LEC12 and LEC29 Gain-of-Function Chinese Hamster Ovary Mutants Reveal Mechanisms for Regulating VIM-2 Antigen Synthesis and E-selectin Binding*

LEC12 and LEC29 are two gain-of-function Chinese hamster ovary glycosylation mutants that express the Fut9 gene encoding α(1,3)fucosyltransferase IX (α(1,3) Fuc-TIX). Both mutants express the Lewis X (LeX) determinant Galβ(1,4)[Fucα(1,3)]GlcNAc, and LEC12, but not LEC29 cells, also express the VIM-2 antigen SAα(2,3)-Galβ(1,4)GlcNAcβ(1,3)Galβ(1,4)[Fucα(1,3)]GlcNAc. Here we show that LEC29 cells transfected with a Fut9 cDNA express VIM-2, and thus LEC29 cells synthesize appropriate acceptors to generate the VIM-2 epitope. Semiquantitative reverse transcription-PCR showed that LEC12 has 10- to 20-fold less Fut9 gene transcripts than LEC29. However, Western analysis revealed that LEC12 has ∼20 times more Fut9 protein than LEC29. The latter finding was consistent with our previous observation that LEC12 has ∼40 times more in vitro α(1,3)Fuc-T activity than LEC29. The basis for the difference in Fut9 protein levels was found to lie in sequence differences in the 5′-untranslated regions (5′-UTR) of LEC12 and LEC29 Fut9 gene transcripts. Whereas reporter assays with the respective 5′-UTR regions linked to luciferase did not indicate a reduced translation efficiency caused by the LEC29 5′-UTR, transfected full-length LEC29 Fut9 cDNA or in vitro-synthesized full-length LEC29 Fut9 RNA gave less Fut9 protein than similar constructs with a LEC12 5′-UTR. This difference appears to be largely responsible for the reduced α(1,3)Fuc-TIX activity and lack of VIM-2 expression of LEC29 cells. This could be of physiological relevance, because LEC29 and parent Chinese hamster ovary cells transiently expressing a Fut9 cDNA were able to bind mouse E-selectin, although they did not express sialyl-LeX.


Antibodies, Chemicals, and Molecular Biology Reagents-Mouse
IgM anti-SSEA-1 (anti-Le X ) monoclonal antibody (mAb) (12) was prepared previously in the laboratory (13). Purified, FITC-conjugated, mouse anti-Le X mAb (clone AHN1.1) was obtained from Calbiochem. Mouse IgM anti-VIM-2 mAb (14) was purchased from Bioresearch GmbH. Anti-sialyl-Le X mouse IgM mAb CSLEX-1 (15) was purified from HB-8580 hybridoma medium obtained from ATCC. Human IgG1, rabbit, or goat fluorescein isothiocyanate (FITC)-conjugated anti-human IgG, anti-goat IgG, anti-mouse IgM, anti-mouse IgGϩAϩM, and horseradish peroxidase-conjugated goat anti-human IgG antibodies were purchased from Zymed Laboratories. Goat allophycocyanin-conjugated anti-mouse IgM (HϩL) was from Caltag Laboratories. Purified recombinant mouse E-selectin fused to human IgG1 Fc fragment and produced in J558 myeloma cells was a kind gift of Drs. Martin Wild and Dietmar Vestweber, University of Munster, Germany (16). Mouse mAbs KM8621 and KM2681 (IgG1) raised against human Fut9 protein produced in Escherichia coli and purified from mouse ascites (KM8621) or hybridoma supernatant (KM2681) were a kind gift of Dr. Hisashi Narimatsu, National Institute of Advanced Industrial Science and Technology, Tsukuba, Japan. Antibody C-17, an affinity-purified goat polyclonal antibody against a carboxyl-terminal peptide of human Fut9, was purchased from Santa Cruz Biotechnology. Mouse anti-Myc monoclonal antibody 9E10 and an anti-mouse ␤-tubulin antibody were obtained from Dr. E. Richard Stanley, Albert Einstein College of Medicine. Amino acid mix was purchased from Promega. Coomassie Brilliant Blue, Ponceau S, creatine phosphate, and creatine phosphokinase purified from rabbit muscle were obtained from Sigma. Hanks' balanced salts mix was from Mediatech. Lysophosphocholine was supplied by Avanti Polar Lipids. Peptide N-glycanase F (PNGase F) purified from Flavobacterium meningosepticum was obtained from New England Biolabs. S-Adenosylmethionine was from ICN. Spermidine was purchased from Fluka. Zeomycin was from Invitrogen.
Semi-quantitation of Fut9 Transcripts by RT-PCR-RNA was prepared using 1 ml of TRIzol TM reagent (Invitrogen) per 10 7 cells. 10 g of total RNA from LEC12 or LEC29 was reverse-transcribed at 48°C using 1 g of oligo(dT) and 100 ng of random hexamers with a mixture of Superscript II TM and avian myeloblastosis virus reverse transcriptases in 40 l. cDNA products were treated with RNase H (Invitrogen), and 1 l of cDNA was subjected to PCR for 27, 30, 33, 36, 39, 42, or 45 cycles. Products were electrophoresed on a 1% agarose gel and stained with ethidium bromide. NIH Image software was used to quantitate by densitometry. Fut9 gene-specific primers were: PS394 (5Ј-TGA CAA CAC AGT GAA GTG GTT CT) and PS229 (5Ј-CCA CAT GAA TGA ATG AAT CAG CTG G). Glyceraldehyde-3-phosphate dehydrogenase gene-specific primers were PS270 (5Ј-TGA ATT CAT TGA CCT CAA CTA CAT) and PS271 (5Ј-AGA ATT CTT ACT CCT TGG AGG CCA).
Isolation of Microsomes from Cells-Cells were grown in suspension to a density of 4 -5 ϫ 10 5 /ml and pelleted by centrifugation at ϳ500 g for 10 min at 4°C. After three washes with isotonic saline, cells were resuspended in 10 mM Tris, pH 7.4, at 6 ϫ 10 7 cells/ml with 250 mM sucrose and placed on ice for 25 min before being homogenized by seven passages through a Balch ball-bearing homogenizer at 4°C. Nuclei and unbroken cells were removed by centrifugation at 1,000 ϫ g for 10 min. The post-nuclear supernatant was centrifuged at 10,000 ϫ g for 10 min at 4°C to pellet the non-microsomal membrane fraction. Microsomes from the supernatant were isolated by centrifugation at 100,000 ϫ g for 1 h at 4°C. Both microsomal and non-microsomal membrane pellets were resuspended (3 ϫ 10 5 cell equivalents/l) in 1.5% Triton X-100 containing Complete EDTA-free protease inhibitor mixture (Roche Applied Science) and stored at Ϫ80°C in 20% glycerol.
Western Analysis-Protein concentrations were determined for membrane fractions and cell lysates by the Bio-Rad Dc assay. Digestion of 100 -200 g of protein with 500 units of PNGase F was performed at 37°C overnight. Digests were diluted in Laemmli SDS-PAGE sample loading buffer (20) containing ␤-mercaptoethanol (final concentration, 0.1 M), boiled 5 min, and electrophoresed under reducing conditions on denaturing Tris-glycine SDS-polyacrylamide gels. Gels (7.5% or 10%) were either cast using Mini-Protean (Bio-Rad) or Sturdier (Hoefer Scientific Instruments) apparatus or were purchased from Bio-Rad. Benchmark, MagicMark, or Mark12 protein molecular weight markers (Invitrogen) were loaded in separate lanes. Staining of proteins was with Coomassie Brilliant Blue dye or Gelcode Blue stain (Pierce Biotechnology). For immunoblotting, proteins were electrotransferred overnight to a Polyscreen polyvinylidene difluoride membrane (PerkinElmer Life Sciences) in Tris-glycine buffer containing 8 -12% methanol. To assess transfer and protein loading, membranes were sometimes stained with Ponceau S. Subsequent incubations were for 1-2 h at room temperature or overnight at 4°C. Membranes were blocked in 5% nonfat dry milk (Nestle) in Tris-buffered saline (150 mM NaCl, 10 -20 mM Tris, pH 7.3-7.5) with 0.05% Tween 20 or Nonidet P-40 detergent before addition of antibodies. KM8621 and C-17 anti-Fut9 antibodies were used at 1:1,000 -2,000 and 1:200 -400 dilutions, respectively. Anti-tubulin antibody was used undiluted and antibody KM2681 was diluted 1:1. Secondary antibodies were used at 1:5,000 -10,000 dilutions. Chemiluminescent detection was by Renaissance TM (PerkinElmer Life Sciences) or SuperWest Pico TM (Pierce Biotechnology) reagents and exposure to Biomax TM MR x-ray films (Kodak).
Anti-VIM-2, Anti-sLe X , and Anti-Le X Binding to Cells by Flow Cytometry-Approximately 2 ϫ 10 5 cells washed twice with PBS (pH 7.2, calcium-and magnesium-free) or Hanks' buffered salt solution (HBSS, 0.4 g/liter KCl, 0.06 g/liter K 2 HPO 4 , 8 g/liter NaCl, 0.048 g/liter Na 2 HPO 4 , 1 g/liter D-glucose) were preincubated in 200 -400 l of PBS or HBSS containing 2% BSA at 4°C for 30 min. Cells were washed once with 3 ml of PBS or HBSS with BSA and resuspended in 200 -400 l of PBS or HBSS with BSA with 1-2 g of CSLEX-1, anti-SSEA-1, or anti-VIM-2 monoclonal antibody. After incubation at 4°C for 30 min, cells were washed once and incubated with 1-2 g of rabbit anti-mouse IgM conjugated to FITC. After 30 min at 4°C, cells were washed once and resuspended in 200 l of PBS/BSA for analysis using a FACScan flow cytometer (BD Biosciences). Propidium iodide was added at 1-2 g/ml just prior to cytometry. All fluorescence data were collected using logarithmic amplification on 10,000 -15,000 events (cell counts). Events considered for analysis were gated for either light scatter or low propidium iodide binding, or both. For double staining, an allophycocyaninconjugated, goat anti-mouse secondary antibody (0.3-1 g) was used instead of the FITC secondary antibody. Cells were washed with HBSS and incubated with 2 g of FITC-conjugated anti-Le X antibody in a volume of 400 l for 1 h at 4°C. Fluorescence cytometry was performed on a FACSCalibur TM cytometer (BD Biosciences) after a final wash with HBSS.
Rapid Amplification of cDNA Ends-The GeneRacer kit (Invitrogen) was used for 5Ј-RACE of LEC12 and LEC29 Fut9 transcripts. Reverse transcription of LEC12 and LEC29 mRNA was performed using a mixture of Superscript II TM and avian myeloblastosis virus reverse transcriptases at 48°C. RNase H-treated cDNA was amplified with Platinum TM Pfx polymerase using Fut9 gene-specific primer PS432 (5Ј-TAG TGA GAT GGC ACC CTT GG) and GeneRacer TM 5Ј-primer specific for the ligated oligonucleotide (5Ј-CGA CTG GAG CAC GAG GAC ACT GA). The PCR products were subjected to nested PCR using Fut9-specific primer PS537 (5Ј-TGT TGG TGG GCT TGA TGT AA) and GeneRacer 5Ј-nested primer (5Ј-GGA CAC TGA CAT GGA CTG AAG GAG TA). PCR products were TA-cloned into pCR2.1 vector, and the inserts of at least three clones were sequenced using vector-specific M13R and T7 primers.
Generation of LEC12 and LEC29 Full-length Fut9 cDNA Constructs-Full-length Fut9 cDNA, containing the 5Ј-UTR, 3Ј-UTR, and coding regions, was amplified from LEC12 cDNA using Platinum Pfx polymerase with primers PS616 (5Ј-CAC TCG AAG TTT CTG CCA GCC TGC CCT CTG CTC TCC TAA TGG A) and PS519 (5Ј-TTT TAA TGA AAG AAT AGA CAT TTC AAG). Taq polymerase was used to add 3Ј-A overhangs to the purified product, which was cloned into pcDNA3.1/V5-His TOPO vector (Invitrogen). Sequencing confirmed that the 5Ј-UTR was intact. The plasmid with a Fut9 cDNA insert of 2251 bp was designated pCMV-Full12F9. Similarly, a cDNA containing the 5Ј-UTR and coding regions of LEC29 Fut9 was obtained using primers PS617 (5Ј-TCT GCT TCT AGG ACA GCG CCG CCA CCG CCG AAA GC) and PS343 (5Ј-TTA ATT CCA AAA CCA TTT CTC TAA ATT ACC CAC AG). This cDNA and pCMV-Full12F9 were digested with KpnI and BspEI restriction enzymes. KpnI cuts pcDNA3.1/V5-His TOPO once, just upstream of the cloning site, whereas BspEI cuts both LEC12 and LEC29 Fut9 cDNA once, at positions 793 and 792, respectively. The KpnI/BspEI-released fragment from the LEC29 clone was cloned into KpnI/BspEI-digested pCMV-Full12F9 to generate the LEC29 full-length construct pCMV-Full29F9 with an insert size of 2250 bp. The LEC12 and LEC29 constructs were identical to each other, except for the first exon in their respective 5Ј-UTRs.
Generation of Myc-tagged Fut9 and Fut11 Constructs-The Fut9 gene ORF was amplified from pCMV-Full12F9 with Platinum Pfx polymerase using primers PS665 (5Ј-CCA CCA TGG AAC AAA AAC TCA TCT CAG AAG AGG ATC TGA CAT CAA CAT CCA AAG GCA TTC TTC GTC C) and PS343 (5Ј-TTA ATT CCA AAA CCA TTT CTC TAA ATT ACC CAC AG). The former also encodes the c-Myc peptide sequence EQKLISEEDL. AmpliTaq polymerase was used to add 3Ј-A overhangs to the purified product, which was then TA-cloned into pCR3.1 (Invitrogen). An N-terminal Myc-tagged mouse Fut11 expression construct was similarly generated by amplification from mouse placenta cDNA using primers PS686 (5Ј-CCA CCA TGG AAC AAA AAC TCA TCT CAG AAG AGG ATC TGG CTG CTC GCT GTA CCG AGG CGG TGC TGG CCG) and PS553 (5Ј-ATC CTC GAG GAT TTT TAT TCC GTT TCA TGA AGA TCT).

Generation of Humanized Renilla Luciferase Reporter Constructs-
The CMV promoter region of pSecTag2C-hygro vector (Invitrogen) was amplified with Platinum Pfx polymerase using primers PS650 (5Ј-ATG CAT CTC GAG CGA TGT ACG GGC CAG ATA TAC GCG, with an XhoI restriction site) and PS651 (5Ј-ATG CAT GGC GCC GTT AGC CAG AGA GCT CTG CTT ATA TAG, with an NarI restriction site). XhoI/ NarI-digested PCR products were cloned into phRL-null vector (Promega) between the XhoI and NarI sites to obtain phRL-CMV. The phRL-null vector has a T7 promoter upstream of humanized Renilla luciferase coding region such that the putative transcription start site is 12-nt upstream of the ATG start codon. For phRL-CMV, the CMV promoter-driven transcriptional start site is 260 nt upstream. Complete 5Ј-UTRs of LEC12 and LEC29 Fut9 were amplified from pCMV-Full12F9 and pCMV-Full29F9 cDNAs, respectively. The forward primers, bearing the T7 promoter region and a NarI restriction site were for LEC12: PS647 (5Ј-ATG CAT GGC GCC TAA TAC GAC TCA CTA TAG GGA GCA CTC GAA GTT TCT GCC AGC CTG CC); and for LEC29, PS649 (5Ј-ATG CAT GGC GCC TAA TAC GAC TCA CTA TAG GTC TGC TTC TAG GAC AGC GCC GCC ACC G). The reverse primer PS639 (5Ј-GCA TGC TAG CAA TTT TTT ACT GTA GAG AAC AGC ATG G) had an NheI restriction site. NarI/NheI-digested PCR products were cloned between the NarI and NheI sites of phRL-null to generate phRL-T7U12 and phRL-T7U29, T7 promoter-driven luciferase constructs fused with LEC12 and LEC29 Fut9 5Ј-UTRs, respectively. In both of them, the transcription start site is 2 nt upstream of the first base of the 5Ј-UTR whose end lies 11 nt upstream of the luciferase ATG start codon. To generate the CMV promoter-driven luciferase con-structs, phRL-CMVU12 and phRL-CMVU29, which have the luciferase coding region fused with the 5Ј-UTRs of Fut9, a similar strategy was used, except that the forward primers lacked the T7 promoter region: LEC12, PS646 (5Ј-ATG CAT GGC GCC GAG CAC TCG AAG TTT CTG CCA GCC TGC C); LEC29, PS648 (5Ј-ATG CAT GGC GCC TCT GCT TCT AGG ACA GCG CCG CCA CCG). In both these constructs, the transcription start site is 5 nt upstream of the first base of the 5Ј-UTR whose end lies 11 nt upstream of the luciferase ATG start codon.
In Vitro Transcription-Capped RNA was produced from linearized plasmid DNA using Ribo m7G cap analog (Promega) with the T7 RNA polymerase-based RiboMAX in vitro transcription kit (Promega). Plasmids were linearized with XhoI for pCMV-Full12F9 and pCMV-Full29F9, and XbaI for phRL-null, phRL-T7U12, and phRL-T7U29. The quality and quantity of purified RNA was assessed by spectrophotometry at 260 nm and electrophoresis on agarose gel followed by ethidium bromide staining.
Preparation of Translational Extracts from CHO, LEC12, and LEC29 Cells-The method of Favre and Trepo (22) that incorporates sucrose and dimeric creatine kinase in lysis buffer was followed. Cells were grown to 90% confluence on a 15-cm plate. After washing once with 12 ml of cold wash buffer (150 mM sucrose, 33 mM NH 4 Cl, 7 mM KCl, 45 mM Mg[CH 3 COO] 2 , 30 mM HEPES, pH 7.4), 4.5 ml of cold wash buffer containing 100 g/ml 1-stearoyl-2-hydroxy-sn-glycero-3-phosphocholine (lysophosphocholine) was added evenly over the plate at room temperature. It was aspirated after 90 s and 0.45 ml of cold extraction buffer (100 mM sucrose, 100 mM HEPES-KOH, pH 7.4, 120 mM KCH 3 COO, pH 7.4, 2.5 mM Mg[CH 3 COO] 2 , 1 mM dithiothreitol, 2.5 mM ATP, 1 mM GTP, 0.1 mM S-adenosylmethionine, 1 mM spermidine, 20 mM creatine phosphate, 40 units/ml creatine phosphokinase, 40 M each amino acid mix) was spread over the cells. After 90 s at 4°C, cells and buffer were scraped off and passed twice through a 26-gauge needle. After centrifugation at 100 ϫ g for 2 min, the supernatant was aliquoted and stored at Ϫ80°C. Protein concentration was measured using Bio-Rad Dc assay with BSA as standard.
In Vitro Translation-The rabbit reticulocyte lysate-based TNT-coupled transcription-translation kit (Promega) was used as per the manufacturer's protocol with both RNA and DNA templates. Translation reactions using 15 l of CHO cell extracts were performed in 20 l containing in vitro transcribed RNA and 1 l (50 g) of creatine phosphokinase. When required, [ 35 S]methionine or EasyTag Express TM 35 S protein labeling mix (PerkinElmer Life Sciences) was added to labeltranslated products. Translation reactions were directly used in luciferase assays or electrophoresed on SDS-PAGE gels. Gels were either used for immunoblotting or dried and exposed to BioMax TM MR x-ray films (Kodak).
Luciferase Assays-Cells grown in 6-well plates were transfected with Renilla reniformis (sea pansy) luciferase reporter constructs using Fugene6 DNA transfection reagent. For normalization of transfection efficiency, a firefly (Photinus pyralis) reporter construct containing the E promoter of the mouse immunoglobulin heavy chain gene (a kind gift of Dr. Barbara Birshtein, Albert Einstein College of Medicine) was cotransfected. Activities from both luciferases were measured sequentially from a single sample using the Dual-Luciferase reporter assay system from Promega. Cells were lysed after 2-3 days using 250 l of Passive Lysis buffer (Promega). Lysates were assayed immediately or after storage at Ϫ20°C. For in vitro translation reactions, samples were diluted 5-to 20-fold in Passive Lysis buffer. Measurements were performed on a Turner Designs TD-20e luminometer with a 2-s pre-read delay and a 10-s measurement period.
Transfection of CHO Cells with RNA-A TransMessenger TM RNA transfection kit from Qiagen was used to transfect adherent cells in 6-well plates as suggested by the manufacturer. For each well, 1-2 g of RNA was complexed with the lipid-based transfection reagent. Cell monolayers, at 40 -60% confluence, were washed once with 4 ml of PBS, and the RNA/transfectant complex was added for 2-3 h. The plates were then washed with 4 ml of PBS before adding 2 ml of ␣-minimal Eagle's medium/10% fetal bovine serum. Cells were harvested 18 -24 h later.
Binding of Mouse E-selectin-IgG Chimera-To measure mouse Eselectin-IgG binding (16), 8 ϫ 10 5 cells preincubated in HBSS/2% BSA for 1 h at 4°C, were incubated with 6 g of human IgG1 or E-selectin-IgG chimera protein for 1 h at 4°C in 400 l of HBSS/2% BSA containing 1 mM CaCl 2 or 1 mM CaCl 2 and 5 mM EDTA. Cells were washed once with HBSS/2% BSA and resuspended in 400 l of HBSS/2% BSA containing 3 g of FITC-conjugated goat anti-human IgG for 1 h at 4°C. They were washed with HBSS and analyzed for FITC fluorescence on a FACSCalibur instrument.

LEC29 Cells Do Not Possess Inhibitors of VIM-2 Synthesis-
LEC29 CHO cells have more Fut9 gene transcripts than LEC12 but ϳ40-times less ␣(1,3)Fuc-T activity (9,11). LEC12 cells express the VIM-2 epitope on cell surface glycoproteins, but VIM-2 is barely detectable on LEC29. To determine whether activation of a negative factor or repression of a positive factor down-regulates expression of VIM-2 on the LEC29 mutant, somatic cell hybrids formed with LEC12 cells were examined for VIM-2 expression. This approach previously showed that CHO Fut6 genes may become active due to a cis mechanism correlated with gene rearrangement or a trans mechanism due to the loss of a negative regulatory factor that represses Fut6B expression in CHO cells (8). Neomycin or zeomycin resistance genes were introduced into LEC12 and LEC29 cells, allowing post-fusion selection for hybrid colonies in medium containing both antibiotics. Upon analysis by fluorescence cytometry, independent LEC12 ϫ LEC29 hybrids were found to be as positive as LEC12 cells for cell surface VIM-2 expression (Fig. 1). Therefore, LEC29 cells do not express a dominant factor that represses VIM-2 synthesis. LEC29 could be due to a defect in the generation of polylactosamines or ␣(2,3)sialylation or ␣(1,3)fucosylation. It is also possible that one or just a few glycoconjugates carry the VIM-2 epitope. Thus the anti-sLe X mAb CSLEX-1 appears to detect one major sLe X -carrying protein in CHO-K1 cells transfected with a Fut7 cDNA (23). To determine whether overexpression of Fut9 in LEC29 would generate VIM-2 as it does in CHO cells (11), the coding region of the Fut9 gene cloned into pCR3.1 (11) was transiently transfected into LEC29 cells. Fut9 transfectants were positive for cell surface VIM-2 expression by fluorescence cytometry ( Fig. 2A). Thus LEC29 cells have the acceptors necessary to generate the VIM-2 antigen. Coexpression of Le X and VIM-2 structures was also examined on individual cells in transfectant populations stably expressing a Fut9 cDNA by fluorescence cytometry. Depending on the site of integration in the genome and number of integrated copies, cells of these stable populations will vary in their expression of Fut9 protein. In Fig. 2B, it can be seen that there was a wide range of expression of fucosylated antigens. However, there was a direct correlation between Le X and VIM-2 expression in both CHO and LEC29 transfectant populations. This indicates that the level of Fut9 must be increased in LEC29 cells to allow synthesis of VIM-2.
Fut9 Gene Transcripts in LEC12 and LEC29 Cells-Although Fut9 gene transcripts can be detected by RT-PCR in LEC12 and LEC29, a signal was not observed for LEC12 by Northern blot analysis (11). Therefore, a semi-quantitative RT-PCR study was performed to compare the relative levels of Fut9 gene transcripts in the two cell lines. Total RNA of LEC12 or LEC29 was reverse-transcribed, and the cDNA was used as template for amplification of glyceraldehyde-3-phosphate dehydrogenase or Fut9 genes by PCR for different numbers of cycles. Forward and reverse primers were designed to recognize sequences on separate exons to ensure amplification of cDNA but not genomic DNA sequences. Glyceraldehyde-3-phosphate dehydrogenase products were detected in the 27-cycle reactions for both LEC12 and LEC29, and product increased almost linearly to 42 cycles (Fig. 3A). Fut9 cDNA products were easily detected in the 39-cycle reaction for LEC29 but not until the 45-cycle reaction for LEC12. After a longer exposure (Fig. 3A), LEC12 products were visible at 33 cycles. Thus, LEC12 expresses ϳ10to 20-fold fewer Fut9 gene transcripts than LEC29.
LEC12 Cells Have Higher Levels of Fut9 Protein Than LEC29 -The CHO Fut9 protein has a deduced molecular mass of 42.2 kDa and three predicted N-glycosylation sites (11). Anti-Fut9 KM8621 antibody identified two major proteins in a reducing SDS-polyacrylamide gel (Fig. 3B). A Fut9 band at ϳ51 kDa was readily detected in LEC12, was present at a very low level in LEC29, and was missing from CHO lysates. The ϳ55-kDa protein marked with a star in Fig. 3B is a nonspecific band unrelated to Fut9, because it is present in CHO cells that do not express the Fut9 gene (11). Fut9 was further resolved from the nonspecific protein by treatment with PNGase F to remove N-glycans prior to SDS-PAGE. The position of the nonspecific band was not altered by this treat-FIG. 3. LEC12 cells have fewer Fut9 transcripts but more Fut9 protein than LEC29 cells. A, semi-quantitative RT-PCR of Fut9 RNA. LEC12 or LEC29 cDNA was amplified using glyceraldehyde-3-phosphate dehydrogenase or Fut9 gene-specific primers in reactions with different numbers of PCR cycles. Products were separated on an agarose gel and stained with ethidium bromide for imaging. The lowest panel is a longer exposure to better reveal LEC12 Fut9 transcripts. B, Western analysis was performed after electrophoresis in a 7.5% reducing SDS-PAGE gel and transfer to membrane, using anti-human Fut9 mAb KM8621. Each lane had 100 g of protein from 1.5% Nonidet P-40 detergent extracts made from CHO, LEC12, or LEC29 cells. PNGase F treatment was used prior to electrophoresis to remove N-glycans. The star indicates a non-Fut9 cross-reacting band present in all CHO cells. C, microsomes and a non-microsomal fraction were prepared from CHO, LEC12, and LEC29 cells as described under "Materials and Methods," and 50 g of protein from each were treated with PNGase F overnight. Immunoblotting to detect Fut9 was performed with mAb KM8621 after electrophoresis in a 7.5% reducing SDS-PAGE gel and transfer to membrane. The lower panel shows a longer exposure of the microsomal fraction. D, Fut9 protein is not selectively degraded in LEC29 cells. CHO, LEC12, and LEC29 cells were transiently cotransfected with expression plasmids for mouse Fut9 and Fut11 proteins with Myc tag at the N terminus. Cell extracts in 1.5% Nonidet P-40 were prepared 3 days after transfection. Western analysis using an anti-Myc antibody was performed after electrophoresis of 100 g of protein in a 7.5% reducing SDS-PAGE gel and transfer to polyvinylidene difluoride membrane. ment while the Fut9 band moved to ϳ43 kDa (Fig. 3B). A probable Fut9 dimer at ϳ80 kDa was present only in LEC12. Further evidence for Fut9 assignment was obtained by cell fractionation. Microsomal membranes from post-nuclear cell extracts were enriched in Fut9 protein and contained much less of the ϳ55-kDa unknown protein compared with the non-microsomal fraction (Fig. 3C). This was the expected location of Fut9, which functions in the Golgi compartment. The amount of Fut9 protein was at least 20-fold higher in the LEC12 microsomal fraction compared with LEC29, and it was not detected in CHO. One explanation for this difference is that Fut9 protein in LEC29 gets degraded fast. Because available antibodies against Fut9 protein were ineffective for immunoprecipitation, pulse-chase studies to estimate the half-life of Fut9 protein in cells could not be performed. However, transfected myc-tagged Fut9 protein was found not to be selectively degraded in LEC29 cells (Fig. 3D). Differences at the level of translation were therefore investigated.
Sequence Differences in the 5Ј-UTR of Fut9 Transcripts in LEC12 and LEC29 -The coding region (1080 bp) of Fut9 cDNA from LEC12 and LEC29 have an identical sequence (11). Sequences of the 3Ј-UTR (969 bp) of Fut9 cDNA from LEC12 and LEC29 and the 92 bases immediately upstream of the coding region were also identical for the two sequences. This region of the hamster 5Ј-UTR is 70 -80% identical to 5Ј-UTR sequences of mouse and rat Fut9 genes. However, the sequence further 5Ј was completely different in LEC12 and LEC29 cDNAs (Fig.  4A). The segment (1-109 nt) of the LEC29 5Ј-UTR sequence is most similar (80 -90%) to 5Ј-UTRs of Fut9 genes from mouse and rat, strongly suggesting that the LEC29 sequence represents the 5Ј-UTR of the Chinese hamster Fut9 gene. By contrast the LEC12 sequence (1-110 nt) appears to have come from another chromosome. A BLAST query showed that regions on mouse chromosome 6 and on rat chromosome 4 have significant similarity (75-85%) to the first 110 bp of the LEC12 5Ј-UTR sequence. Because the Fut9 gene is located on a different chromosome from this sequence in mouse and rat (chromosomes 4 and 5, respectively), it appears that a translocation occurred upstream of the Fut9 gene in LEC12 cells.
A 217-nt 5Ј-UTR from the Fut9 gene expressed in the LEC30 CHO mutant (11) was also obtained by the GeneRacer strategy. This mutant has more Fut9 gene transcripts and ϳ5ϫ more Fut9 protein than LEC12 (11). 3 The first eight bases upstream of the LEC30 Fut9 coding region were identical to the 5Ј-UTR sequence of LEC12 and LEC29. However, the next stretch of 180 bp is 87% identical to the 5Ј-region of the 6.4-kb Chinese hamster Intracisternal A Particle 34 element sequence (24) postulated to be a defective retroviral sequence. It is estimated that there are 80 copies of the Chinese hamster Intracisternal A Particle element per haploid genome of CHO cells (24).
The identity of the first eight bases upstream of the coding region of the Fut9 gene in all three mutants, and of the next 84 bases in LEC12 and LEC29, indicates that the CHO Fut9 gene is composed of at least three exons, similar to human, mouse, and rat, which have three Fut9 exons separated from each other by large introns (ϳ90 -120 kb). The coding region and 3Ј-UTR lie in the third exon in each species. RT-PCR using a forward primer specific for either the LEC12 or LEC29 Fut9 exon 1 (PS616 or PS617, respectively) and a reverse primer specific from the Fut9 coding region showed that the LEC12 and LEC29 5Ј-UTRs differ from each other in the first exon (Fig. 4B). No products were obtained using the same primers with genomic DNA as template. A newly identified 121-nt exon that is present in mouse cDNAs from embryonic stem cells (GenBank TM accession number AK049122) and cecum (AK033607) between mouse exons 1 and 2 was not present in the LEC12 or LEC29 cDNAs.
The sizes of the LEC12 and LEC29 5Ј-UTR sequences ob- tained using the GeneRacer 5Ј-RACE strategy to amplify cDNA molecules containing a capped 5Ј-end are consistent with the sizes of full-length Fut9 transcripts in CHO cells (11). In addition, primer extension analysis performed on LEC29 mRNA with two primers gave major products that were ϳ10 bases longer or ϳ35 bases shorter than obtained by 5Ј-RACE (data not shown). Fut9 RNA from both LEC12 and LEC29 was found equivalently associated with polysomes by RT-PCR of fractions across a polysome gradient (data not shown), and thus effects of the different 5Ј-UTRs on translation were investigated.
The LEC29 Fut9 5Ј-UTR Efficiently Translates a Reporter Transcript-To investigate effects of the different LEC12 and LEC29 5Ј-UTRs on translation efficiency, reporter cDNA constructs containing the respective 5Ј-UTRs upstream of Renilla luciferase and downstream of a T7 or CMV promoter were generated. CHO, LEC12, or LEC29 cells were transfected, and cell lysates were analyzed after 2-3 days. Surprisingly, all three cell types expressing the LEC29 5Ј-UTR construct had 3-5 times more activity than those transfected with the LEC12 5Ј-UTR reporter (Fig. 5A). There was no apparent difference in reporter gene transcripts between transfectants based on Northern analysis (data not shown). When the same cDNA constructs were assayed by coupled transcription-translation using the rabbit reticulocyte lysatebased TNT system, the presence of either the LEC12 or LEC29 5Ј-UTR inhibited translation ϳ4 -5 times compared with a construct that had no 5Ј-UTR. However, in this case there was no significant difference between luciferase activity generated from LEC12 and LEC29 5Ј-UTR constructs (data not shown).
Similar results were obtained from translation of in vitro generated RNA by translation extracts prepared from CHO, LEC12, or LEC29 cells, respectively. m 7 G-capped reporter RNA synthesized in vitro using T7 RNA polymerase was translated by each translation extract (Fig. 5B) and by a rabbit reticulocyte lysate (data not shown). As expected, the absence of a 5Ј-UTR improved translation considerably. However, the different LEC12 and LEC29 5Ј-UTRs did not differentially affect translation of the in vitro synthesized reporter luciferase RNA by any of the cell extracts (Fig. 5B).
The LEC29 5Ј-UTR Inhibits Translation from Full-length Fut9 cDNA Constructs-The absence of an inhibitory effect of the LEC29 Fut9 5Ј-UTR in luciferase reporter studies suggested that the 5Ј-UTR may interact with other regions of the transcript to regulate Fut9 mRNA translation. To address this, full-length Fut9 cDNA from LEC12 and LEC29 containing the 5Ј-UTR, coding region, and 3Ј-UTR, and differing only in the region corresponding to exon 1, were cloned into the pcDNA3.1/V5-His TOPO mammalian expression vector. As shown in Fig. 6A, lysates from LEC29 cells transfected with vector had a small amount of Fut9 protein detected by a polyclonal antibody. There was no nonspecific band with this commercial antibody. LEC29 cells transfected with the Fut9 ORF had the largest amount of Fut9 protein, including the dimer form. However, full-length LEC29 cDNA gave rise to much less Fut9 protein than observed in LEC29 cells transfected with full-length LEC12 cDNA. The levels of ␤-tubulin in each lysate were equivalent. In vitro ␣(1,3)Fuc-T enzyme activity in cell lysates were 90.2 (Fut9 coding region), 17.2 (full-length LEC12), and 5.3 (full-length LEC29 cDNA) pmol/mg of protein/ min, respectively, correlating approximately with differences in Fut9 protein levels.
To determine if the reduced translation of LEC29 Fut9 cDNA observed in transfectants was reflected by in vitro assays, full-length Fut9 cDNAs were subjected to in vitro transcription/ translation in the coupled TNT reticulocyte lysate system. The coding region of the hamster Fut9 gene has six in-frame translation start codons that could generate multiple products (Fig.  6B). Three major species were obtained from the Fut9 coding region construct. Both LEC12 and LEC29 full-length Fut9 constructs gave fewer products, with a major band at 37.4 kDa in each case. The intensity of this band was less for full-length LEC29 Fut9 cDNA.
To investigate further, in vitro translation of full-length, capped Fut9 RNAs was examined using translation extracts prepared from CHO, LEC12, or LEC29 cells (Fig. 6C). Addition of in vitro synthesized RNAs to each translation extract resulted in several translation products of different sizes. The nonspecific band at ϳ55 kDa detected by this mAb (Fig. 3B) is marked with a star and provides evidence for relative loading. The ϳ42-kDa band present only in LEC12 extracts probably represents endogenous, unglycosylated Fut9 protein, because it was absent in the "-RNA" reactions of CHO and LEC29 extracts. Also, the intensity of this band was not increased when RNA was added, apparently because translation of this species did not occur efficiently in the extracts. Signals from the ϳ37-, ϳ32-, and ϳ26-kDa species with full-length LEC12 Fut9 RNA were more intense than those with full-length LEC29 Fut9 RNA in reactions with all three translation extracts, even though the CHO extract was not as active as LEC12 or LEC29 extracts. It is clear from these in vitro translation assays that full-length LEC29 Fut9 RNA with 5Ј-and 3Ј-UTRs is translated less efficiently than full-length LEC12 Fut9 RNA.
To assay full-length Fut9 RNAs in vivo, LEC29 cells were transfected with capped, in vitro synthesized, full-length LEC12 and LEC29 Fut9 RNA. Cell lysates prepared 20 h after transfection were assayed for ␣(1,3)Fuc-T activity (Fig.  6D). LEC12 RNA gave more ␣(1,3)Fuc-T activity than the LEC29 RNA. Taken together, these findings suggest that the 5Ј-UTR of LEC29 Fut9 mRNA acting together with its coding or 3Ј-UTR region reduces translation of Fut9 gene transcripts giving rise to less Fut9 protein and less ␣(1,3)Fuc-T activity.
CHO Cells Expressing Fut9 Bind a Mouse E-selectin-IgG Chimera-LEC11 CHO cells express sialyl-Le X (9,13,25) and bind E-selectin (8,26). Although LEC12 cells do not express sialyl-Le X (9,13,25) and bind at only background levels to activated human endothelial cells (26), LEC12 cells bind to human E-selectin overexpressed in CHO cells (albeit less well than LEC11 (27), and a small subset of glycoproteins from LEC12 bind a mouse E-selectin-IgG chimera (28). Flow cytometry with the latter mouse E-selectin-IgG chimera showed that LEC12 CHO cells bound mouse E-selectin (Fig. 7A), 4 and LEC12 cells bound ϳ2% of the amount of E-selectin bound by LEC11 cells (Fig. 7A). LEC29 cells that make good amounts of Le X but almost no VIM-2 ( Fig. 2) did not bind mouse E-selectin (Fig.  7A). However, when CHO or LEC29 cells were transiently transfected with an Fut9 coding region cDNA, binding of mouse E-selectin was induced. Control human IgG1 did not bind to the transfectants. Furthermore, E-selectin binding was abolished in the presence of 1 mM CaCl 2 and 5 mM EDTA, a condition that does not induce membrane shedding of glycoproteins (data not shown). Both mouse and human Fut9 have been shown not to generate sLe X following transfection, nor to fucosylate ␣(2,3)sialylated LacNAc structures in vitro (29,30). In agreement with this, LEC29 cells overexpressing Fut9 did not bind the CSLEX-1 mAb that recognizes sLe X , although LEC29 cells were capable of synthesizing sLe X if transfected with a Fut6 cDNA (Fig. 7B). This shows that levels of Fut9 expressed in LEC12 cells and induced by transfection of a Fut9 cDNA in CHO or LEC29 cells cause the synthesis of a fucosylated ligand for mouse E-selectin likely to be VIM-2. In vitro transcribed, capped, full-length LEC12 (12) or LEC29 (29) Fut9 RNA (1.5 g) was translated in translation extracts prepared from CHO, LEC12, or LEC29 cells, and products were separated by SDS-PAGE, transferred to membrane, and immunoblotted with monoclonal KM2681 anti-Fut9 mAb. The star identifies the nonspecific band identified by this antibody. D, LEC29 cells were transfected with in vitro transcribed, full-length, LEC12, or LEC29 Fut9 RNA (2 g) using the TransMessenger reagent. Cell lysates in 1.5% Triton X-100 were prepared after 20 h (50 l of lysate per plate) and assayed for ␣(1,3)Fuc-T activity using LacNAc as substrate. Error bars indicate mean and range of data acquired in duplicate assays of lysates from duplicate transfections. Similar results were obtained in two experiments. and it was considered unlikely that they expressed the same Fut gene (9). They have different lectin resistance phenotypes, express a different spectrum of ␣(1,3)fucosylated cell surface antigens, and have very different levels of ␣(1,3)Fuc-T activity in cell extracts. Thus it was surprising to find that both mutants express the Fut9 gene (8,11). However, the mutant with the most Fut9 gene transcripts, LEC29, had the least ␣(1,3)Fuc-T activity. Here we identify a molecular basis for the discrepant phenotypes of LEC12 and LEC29, provide insight into translational control of Fut9 gene transcripts, and reveal potential biological consequences of regulated expression of ␣(1,3)Fuc-TIX. A comparison of LEC12 and LEC29 properties is summarized in Table I.
Here we show that LEC12 and LEC29 Fut9 genes differ in the sequence of their first exon. The LEC29 Fut9 exon 1 is highly related to exon 1 in mouse and rat Fut9 and human FUT9 and thus most likely represents the endogenous hamster Fut9 sequence. The LEC12 Fut9 exon 1 sequence differs completely, and, based on sequence comparisons, probably arose from a chromosomal translocation, as did the exon 1 of FIG. 7. A, mouse E-selectin binds to CHO cells overexpressing Fut9. LEC11 and LEC12 CHO mutants and LEC29 or CHO cells transiently transfected with empty vector (ϩ vector), or with vector containing the coding region of Fut9 (ϩ Fut9) were analyzed by fluorescence cytometry for binding of 6 g of mouse E-selectin-IgG fusion protein or human IgG1 detected with anti-human IgG1 secondary antibody. LEC29ϩFut9 cells were also tested in the presence of EDTA as described under "Materials and Methods." The inset enlarges the area of difference in the LEC29 ϩ Fut9 panel. B, sialyl-Le X is not generated on LEC29 cells overexpressing a Fut9 cDNA. LEC29 cells that were transiently transfected with empty vector (ϩ vector), or with vector containing the coding region of the Fut6B (8) or Fut9 Chinese hamster genes (LEC29 ϩ Fut9, left panel; LEC29 ϩ Fut6, right panel) were analyzed by fluorescence cytometry for binding of 3 g of CSLEX-1 mAb.
Fut9 expressed in LEC30 CHO cells. Fut9 mRNA in LEC12 and LEC29 associated to the same extent with polysomes, and there was no apparent difference in the intracellular stability of Myc-tagged Fut9 protein in these cells (Fig. 3D). Therefore, the translational abilities of the respective 5Ј-UTRs were compared. Interestingly, the LEC29 5Ј-UTR ligated to a luciferase reporter facilitated translation better than the LEC12 5Ј-UTR, whereas the opposite result was expected. It was subsequently found that the LEC29 5Ј-UTR is inhibitory to translation if it is expressed in the context of a full-length Fut9 cDNA or RNA. In this context the LEC29 5Ј-UTR inhibited in vitro translation of Fut9 transcripts and produced less Fut9 protein in transfectants than the LEC12 5Ј-UTR in full-length constructs. These results are summarized in Table II.
The characteristics of 5Ј-UTR sequences that affect translation include their length, Kozak consensus residues near the ATG start codon, secondary structure, and upstream ORFs (31). LEC12 and LEC29 Fut9 5Ј-UTR sequences are 210 and 209 nucleotides, respectively, but differ in sequence in exon 1. LEC12 exon 1 is 55% GC-rich, whereas LEC29 exon 1 is 68% GC-rich. Based on the mFold algorithm (32), LEC12 exon 1 has a ⌬G value of Ϫ46.33 kCal/mol, whereas the LEC29 exon 1 has a ⌬G value of Ϫ50.36 kCal/mol. The same start codon appears to be used in both cells, because Fut9 proteins from the two cell lines are indistinguishable in molecular weight. The LEC12 5Ј-UTR has three translation start codons, only one of which lies in a Kozak consensus sequence. The small upstream ORF of 20 codons does not have ATG in a Kozak consensus sequence. The LEC29 5Ј-UTR has four upstream ATGs. The only one in a Kozak consensus sequence initiates a small ORF of 15 nucleotides. The ability of eukaryotic ribosomes to reinitiate translation at a downstream ATG is limited by the size of upstream ORFs. Luukkonen and colleagues (33) have estimated the cut-off length as 30 codons. The LEC29 upstream ORF of 5 codons is 29 codons upstream of the start codon of the Fut9 coding region, and it is possible that it inhibits reinitiation of translation from that codon. The LEC29 exon 1 sequence could bind factors that inhibit translation or it could interact with the coding or 3Ј-UTR region to form an inhibitory structure. For example, cis inhibition of p53 RNA translation in yeast is caused by a pseudoknot structure that forms by interactions between the 5Ј-UTR and the ORF (34). The cis regulatory sequences that affect the translation of mRNA are generally located in the 3Ј-UTR and recruit factors that repress translation by different mechanisms (35).
Based on databases searches, human and mouse Fut9 gene transcripts include 5Ј-splice variants that might be differentially translated and result in variable levels of ␣(1,3)Fuc-TIX activity in vivo. ␣(1,3)Fuc-TIX is primarily responsible for the synthesis of the Le X determinant in mature granulocytes (37), in brain (38), and in the mouse embryo (39). However, our results show that sufficient levels of ␣(1,3)Fuc-TIX may also synthesize VIM-2 in a cell, and this could be biologically relevant in vivo. Thus, LEC12 CHO cells that express VIM-2 bind low levels of mouse E-selectin (39 and Fig. 7), whereas LEC29 cells that express only Le X do not (Fig. 7). In addition high level expression of Fut9 in CHO transfectants correlates with increased levels of VIM-2 expression (Fig. 2) and induces a low degree of binding of mouse E-selectin (Fig. 7). VIM-2 on glycolipids has been identified as a binding determinant for Eselectin (40 -42). It seems possible, therefore, that VIM-2 glycans mediate binding of E-selectin to the CHO Fut9 cDNA transfectants and thus may also do so under certain circumstances in vivo. However, regardless of the nature of glycans involved, the data presented here suggest that ␣(1,3)Fuc-TIX, like ␣(1,3)Fuc-TIV and ␣(1,3)Fuc-TVII, may act as a regulator of E-selectin-mediated processes.