced stability of the transcripts. These data strongly demonstrate that TNFα modulates the glycosylation of endothelial cells by a mechanism that directly enhances the stability of β4GalT-1 mRNA transcripts.

The cytokine TNF␣ 1 is released during the inflammatory response by activated macrophages (acute inflammation) or lymphocytes (chronic inflammation) situated at the site of injury. It activates endothelial cells to expose adhesion molecules that are necessary for the rolling, adhesion, and transmigration of leukocytes through the endothelial cell wall (1). Glycans play important roles in the recruitment of leukocytes because they are involved in many of the leukocyte-endothelial interactions. It has been reported previously that the glycan biosynthesis changes during an immune response (2). Accordingly, in recent preliminary data, we observed that TNF␣ induces changes in the expression levels of several glycosyltransferase genes 2 in primary HUVECs, of which up-regulation of the mRNA of ␤4GalT-1 was the most prominent one.
To date, ␤4GalT-1 (E.C. 2.4.1.22/38) is one of the best-studied glycosyltransferases. Although several aspects of its regulation and function have been extensively studied (3)(4)(5), there are still many uncharacterized features. ␤4GalT-1 is constitutively expressed in all tissues, with the exception of the brain (6), as a Golgi-resident protein (7). It catalyzes the transfer of galactose from the activated sugar donor, UDP-galactose, to oligosaccharides carrying terminal N-acetyl-glucosamine. This reaction results in the formation of ␤4-N-acetyllactosamine (Gal␤1,4-GlcNAc) or poly-␤4-N-acetyllactosamine sequences, also known as type 2 chains, that are abundantly present in N-and Olinked glycans as well as in glycolipids (8). Thus, ␤4GalT-1 plays an important role in the synthesis of the backbone structure of important carbohydrate epitopes involved in leukocyteendothelial cell interactions, such as selectin ligands, and also galectin and siglec ligands (9). Additionally, in the lactating mammary gland, ␣-lactalbumin converts ␤4GalT-1 to lactose synthase. Interaction of ␤4GalT-1 with this cofactor greatly lowers the K m of the enzyme for glucose, so that glucose can be effectively utilized as acceptor substrate at physiological concentrations (10,11).
Although ␤4GalT-1 is ubiquitously expressed in human tissues, the levels of expression vary and are subject to different stimuli, as in lactation (12,13), T-lymphocyte activation and differentiation (14,15), and activation of monocytic cells (16). In the case of the lactating mammary gland, the regulatory mechanism involves two steps: an initial ϳ10-fold increase in mRNA levels is achieved by transcriptional regulation (12,17), and an additional ϳ5-fold increase is obtained by translational control (13). During mid-and late pregnancy, the expression of lactating gland-specific transcription factors induces a change in the transcription initiation site, leading to the expression of a shorter transcript. These shorter mRNAs contain less stable secondary structures in the 5Ј-UTR and, as a result, are translated more efficiently than the longer and highly structured transcript of the non-lactating gland. However, the regulatory mechanisms in tissues other than the lactating mammary gland differ completely and are not yet fully elucidated. The regulatory mechanism in activated HL60 cells, in contrast with lymphocytes, seems to involve post-transcriptional control at the level of mRNA stability (14,16). A further indication that the expression of ␤4GalT-1 might be regulated at the posttranscriptional level is found in the short half-life of ϳ80 min of * 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.
The its mRNA in mice (18), similar to other labile mRNAs that are also regulated at the level of mRNA stability (19).
In the present study, we set out to investigate the mechanism involved in the TNF␣-mediated up-regulation of ␤4GalT-1 in primary HUVECs and to explore the effect on the activity and subcellular localization of the enzyme. Our data show that ␤4GalT-1 is up-regulated in a time-and concentrationdependent manner in response to TNF␣ stimulation due to an increase in the stability of the mRNA transcript. Furthermore, the ␤4GalT-1 mRNA up-regulation correlated with an increase in the Golgi-restricted expression of the enzyme and an augmented in vitro ␤4GalT-1 activity and resulted in an enhanced galactose incorporation in newly synthesized glycoproteins.

EXPERIMENTAL PROCEDURES
Cell Cultures and Reagents-HUVECs were isolated from five healthy donors by a modification of the method of Jaffe et al. (20). The cells were cultured in M199 supplemented with 100 units/ml penicillinstreptomycin, 10% human serum, 10% newborn calf serum, 5 units/ml heparine (Leo Pharmaceutical Products), and 10 ng/ml basic fibroblast growth factor (Sigma) on 1% gelatin (Fluka)-coated 6-well plates (Costar). When confluence was reached, cells were trypsinized (0.18% trypsin, 10 mM EDTA) and plated again to one-third of their density. All experiments were performed 2 passages after isolation. Cells were serum-starved for 2 h and exposed to 100 units/ml TNF␣ (Strahtmann Biotech) for 6 h or as otherwise indicated in the text.
The human cervical epithelial carcinoma cell line HeLa Tet-Off (Clontech), which is stably transfected to continuously express the tetracycline-controlled transactivator, was cultured in high-glucose Dulbecco's modified Eagle's medium (Invitrogen) containing 10% fetal calf serum (Clontech) supplemented with 2 mM L-glutamine, 4.5 g/liter D-glucose, 100 units/ml penicillin, and 100 units/ml streptomycin. Cells were serum-starved for 2 h prior to experiments.
Transfection-HeLa Tet-Off cells were grown to 90% confluence before transfection. The transfection with CLONfectin (Clontech) was performed as recommended by the manufacturer, using 2 g of reporter plasmid and 1 g of CLONfectin per 2.5 ϫ 10 5 cells. Cells were used for experiments 24 h after transfection.
mRNA Half-life Determination-HUVECs were stimulated as previously indicated, and 10 g/ml actinomycin D (Sigma) was added to block transcription. After stimulation of transfected HeLa Tet-Off, 1 g/ml doxycycline (Clontech) was added to block tetracycline-controlled transactivator-dependent transcription of chimeric mRNA. Lysates for mRNA isolation were taken 0, 30, 60, 120, 180, and 240 min after the addition of either actinomycin D or doxycycline.
Isolation of RNA and cDNA Synthesis-mRNA was specifically isolated by capturing of poly(A ϩ ) RNA in streptavidin-coated tubes with a mRNA Capture kit (Roche Applied Science), and cDNA was synthesized with the reverse transcription system kit (Promega) following the manufacturer's guidelines. Cells (2 ϫ10 5 cells/well) were washed twice with ice-cold PBS and harvested with 200 l of lysis buffer. Lysates were incubated with biotin-labeled oligo(dT) 20 for 5 min at 37°C, and then 50 l of the mix were transferred to streptavidin-coated tubes and incubated for 5 min at 37°C. After washing three times with 250 l of washing buffer, 30 l of the reverse transcription mix (5 mM MgCl 2 , 1ϫ reverse transcription buffer, 1 mM deoxynucleotide triphosphate, 0.4 unit of recombinant RNasin ribonuclease inhibitor, 0.4 unit of avian myeloblastosis virus reverse transcriptase, and 0.5 g of random hexamers in 30 l of nuclease-free water) were added to the tubes and incubated for 10 min at room temperature followed by a 45-min incubation at 42°C. To inactivate avian myeloblastosis virus reverse transcriptase and separate mRNA from the streptavidin-biotin complex, samples were heated at 99°C for 5 min, transferred to microcentrifuge tubes, and incubated in ice for 5 min. Then, they were diluted 1:2 in nuclease-free water and stored at Ϫ20°C until analysis.
Real-time PCR-Oligonucleotides (Table I) have been designed by using Primer Express 2.0 computer software (Applied Biosystems). All oligonucleotides were provided by Invitrogen. Oligonucleotide specificity was computer tested (BLAST; National Center for Biotechnology Information) by homology search with the human genome and specifically with all the known galactosyltransferases (CLUSTALW; European Molecular Biology Laboratory), and later confirmed by dissociation curve analysis and resolving the PCR products in agarose electrophoresis. The efficiency (22) of the oligonucleotides was determined using the computer program LinReg (23) and resulted in an average of 90%. PCRs were performed in an ABI 7900HT sequence detection system (Applied Biosystems) using SYBR Green Master Mix (Applied Biosystems) according to the manufacturer's guidelines. The fluorescence monitoring occurred at the end of each cycle. Additionally, dissociation curve analysis was performed at the end of every run, showing in all cases one single peak at the Tm (melting temperature of the amplicon) predicted by the Primer Express 2.0 software. The Ct (cycle at the threshold) value is defined as the number of PCR cycles where the fluorescence signal exceeds the detection threshold value, which is fixed above 10 times the S.D. of the fluorescence during the first 15 cycles and typically corresponds to 0.2 relative fluorescence unit. This threshold is set constant throughout the study and corresponds to the log linear range of the In the case of the primers used for quantitative real-time PCR, the location of the oligonucleotides within the coding sequence is provided in parentheses. The ATTTA sequences within the ␤4GalT-1 3Ј-UTR are depicted in italics, and the mutations generated are depicted in bold italics. amplification curve. The normalized amount of target (24) reflects the relative amount of target transcripts with respect to the expression of the endogenous reference gene. In this study, the endogenous reference gene chosen was glyceraldehyde 3-phosphate dehydrogenase, based on previous results (25). Immunofluorescence Microscopy-HUVECs were seeded in crosslinked gelatin-coated Labtek II chamber slides (Nalge Nunc Inc.) and cultured to confluence. Subsequently, cells were serum-starved for 2 h and incubated in the presence or absence of 100 IU/ml TNF␣. After 8 h, cells were washed with PBS and fixated with 3% paraformaldehyde/ PBS. Immediately, slides were permeabilized in PBS containing 0.075% saponin and 0.2% bovine serum albumin (permeabilizing buffer). Subsequently, the preparations were incubated for 45 min with a primary antibody solution containing GT2/36/118 (a kind gift of Dr. E. Berger, Zurich, Switzerland) (26) and phalloidin-rhodamine (a kind gift of Dr. H. de Vries, Amsterdam, The Netherlands) in the permeabilizing buffer, washed in the same buffer, and incubated for 45 min with goat anti-mouse Alexa 488 (Molecular Probes) and Hoechst (Molecular Probes). Finally, preparations were mounted in 15% polyvinyl alcohol/ 33% glycerol in PBS and analyzed under a Nikon Eclipse E800 microscope (Nikon Europe bv) and a Leica TCS SP2 confocal laser-scanning fluorescence microscope (Leica).
Flow Cytometry-Cells were washed twice in cold PBS and resuspended in PBS. Subsequently, cells were fixed with 3% paraformaldehyde/PBS (10 min, room temperature), centrifuged, and treated with 20 mM glycine/PBS (10 min, room temperature) in order to quench free aldehyde groups. Subsequently, cells were permeabilized with 0.1% saponin/PBS for 30 min, incubated for 30 min at room temperature with 25 l of the primary antibody (GT2/36/118) diluted in 0.1% saponin/1% bovine serum albumin-PBS, washed twice with 0.1% saponin/PBS, and incubated for 30 min at room temperature with secondary antibody (PE-rabbit anti-mouse) diluted in 0.1% saponin/1% bovine serum albumin-PBS. After the second incubation, cells were washed twice with 0.1% saponin/PBS and resuspended in a final volume of 100 l of 0.1% saponin/1% bovine serum albumin-PBS for analysis in the FACSCalibur (BD Biosciences). Cells were analyzed for immunofluorescence by collecting data for 10 4 cells/histogram. Corresponding negative controls were performed by using a secondary antibody alone.
Metabolic Labeling and Autoradiography-HUVECs were grown to confluence in T175 flasks, serum-starved for 2 h, and incubated for 10 h with 5 Ci/ml [ 3 H]galactose (American Radiolabeled Chemicals) in the presence or absence of 100 units/ml TNF␣ (Strahtmann Biotech). Cells were washed and resuspended in PBS and lysed in Laemmli sample buffer. The cell lysate was subjected to SDS-PAGE in a 10% polyacrylamide gel according to Laemmli (28) and blotted onto a polyvinylidene difluoride membrane (Millipore). The membrane was exposed for autoradiography in a tritium screen (Amersham Biosciences) and analyzed with a STORM TM 860 system (Amersham Biosciences).
Statistics-Results are shown as the average Ϯ S.D. of five independent measurements. Statistical significance was evaluated using the Kruskal-Wallis test in the SPSS 11.0 software (license number AZVU-7061649). Computer analysis of the ␤4GalT-1 gene and mRNA sequences was performed using TRANSFAC (29) and the CpGplot application at European Molecular Biology Laboratory-European Bioinformatics Institute (www.ebi.ac.uk/Tools/sequence.html).

RESULTS
TNF␣ Up-regulates ␤4GalT-1 mRNA Expression in HUVECs in a Time-and Concentration-dependent Manner-One of the functions of TNF␣ during the inflammatory response is to activate endothelial cells and to prompt them to produce adhesion molecules, which are necessary for the recruitment of leukocytes to the injured tissue. Preliminary observations showed that TNF␣ modulates the expression of several glycosylation-related genes. 2 Among the genes up-regulated by TNF␣ in endothelial cells were ␤4GalT-1, -5, and -6, whereas no effect was apparent in other ␤3or ␤4-galactosyltransferases. The expression of ␤4GalT-1 was considerably higher than that of other ␤4-galactosyltransferases. Expression of ␤3galactosyltransferases was more than 50-fold less abundant than ␤4GalT-1 expression. Hence, we decided to focus on ␤4GalT-1. To establish the kinetics of the TNF␣-mediated upregulation of ␤4GalT-1 mRNA levels, confluent monolayers of primary HUVECs were incubated in the presence of increasing concentrations of TNF␣ and for different times. Prior to this study, glyceraldehyde 3-phosphate dehydrogenase was determined as the optimal endogenous reference gene for normalization in the experimental conditions used in the present work (25). The expression of E-selectin was determined as a positive control for the activation of HUVECs. Low concentrations of TNF␣ (20 units/ml) were ineffective in up-regulating both Eselectin and ␤4GalT-1, whereas concentrations between 50 and  1A). For analysis of the time dependence of this effect, HUVECs were incubated with culture medium alone or with medium containing 100 units/ml TNF␣, and the relative abundance of E-selectin and ␤4GalT-1 transcripts was determined by quantitative real-time PCR. The activation of HUVECs by TNF␣ resulted in a Ͼ2-fold increase in the expression of ␤4GalT-1 at 6 h after the addition of the cytokine (Fig. 1B) and a Ͼ500-fold increase in the E-selectin mRNA. After this time point, mRNA levels slowly decreased, returning to the basal levels, although a small degree of up-regulation was still visible after 12 h of incubation.
␤4GalT-1 Is Exclusively Localized in the Golgi System and Enhanced in Response to TNF␣-Several functions have been assigned to ␤4GalT-1, some of which depend on the localization of the enzyme. Both the expression and localization of ␤4GalT-1 in control and TNF␣-treated HUVECs were examined by immunocytochemistry after staining with the monoclonal antibody GT2/36/118, raised against the stalk region of ␤4GalT-1 (26). Phalloidin-rhodamine staining was employed to define the limits of the cells and control for the activation of HUVECs, as seen by the formation of stress fibers (Fig. 2B), whereas Hoechst was used to stain the nuclei. The staining patterns correspond to a juxtanuclear localization, which is compatible with an exclusive Golgi localization (Fig. 2, A and  B). No extracellular membrane staining could be detected. The intracellular staining was more intense in TNF␣-treated HUVECs than in control cells (Fig. 2, A and B). Fluorescenceactivated cell-sorting analysis with the monoclonal antibody GT2/36/118 confirmed the ϳ2-fold increase in intracellular ␤4GalT-1 expression after TNF␣ treatment (Fig. 2C). These data indicate that the increase in ␤4GalT-1 mRNA results in an increase in the amount of the enzyme in the Golgi compartment.
The Up-regulation of ␤4GalT-1 mRNA Correlates with an Increase in ␤4GalT Enzyme Activity-In order to determine whether the increase in the intracellular expression of ␤4GalT-1 coincided with an increase in enzyme activity, cell lysates of unstimulated and TNF␣-stimulated HUVECs were assayed for specific ␤1,4-galactosyltransferase activity using N-acetyl-glucosamine and UDP-[ 3 H]galactose as substrates (27). TNF␣ induced a 2-fold increase in the in vitro ␤4-galactosyltransferase activity (from 0.18 Ϯ 0.02 to 0.36 Ϯ 0.03 nmol/g protein, p Ͻ 0.05), whereas the ␤3-galactosyltransferase activity remained unchanged (0.06 Ϯ 0.01 nmol/g protein). In addition, autoradiography after SDS-PAGE of lysates of cells grown in the presence of [ 3 H]galactose revealed that several bands in the molecular mass range between 45 and 80 kDa were more strongly radiolabeled after TNF␣ treatment (Fig. 3). These data indicate that TNF␣ stimulation of HU-VECs results in a 2-fold increase in the in vitro measured ␤4-galactosyltransferase activity and in the incorporation of galactose into newly synthesized glycoproteins. Our data strongly suggest that the observed increase in ␤4-galactosylation can be attributed to an increase in ␤4GalT-1 enzyme levels. The contribution of the up-regulation of ␤4GalT-5 and -6 2 to the increase in ␤4-galactosylation is expected to be low because it has been shown that the enzymatic activity of ␤4GalT-5 is several orders of magnitude lower than that of ␤4GalT-1 (30), whereas ␤4GalT-6 prefers glycolipids as acceptors (31).
TNF␣-mediated Increase of ␤4GalT-1 Expression Is Regulated by mRNA Stabilization-In order to identify potential regulatory sequences responsible for the up-regulation of ␤4GalT-1, the 2000 bp upstream of the transcription initiation site were analyzed using the computer program TRANSFAC and the European Molecular Biology Laboratory application CpG Plot (European Molecular Biology Laboratory-European Bioinformatics Institute). TRANSFAC was employed to identify potential binding sites for transcription factors in the 500 bp upstream of the transcription initiation site, whereas the CpG Plot application predicts the presence of CpG islands. The analysis revealed the existence of a relatively long CpG island close to the site of initiation of transcription (Ϫ679 to Ϫ57), which correlated with a high density of binding site with the transcription factor Sp1. Importantly, no binding sites for transcription factors activated by the TNF␣ signaling pathway were found within the 5Ј-flanking region of the ␤4GalT-1 gene.
In the lactating mammary gland, there is a change in the transcription initiation site that results in ␤4GalT-1 transcripts with a lower GC content in the 5Ј-UTR. The resulting

FIG. 3. TNF␣-induced increase in galactose incorporation in newly synthesized glycoproteins in HUVECs.
HUVECs were grown to confluence in T175 flasks and incubated with 5 mCi of (C 6 )-[ 3 H]galactose in the presence or absence of 100 units/ml TNF␣. After 24 h, cells were lysed in CHAPS, and identical amounts of protein were subjected to SDS-PAGE, followed by transfer onto polyvinylidene difluoride membrane and autoradiography using a tritium screen. The fold change values represent the ratio of TNF␣-treated to non-treated radioactivity for the indicated bands. Results are representative of three separate experiments.
FIG. 2. The localization of ␤4GalT-1 is restricted to the Golgi and up-regulated by TNF␣. HUVECs were grown to confluence in GlassTek II slides, serum-starved for 2 h, and incubated for 8 h in the absence (A) or presence (B) of 100 units/ml TNF␣. After fixation, the cells were permeabilized; stained with GT2/36/118, phalloidin, and Hoechst; and analyzed for immunofluorescence in a Nikon Eclipse E800 microscope. C, HUVECs were grown to confluence in T25 flasks, serum-starved for 2 h; incubated for 8 h in the presence or absence of 100 units/ml TNF␣; mechanically detached and subsequently fixated, permeabilized, and stained with GT2/36/118; and analyzed for immunofluorescence using a FACSCalibur. Results are representative of three separate experiments. less structured transcripts are translated more efficiency (13). However, this is unlikely to occur in endothelial cells, so we directed our attention to the 3Ј-UTR. Interestingly, the 3Ј-UTR of ␤4GalT-1 presents four AUUUA sequences (AREs), which are often associated with the regulation of mRNA stability (19). To test whether mRNA stabilization is involved in the upregulation of ␤4GalT-1, the half-life of ␤4GalT-1 transcripts was determined in resting and TNF␣-treated HUVECs using the transcription inhibitor actinomycin D. The results clearly showed that TNF␣ stimulation induced an increase in the half-life from 147 Ϯ 24 to 288 Ϯ 96 min (p Ͻ 0.01) after TNF␣ treatment (Fig. 4A). These data imply that the 2-fold increase in mRNA levels as shown in Fig. 1 is due to an enhanced mRNA stability. Next, the kinetics of this TNF␣-mediated mRNA stabilization were investigated by determining the half-life of ␤4GalT-1 at different time points after TNF␣ stimulation. The results clearly indicate that after 30 min, the half-life had already increased significantly (p Ͻ 0.01) (Fig. 4B).
The TNF␣-induced mRNA Stabilization of ␤4GalT-1 Involves a Single ARE-In order to further prove the involvement of the 3Ј-UTR of ␤4GalT-1 in the induction of mRNA stabilization, a chimeric ␤-globin mRNA containing the ␤4GalT-1 3Ј-UTR (␤4GalT1-3ЈUTR) was designed and transfected into HeLa Tet-Off cells. This chimeric mRNA is continuously transcribed under the control of a tetracycline-responsive element in HeLa Tet-Off cells, whereas its synthesis can be selectively terminated by the addition of doxycycline. Chimeric ␤Globin-␤4GalT1-3ЈUTR mRNA had a half-life of 135 Ϯ 28 min, similar to that of endogenous ␤4GalT-1 mRNA in resting HUVECs, indicating that the 3Ј-UTR of ␤4GalT-1 mRNA contains all regulatory domains determining its stability. Furthermore, TNF␣ treatment of HeLa Tet-Off cells transfected with the ␤Globin-␤4GalT1-3ЈUTR construct lead to an ϳ2ϫ increase in the half-life of the chimeric construct (Fig. 5C), as previously seen for HUVECs (Fig. 4, A and B).
As previously mentioned, the 3Ј-UTR of ␤4GalT-1 contains four AREs, termed AU1 through AU4 for simplicity, with AU1 being closest to the stop codon of ␤4GalT-1 (Fig. 5A). In order to study the individual contribution of each ARE to the TNF␣induced mRNA stabilization of ␤4GalT-1, every ARE was con-secutively mutated in the ␤Globin-␤4GalT1-3ЈUTR construct (Fig. 5A). The elimination of binding sites for AU-binding proteins by individually mutating the AUUUA sequences of AU1 through AU4 to ACGUA only affected the half-life of the chimeric mRNA when AU2 was mutated (Fig. 5B). Chimeric mRNA harboring the mutated AU2 (␤Globin-␤4GalT1-mAU2) showed an enhanced half-life (257 Ϯ 55 min, p Ͻ 0.01), similar to endogenous ␤4GalT-1 mRNA in TNF␣-stimulated HUVECs (Fig. 4A). Furthermore, TNF␣ did not induce an additional stabilization of chimeric ␤Globin-␤4GalT1-mAU2 mRNA (Fig.  5C). These results indicate that the stability of the ␤4GalT-1 mRNA is modulated solely through one ARE within its 3Ј-UTR (AU2). Finally, these results suggest that in resting cells a destabilizing factor is bound to AU2.

DISCUSSION
Although ␤4GalT-1 has long been considered a housekeeping gene, evidence is accumulating showing that its expression is regulated and involved in important functions (3). In this report, we provide new evidence showing that TNF␣ up-regulates the expression of ␤4GalT-1 in endothelial cells by a post-transcriptional mechanism involving the control of mRNA stability. Furthermore, the time-and concentration-dependent increase in mRNA levels is accompanied by an increase in the Golgi localization of the enzyme and the enzyme activity, suggesting a role for ␤4GalT-1 in the changes occurring in the glycosylation of endothelial cells during inflammation.
␤4GalT-1 mRNA expression is regulated by TNF␣ in a time-and concentration-dependent manner, indicative of a direct effect of TNF␣ signaling. The involvement of delayed regulatory systems, such as the synthesis of intermediate regulatory molecules or the effect of secondary cytokines secreted by HUVECs under the influence of TNF␣ (32), is excluded by the short time needed to reach a maximum effect as well as by the observation that the activation by TNF␣ appears to be saturable.
Computational analysis of the 5Ј-flanking region of the ␤4GalT-1 gene revealed the presence of a CpG island and a high frequency of binding sites for the transcription factor Sp1. CpG islands have been associated with a low degree of methylation and, consequently, with an "active" state of transcription (33), whereas Sp1 is a transcription factor generally associated with the transcription of so-called "housekeeping genes" (34). Because no binding sites for transcription factors related to TNF␣ signaling pathways were found in the 5Ј-flanking region of the ␤4GalT-1 gene, we hypothesized that the transcription of ␤4GalT-1 is set at a basal level, and consequently, it is likely that other mechanisms are involved in the regulation of gene expression of ␤4GalT-1 by TNF␣. Indeed, we could observe an increase in the half-life of ␤4GalT-1 mRNA after the treatment of HUVECs with TNF␣. This finding is in agreement with the presence of AUUUA sequences in the 3Ј-UTR of ␤4GalT-1. Furthermore, the rapid effect of TNF␣ in increasing the ␤4GalT-1 mRNA half-life (30 min) suggests the involvement of a rapid activation system. The analysis of the ␤4GalT-1 3Ј-UTR in the well-known Tet-Off system (21) clearly demonstrated that AU2 is the ARE responsible for the TNF␣-induced increase in mRNA stability. AUUUA sequences are also known as AREs and are potential binding sites for mRNA stabilitymodulating proteins (19). Several RNA-binding proteins, such as tristetraproline, K homology-type splicing regulatory protein, AUF1, and HuR, have been shown to form stable complexes with AREs and influence in this way the degradation rate, either positively or negatively, of mRNA (19). In particular, tristetraproline and HuR should be considered as a priori candidates for the modulation of the ␤4GalT-1 mRNA stability because their binding is regulated by the p38 (35) and phos- phatidylinositol 3-kinase (36) signaling pathways, which are also activated by TNF␣ (37). Furthermore, tristetraproline has long been known as a destabilizing factor (38). The identities of the ␤4GalT-1 ARE-binding proteins and how they are affected by TNF␣ signaling pathways are currently under investigation. It is still unclear why only AU2 is involved in the stabilization of the ␤4GalT-1 mRNA. It has previously been suggested that the function of AU-binding proteins might be dependent on the density of ARE in a particular sequence and the formation of mRNA secondary structures (39,40).
The TNF␣-induced stabilization of the ␤4GalT-1 transcripts in HUVECs results in an increase in the expression of the enzyme in the Golgi, the in vitro enzyme activity in cell lysates, and the incorporation of galactose into newly synthesized glycoproteins. Immunostainings of ␤4GalT-1 in HUVECs showed a juxtanuclear pattern compatible with localization in the Golgi apparatus and are in agreement with previous reports (41). Although ectopic localization of ␤4GalT-1 has been described previously (3,42,43) and associated with adhesive, lectin-like properties of the enzyme in the absence of the sugar donor UDP-galactose, we could not demonstrate any plasma membrane localization after sequential analysis of consecutive planes ranging from the basolateral to the apical membrane (data not shown). We conclude from these data that the upregulation of ␤4GalT-1 transcripts results in a functional increase in ␤4-galactosylation mediated by the increase in the Golgi expression of the enzyme. Indeed, TNF␣-treated HUVECs demonstrated an increase in the incorporation of galactose in proteins with a molecular mass between 45 and 100 kDa, which coincides with the molecular mass of numerous integrins and adhesion molecules (intercellular adhesion molecule-1, intercellular adhesion molecule-2, and vascular cell adhesion molecule-1) that are expected to act as scaffolds for the presentation of carbohydrate ligands for selectins, galectins, or siglecs. Interestingly, it has been shown that ␣6Ј-sialyllactosamine, the ligand for the B-cell molecule CD22, is up-regulated in TNF␣-treated HUVECs, which was has been attributed to an increase in the expression of ST6Gal I (44,45). However, the TNF␣-induced up-regulation of ␤4GalT-1 de-scribed in the present study can be expected to contribute to the change in glycan synthesis. This might be of special importance in the migration of B cells to the inflamed joints in rheumatoid arthritis, where B cells are found in high numbers and partly contribute to the progression of the disease (46). Accordingly, TNF␣ blockade therapy reduces the numbers of synovial CD22ϩ B cells in rheumatoid arthritis patients (47). This may involve blocking of the TNF␣-dependent up-regulation of ␤4GalT-1 and ST6Gal I. The TNF␣-dependent up-regulation of ␤4GalT-1 might also be critical in the expression of 6Јsulfo-␣3Јsialyl-␣3fucosyl-lactosamine, the L-selectin ligand, in activated endothelial cells (48). The up-regulation of L-selectin ligands is the main mechanism for the migration of monocytes and lymphocytes in chronic inflammation, as has been demonstrated in several models, such as contact hypersensitivity (49), Arthus reaction (50), Sjögren's syndrome (51), or ulcerative colitis (52). The further dissection of the mechanisms involved in regulating the expression of ␤4GalT-1 is important to increase our understanding of the biosynthesis of complex-type glycans that constitute ligands for adhesion molecules during inflammation, which will enable the design of more specific therapies to fight chronic inflammatory diseases.