Human a (1,3)-Fucosyltransferase IV ( FUTIV ) Gene Expression Is Regulated by Elk-1 in the U937 Cell Line*

The a 1,3-fucosyltransferase IV (FucTIV) encoded by its gene ( FUTIV ) is responsible for synthesis of Le x (Gal b 4[Fuc a 3]GlcNAc b 3Gal b 1,R), which causes compaction in the morula stage of the preimplantation mouse embryo, as well as a 1,3-fucosylation at multiple internal GlcNAc of unbranched poly- N -acetyllactosamine, termed “myeloglycan,” the physiological epitope of E-selectin. Since myeloglycan-type structure is also expressed in various types of human cancer and may mediate E-selectin-dependent metastasis, expression of FUTIV is oncodevelopmentally regulated. The mecha-nisms controlling FUTIV expression remain to be clari-fied. In this report, we further characterize FUTIV gene structure and define a non-TATA box-dependent transcriptional start region just upstream from the translational start. FUTIV promoter/reporter fusion constructs defined a “full-length” promoter and highly active fragments in the macrophage-derived U937 and myeloid HL60 cell lines. One highly active fragment contains a consensus binding site for the Ets-1 transcription factor (Withers, D. A., and Hakomori, S. (1997) Glycoconj. J. 14, 764). Gel shift analysis shows specific binding to this site in nuclear extracts from U937 cells. Mutation of the Ets consensus site significantly reduces FUTIV promoter activity in both cell lines. Gel supershift and dominant negative cotransfection experiments identified the Ets family member Elk-1 as one component binding and reg-ulating the FUTIV promoter in U937 cells. The signifi-cance of FUTIV regulation by Elk-1 is discussed.

The ␣1,3-fucosyltransferase IV (FucTIV) encoded by its gene (FUTIV) is responsible for synthesis of Le x (Gal␤4[Fuc␣3]GlcNAc␤3Gal␤1,R), which causes compaction in the morula stage of the preimplantation mouse embryo, as well as ␣1,3-fucosylation at multiple internal GlcNAc of unbranched poly-N-acetyllactosamine, termed "myeloglycan," the physiological epitope of Eselectin. Since myeloglycan-type structure is also expressed in various types of human cancer and may mediate E-selectin-dependent metastasis, expression of FUTIV is oncodevelopmentally regulated. The mechanisms controlling FUTIV expression remain to be clarified. In this report, we further characterize FUTIV gene structure and define a non-TATA box-dependent transcriptional start region just upstream from the translational start. FUTIV promoter/reporter fusion constructs defined a "full-length" promoter and highly active fragments in the macrophage-derived U937 and myeloid HL60 cell lines. One highly active fragment contains a consensus binding site for the Ets-1 transcription factor ( Among nine fucosyltransferases (FucTs) 1 expressed in animal cells (FucTI-FucTIX; see "Discussion"), FucTIV is of great biological interest based on results from two areas of study. One area is early embryogenesis, in which expression of Le x epitope becomes maximal at morula-stage preimplantation embryo, and declines quickly thereafter, by conversion to Le y expressed at the surface of blastocyst. Recent studies indicate that Le x expression in embryonal carcinoma F9 cells is due to expression of the FUTIV gene (1). Le x at morula stage is thought to cause compaction, the first cell adhesion event during embryogenesis, in which Le x -Le x interaction plays a major role (2,3) prior to uvomorulin (cadherin)-dependent adhesion (4). Cell adhesion based on Le x glycolipid was confirmed recently (5). FucTIV is also involved in synthesis of the cell surface epitope, so-called myeloglycan, consisting of unbranched long chain type 2 poly-N-acetyllactosamine having ␣133 fucosylation at multiple internal G1cNAc with ␣233 sialylation at terminal Gal (6). Myeloglycan is considered the major physiological E-selectin epitope controlling tethering and rolling of neutrophils on solid phase E-selectin, particularly under dynamic flow conditions (7). Expression of FucTVII is responsible for ␣133 fucosylation at penultimate G1cNAc, but penultimate fucosylation by FucTVII is inhibited by internal ␣133 fucosylation by FucTIV. Thus, FucTIV is responsible mainly for synthesis of myeloglycan without fucosylation at penultimate GlcNAc, i.e., without sialyl-Le x (8). Various types of tumor express myeloglycan-type structures (9), which may mediate E-selectin-dependent metastasis (for review, see Ref. 10).
Expression of the ␣1,3-FUT gene family has been examined for lung (11), colorectal (12,13), and gastric (14) tumors. In all cases, FUTIV expression is significantly higher in tumors than in adjacent non-tumorous tissue. In lung tumors, FUTIV and FUTVII expression levels were inversely correlated with patient prognosis (11). FUTIV expression has also been studied in purified myeloid lineage cells and cell lines induced to differentiate (15)(16)(17). Typically, differentiated cells lose surface Le x expression accompanied by a decrement in FUTIV transcript levels as compared with undifferentiated cells. Conversely, the colon adenocarcinoma cell line HT-29 shows elevated FUTIV mRNA levels when apoptosis is induced (18).
The collective data suggest that Le x , determined by "myeloid type" fucosyltransferase (FucTIV) expression, is an oncofetal marker. Expression is regulated during normal development (16,19,20) and in tumors (11)(12)(13). We sought to identify the trans-acting factors responsible for up-regulating FUTIV expression in tumor cell lines. Preliminary studies localized upstream genomic DNA fragments with high promoter activity in U937 and HL60 tumor cell lines (21). 2 The data presented here identifies the Ets family member Elk-1 as a positive regulatory factor acting on the FUTIV promoter in U937 cells.

MATERIALS AND METHODS
Isolation of FUTIV Genomic DNA, Plasmid Construction, and Sequence Analysis-A FUTIV genomic clone, 20, was isolated from a human placental genomic DNA library (Stratagene) while screening with a FUTIII probe. The EcoRI/NcoI 3.1-kb and BamHI/NcoI 1.65-kb * This work was supported in part by Grant OIG CA42505 from the NCI, National Institutes of Health, and by a grant from the Biomembrane Institute. 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 nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF305082 and AF305083.
The consensus Ets-1 site at Ϫ350 (from the translational start site) was mutated in promoter fusion constructs using the recombinant PCR technique (22). Primary PCR reactions used the Ϫ518 (XhoI/NcoI) construct in the pOGH vector (Nichols Institute, San Juan Capistrano, CA) as template. 3Ј-Primary PCR used a mutated Elk-1 site primer, 3ЈELKmut (Ϫ361 to Ϫ338; 5Ј-GGGCGCGGCCCTTCAGCCCTCGGG-3Ј, mutated bases underlined), and a downstream vector primer (5Ј-TG-CAGCTAGGTGAGCTGTC-3Ј). 5Ј-Primary PCR used 5ЈELKmut primer (complementary to 3ЈELKmut) and an upstream vector primer (5Ј-CCCAGTCACGACGTTGTAAAACG-3Ј). Secondary PCR used primary PCR fragments as templates and the flanking vector primers. The resulting fragment was digested with BstXI and SfiI and cloned into pG-1647Bam deleted for BstXI/SfiI. Resulting clones were sequenced through the BstXI/SfiI interval. DNA was sequenced either manually using Sequenase version 2.0 (Amersham Pharmacia Biotech) or with an ABI310 automated sequencer. Putative promoter sequences were analyzed for consensus transcription factor binding sites using Transcription Element Search Software (TESS) (23).
Cell Culture and Transient Transfections-U937 and HL60 cell lines were originally purchased from ATCC and maintained in this laboratory. Cells were cultured in RPMI containing 10% heat-inactivated fetal bovine serum at cell density Յ8 ϫ 10 5 cells/ml. Cultures were split 1 day prior to electroporation and collected at a culture density of ϳ5 ϫ 10 5 cells/ml. For U937 electroporation, 5 ϫ 10 6 cells were resuspended in 250 l of room temperature RPMI containing 10% fetal bovine serum and 10 g of plasmid DNA and transferred to a 0.4-cm gap cuvette. HL60 cells (2 ϫ 10 6 ) were resuspended in 400 l of medium containing 20 g of DNA. Each electroporation included 1 g of pXGH5 (growth hormone expression vector) to control for transfection efficiency. Electroporation at 242 V and 975 microfarads (U937) or 260 V and 1,050 microfarads (HL60) was done with a Gene Pulser II electroporator (Bio-Rad). After electroporation, cells were cultured in a 5-ml final volume of RPMI with 10% fetal bovine serum for 48 h. For cotransfection experiments, the Elk-1 dominant negative-expressing plasmid in the pCMV5 vector was used. This plasmid encodes the DNA binding, but lacks the transactivation domain of Elk-1. pCMV5 or pGL3 vectors were used as controls. Cotransfection experiments used 8 g of reporter plasmid, 2 g of control or experimental plasmid (pGL3basic, pCMV5, or pCMV5-Elk-Dominant negative), and 1 g of pXGH5.
Luciferase, Growth Hormone, and Protein Quantitation from Transient Transfections-Electroporated cell extracts were prepared and assayed using the luciferase assay system (Promega) according to the manufacturer's instructions. Firefly luciferase levels were determined on an EG&G Berthold Lumat LB 9507 luminometer with a 100-l D-luciferin injection. Growth hormone levels were assayed from the culture medium using the HGH-TGES kit purchased from Nichols Institute (San Juan Capistrano, CA). Extract protein levels were determined by the bicinchonic acid method (Pierce). Luciferase activity was normalized by both growth hormone (transfection efficiency control) and protein levels (cell number control).
Nuclear Extract Preparation and Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts were prepared by the method of Dent and Latchman (26). EMSA and "supershift" experiments were performed essentially as described (26), except that Tris-glycine (25 mM Tris, pH 8.5, 190 mM glycine, 1 mM EDTA) was used as gel running buffer. Antibodies for supershift were purchased from Santa Cruz Biotechnology, Inc.
RNA isolation used the Ultraspec (Biotecx Laboratories, Inc.) acidphenol procedure, and poly(A) ϩ RNA was selected using the poly(A)tract mRNA isolation system IV kit from Promega Corp. Northern blots were prepared as described (27).

RESULTS
FUTIV Gene Structure-The FUTIV gene produces three transcripts of 2.3, 3.0, and 6.0 kb as detected on Northern blots (15,28,29). The 2.3-and 6.0-kb transcripts are typically detected at much higher abundance than the 3.0-kb species (Fig.  2B). cDNAs corresponding to the 2.3-and 3.0-kb transcripts were described previously (28) (see Fig. 1). The derivation of the 6.0-kb transcript, however, is not known and has been suggested to come from an as yet undescribed FUT gene. This possibility led us to further investigate FUTIV gene structure. DNA sequence of a FUTIV 3Ј-genomic DNA fragment and the FUTIV 3Ј-untranslated region (Fig. 1) were used to synthesize downstream and upstream primers, respectively, for RT-PCR amplification of the putative 6.0-kb transcript. This experiment, shown in Fig. 2A, clearly detects the presence of another FUTIV transcript containing sequences common to the 3.0-and 2.3-kb FUTIV transcripts (since the upstream primer falls within the 3Ј-untranslated region of the 2.3-and 3.0-kb transcripts) as well as additional 3Ј-untranslated sequences. In addition, this RT-PCR-generated fragment detects only a ϳ6.0-kb transcript on Northern blots (Fig. 2B).
Using 3Ј-RACE analysis (24), the remainder of this cDNA  NotI (N), and EcoRV (V); large central arrow represents protein coding region. Large arrows above genomic map represent previously described cDNAs (28). The large arrow below the genomic map represents a cDNA described within. Primers used for RT-PCR and RACE experiments are shown as small arrows. Derivation of probes used to detect FUTIV transcripts by Northern analysis (Fig. 2B) is shown at the bottom. Sequences of the 3Ј-untranslated region of the ϳ6.0-kb transcript and the promoter region have been deposited with GenBank under accession numbers AF305083 and AF305082, respectively. was cloned. Sequence analysis showed a 3,098-base pair extension from the 3Ј-most sequences of the 3.0-kb transcript (ELFT-L) to the poly(A) addition site. This cDNA is colinear with genomic DNA (data not shown). The 6.0-kb transcript, which matches previously unassigned clones in the human expressed sequence tag data base, uses an alternate poly(A) addition site from the shorter FUTIV transcripts and appears to differ from them only by the length of 3Ј-untranslated sequences.
RACE was also used to locate the 5Ј end of FUTIV transcripts in RNA from U937 cells. Results from this experiment (Fig. 3) show that ϳ80% of recovered RACE clones extended to a region approximately 85 base pairs upstream from the start of translation. Two further upstream start sites at Ϫ129 and Ϫ145 were also detected by this method. An additional low frequency FUTIV start site, corresponding to the 3.0-kb transcript, is located approximately 550 base pairs upstream from the start of translation (28). As is typically true for genes with scattered transcriptional start sites, the FUTIV promoter contains no canonical TATA box.
Promoter Analysis-A genomic clone containing FUTIV was isolated from a placental DNA library (CLONTECH) using FUTIII gene sequences as probe. 3.1 kb extending upstream from the start of translation was subcloned and sequenced (Fig.  3). This fragment matches the restriction map of the FUTIV upstream region (29 -31) and contains 5Ј sequences from the FUTIV 2.3-and 3.0-kb cDNAs (28) (Fig. 1), thus confirming its identity as FUTIV. Comparison of the FUTIV upstream sequence to a transcription factor consensus binding site data base revealed many matches to described factor binding sites (Fig. 3).
The 3.1-kb fragment was cloned into the luciferase reporter vector (pGL3basic) to make "pG-3092Eco." This "full-length" fragment and deleted or mutated derivatives were transfected into FUTIV-expressing cell lines U937 and HL60, and luciferase reporter levels were assayed. The longest constructs (pG-3092Eco, pG-2444Nsi, and pG-2145Pac) have undetectable promoter activity in both cell lines (data not shown). Deletion derivatives, such as pG-1800HindIII (data not shown) and pG-1647Bam have high activity in U937 (Fig. 4) and HL60 (data not shown) cell lines. In addition, the internal deletion removing sequences from Ϫ424 to Ϫ273 from the full-length promoter has significantly reduced activity in U937 cells. Mutation of a consensus Ets-1 binding site at position Ϫ350 reduces promoter activity further (Fig. 4). This region was further studied using EMSA in order to localize potential transcription factor binding sites.
Electrophoretic Mobility Shift Assays Using the Ets Consensus Site at Ϫ350 -The 151-base pair interval from Ϫ424 to Ϫ273 contains a consensus binding site for the Ets family of transcription factors located at Ϫ350. Oligonucleotides corresponding to FUTIV promoter region Ϫ350 and containing wild type or mutant Ets binding sites were used in EMSA to detect binding activity in U937 cells (Fig. 5). The Ets mutant oligonucleotide differs from wild type by a three-base substitution at the Ets consensus core sequence (see "Materials and Methods"). At least two binding activities (see s1 and s2 in Fig. 5) were detected with the wild type oligonucleotide. These activities were specific for the Ets consensus site since binding was competed with excess unlabeled wild type, but not mutant, oligonucleotide (Fig. 5, left panel, lanes 2 and 3). In addition, s1 and s2 bands were absent when the mutant oligonucleotide was used as probe (Fig. 5, left panel, lane 4).
The identity of the factor(s) bound to the FucTIV Ets consensus site was investigated using EMSA and antibodies to various Ets family members (Fig. 5, middle panel). The most abundant specific binding activity (s2) is shifted in mobility upon addition of anti-Elk-1 antibody to the binding reaction (middle panel, lane 3) . Ets-1, pu.1 (middle panel, lanes 2 and 4,  respectively), Ets-2, and SAP-1 antibodies failed to recognize U937 nuclear proteins bound to this site (data not shown). In addition, the s2 band is absent from FUTIV-nonexpressing HepG2 cells (Fig. 5, right panel).
Role of the Elk-1 (Ϫ350) Binding Site-A promoter fusion construct (pG-1647Bam/Ϫ350(Elk Ϫ )) was made containing the same Ets-1 site mutation used above for EMSA. When compared with wild type (pG-1647Bam), promoter activity of the Elk-1 site mutant is significantly reduced in U937 cells (see Fig. 4).
The role of Elk-1 was further investigated by co-transfection experiments using the FucTIV promoter fusion construct pG-1647Bam, containing either the wild type or mutated Elk-1 Ϫ350 site, along with an expression vector encoding dominant

Elk-1 Regulates Human FUTIV
negative Elk-1 protein (32). Results of these experiments are shown in Fig. 6. The data show a dramatic reduction (approximately 5-fold) in promoter activity of the pG-1647Bam construct when co-transfected with the dominant negative Elk-1encoding, but not the empty vector plasmid. When the mutant Elk site construct was used as reporter, DN-Elk cotransfection reduced promoter activity to the same level seen for the wild type reporter plus DN-Elk cotransfection. The latter data suggest that Elk-1 binds to and regulates other sites in the FUTIV promoter, possibly a weaker consensus site at Ϫ610 base pairs (see Fig. 3). Collectively, these data show that Elk-1 binds specifically to the Ϫ350 site and regulates expression from the FUTIV promoter.
We have localized the predominant promoter for the FUTIV gene to just upstream from the translational start site. The FUTIV promoter lacks a TATA box and fits the criteria of a CpG island (50). Typical of such promoters, multiple, scattered transcriptional start sites were identified. In addition, a key regulator of the FUTIV promoter, Elk-1, was identified by DNA binding and cotransfection experiments.
Elk-1 was originally found by homology with the oncogene for Ets-1 (51) and is a member of the large family of Ets-related transcription factors. The Ets family regulates expression of other glycosyltransferases such as ␤4-galactosyltransferase-I by TASS-I (52), ␤-1,2-N-acetylglucosaminyltransferase II and N-acetylglucosaminyltransferase-V by Ets-1 and Ets-2 (53-56), as well as members of the metalloproteinase gene family (57)(58)(59)(60)(61). Activity of N-acetylglucosaminyltransferase-V and metalloproteinases contributes to the metastatic capacity of tumors.
Ets family proteins bind to DNA either autonomously or as a member (called the ternary complex factor) of a complex with the dimeric serum response factor. Three Ets family members are capable of acting as ternary complex factors (Elk-1, SAP-1, and SAP-2/Erp/Net; reviewed in Ref. 62). Ets family proteins bind DNA sites containing a core GGA consensus sequence via a winged helix-turn-helix DNA binding motif (reviewed in Ref. 63). Binding site specificity for members of the Ets family is partially determined by DNA sequence flanking the core site. The FUTIV Elk-1 binding site at Ϫ350 (GCCCGGAAGCC) matches the 7 central bases of the 11-base consensus sequence (AACCGGAAGTG/a) to which Elk-1 binds autonomously (64). The latter sequence was also selected from randomized oligonucleotides by the Elk-1 DNA binding domain (65). SAP-1 is theoretically capable of autonomous binding to the FUTIV Ϫ350 site (65) but does not occur in U937 extracts (data not shown), so the identity of the s1 binding activity seen in Fig. 4 is unknown. Since a serum response element is absent in the surrounding FUTIV promoter region, Elk-1 may bind to this site autonomously rather than as a component of serum response factor. If so, this is the only functional gene promoter site yet described capable of autonomous regulation by Elk-1.
Elk-1 is a target of central importance in both Ras-dependent and -independent mitogen-activated protein kinase signaling pathways in numerous cell types (32, 66 -69). Both growth factor-regulated kinases and two groups of stress-activated kinases (c-Jun N-terminal kinases and p38 kinase) have been Extracts were prepared from each sample and luciferase, growth hormone, and protein levels assayed. Luciferase activity was normalized by both growth hormone (transfection efficiency control) and protein levels (cell number control). Normalized activities are expressed relative to the pGbasic sample, which was arbitrarily assigned the value 1.0. Plotted values represent the average of at least three independent experiments for each construct.

Elk-1 Regulates Human FUTIV
found to activate Elk-1 (32, 70 -72). Phosphorylation of Elk-1 induces a conformational change that increases its DNA binding, transcriptional activating and serum response factor binding activities, thus activating both the ternary complex and autonomous transcriptional regulatory functions of this protein (32,66,73,74). The multiplicity of signaling pathways that activate Elk-1, and their deranged regulation in tumors, may explain the common expression of an autonomous Elk-1 target such as FUTIV in many tumors and tumor cell lines of various derivation (38,43). In contrast, adults express the myeloid-type fucosyltransferase activity attributed to FucTIV only in adult human leukocytes and brain tissue (20).
Whether Elk-1 plays a role in FUTIV expression during normal development is not known. It is of interest, however, that a fucosyltransferase activity with acceptor and inhibitor specificity characteristic of FucTIV is expressed in the early stages of development of many human tissues. The embryonic fucosyltransferase activity gives way to a distinct tissue characteristic activity in the adult organ (20). This expression pattern is mirrored in the developing myeloid lineage, where immature promyelocytes express predominantly FUTIV while mature granulocytes express mostly FUTVII but very low levels of FUTIV (15). The developing human lung also shows stage specific patterns of Le x , Le y , and SLe x expression. Le x is abundantly expressed in the early developmental stages, in cells of the presumptive bronchiolus, but declines in abundance after overt development of this tissue (19). Finally, down-regulation of FUTIV mRNA levels is seen upon Me 2 SO-induced differentiation of the promyelocytic HL60 cell line (16). Recently, a sixth ␣1,3-fucosyltransferase (FucTIX) (39) has been described, which apparently shares acceptor specificity and NEM insensitivity (see analysis in Ref. 40) with FucTIV. Although some of the above studies specifically assay FUTIV mRNA expression (by RT-PCR), caution is warranted in attributing enzyme activity or end product (Le x ) expression to FucTIV. Given all this information, it is likely that Elk-1 plays a role in establishing or maintaining expression of FUTIV in embryonic tissues or pluripotent lineages, but our data do not support a role for Elk-1 in FUTIV down-regulation during differentiation (data not shown).
Cell surface, stage-specific embryonic expression of Le x , determined by FucTIV activity, may mediate cell adhesive interactions during tissue morphogenesis. Evidence in support of this comes from studies of mouse preimplantation embryos, the blastomeres of which normally have a tightly compact form, but undergo decompaction when incubated with multivalent Le x in solution (75,76). Blastomere compaction is known from gene knock-out experiments to be mediated by E-cadherin (77). Using mutant embryocarcinoma (EC) cell lines, other molecules involved in this process have been identified. Cells from two independent mutant lines, deficient either for embryoglycan (the major carrier of Le x in EC cells (Ref. 78)) or for ␣(1,3)-fucosyltransferase activity, could not aggregate with wild type cells (48,79), despite a normal capacity for homotypic aggregation. These studies suggest that cell surface Le x plays a role in the initial stages of E-cadherin-mediated cell adhesion.
Enhanced FUTIV expression may also play a role in tumor progression, since the myeloglycan-type structure produced by FIG. 5. EMSA with probes from FU-TIV ؊350 region encompassing the Ets/Elk-1 consensus binding site using extracts from FUTIV-expressing U937 or nonexpressing HepG2 cell lines. Left panel, wild type (W) or mutant (M) oligonucleotide probes were incubated without or with 100-fold molar excess of the indicated unlabeled oligonucleotide. Middle and right panels, EMSA after incubation in the absence (Ϫ) or presence of anti-Ets-1 (E), -Elk-1 (L), or -Pu.1 (P) antisera using extracts from U937 (middle panel) or HepG2 (right panel). s1 and s2 indicate specifically bound complex; ss indicates a "supershifted" complex.
FIG. 6. U937 cotransfection with dominant negative Elk-1-expressing plasmid versus control cotransfections with pGL3 or pCMV5 vector plasmids. Reporter constructs were either wild type (pG-1647(Bam)) or the Elk site-mutated (pG-1647(Bam Ϫ350Elk Ϫ )) constructs. Extracts were prepared from each sample and luciferase, growth hormone, and protein levels assayed. Luciferase activity was normalized by both growth hormone (transfection efficiency control) and protein (cell number control) levels. Normalized activities are expressed relative to the pG-1647-Elk ϩ (cotransfected with pGbasic) sample, which was arbitrarily assigned the value of 100. Plotted values represent the average of three independent experiments.
FUTIV accumulates in metastatic colonic adenocarcinoma as well as lung and gastric tumor cell lines, but not in normal colonic mucosa or normal fibroblasts (9). Tumor cells expressing cell-surface myeloglycan structures might be better able to metastasize to tissues containing counter-receptors such as E-selectin or Le x .