Purification and characterization of an alpha1,2,-L-fucosyltransferase, which modifies the cytosolic protein FP21,from the cytosol of Dictyostelium.

A novel fucosyltransferase (cFTase) activity has been enriched over 106-fold from the cytosolic compartment of Dictyostelium based on transfer of [3H]fucose from GDP-[3H]fucose to Galβ1,3GlcNAcβ-paranitrophenyl (paranitrophenyl-lacto-N-bioside or pNP-LNB). The activity behaved as a single component during purification over DEAE-, phenyl-, Reactive Blue-4-, GDP-adipate-, GDP-hexanolamine-, and Superdex gel filtration resins. The purified activity possessed an apparent M of 95 × 103, was Mg-dependent with a neutral pH optimum, and exhibited a K for GDP-fucose of 0.34 μM, a K for pNP-LNB of 0.6 mM, and a V for pNP-LNB of 620 nmol/min/mg protein. SDS-polyacrylamide gel electrophoresis analysis of the Superdex elution profile identified a polypeptide with an apparent M of 85 × 103, which coeluted with the cFTase activity and could be specifically photolabeled with the donor substrate inhibitor GDP-hexanolaminyl-azido-I-salicylate. Based on substate analogue studies, exoglycosidase digestions, and co-chromatography with fucosylated standards, the product of the reaction with pNP-LNB was Fucα1, 2Galβ1,3GlcNAcβ-pNP. The cFTase preferred substrates with a Galβ1,3 linkage, and thus its acceptor substrate specificity resembles the human Secretor-type α1,2-FTase. Afucosyl isoforms of the FP21 glycoprotein, GP21-I and GP21-II, were purified from the cytosol of a Dictyostelium mutant and found to be substrates for the cFTase, which exhibited an apparent K of 0.21 μM and an apparent V of 460 nmol/min/mg protein toward GP21-II. The highly purified cFTase was inhibited by the reaction products Fucα1,2Galβ1,3GlcNAcβ-pNP and FP21-II. FP21-I and recombinant FP21 were not inhibitory, suggesting that acceptor substrate specificity is based primarily on carbohydrate recognition. A cytosolic location for this step of FP21 glycosylation is implied by the isolation of the cFTase from the cytosolic fraction, its high affinity for its substrates, and its failure to be detected in crude membrane preparations.

The glycosylation of proteins traversing the secretory pathway of eukaryotic cells has been well studied (1)(2)(3). Evidence has been steadily accumulating that some amino acid side chains of proteins of the cytosol and nucleoplasm are also modified, with either a single sugar, O-GlcNAc (3,4), or oligosaccharides (5)(6)(7)(8)(9). The frequency of O-GlcNAc modifications of cytosolic and nucleoplasmic proteins may approach that of phosphorylation (10), and there is evidence that cytosolic glycosylation may be regulatory (5, 8, 10 -12), based on the existence of glycoforms which vary with different physiological states and protein localizations, the presence of specific cytosolic glycosidases, half-lives of sugar modifications which vary relative to the host protein, and potential competition for attachment sites between O-GlcNAc and phosphate groups.
FP21 is a protein found both in the cytosol and the nucleus, and is modified by an oligosaccharide in the cellular slime mold Dictyostelium discoideum (7,8). FP21 is a highly conserved protein whose amino acid sequence shows 68% identity, 78% similarity, and one amino acid residue difference in length between Dictyostelium and humans. FP21 has been detected in the cyclin A/cdcK2 complex associated with the G1 checkpoint of the human HeLa cell cycle (13), the kinetechore complex (14) of budding yeasts, 1 and in association with certain membranes of Dictyostelium. 2 FP21 has a particularly high abundance in the inner ear organ of Corti (15)(16)(17). An FP21 gene is present in a green alga virus genome (18), and two copies are frequently found in eukaryotic genomes, including Dictyostelium. 3 Dictyostelium FP21 possesses a single tetra-or pentasaccharide, which appears to contains Fuc, 4 Xyl, and Gal (8) and is probably O-linked based on susceptibility to release by mild base (7) or hydrazinolysis. 2 Two FP21 isoforms have been purified which vary in their proportions of Xyl and Gal (8).
Since a GDP-Fuc synthesis mutant in Dictyostelium exhibits slow growth which can be partially rescued by exogenous Fuc, 2 we have investigated the fucosylation of FP21 for its possible involvement in this phenotype. Using afucosyl-FP21 from the fucosylation mutant as an acceptor substrate, an ␣-cFTase (cytosolic fucosyltransferase) activity dependent upon GDP-␤-Fuc was detected in the cytosolic S100 fraction but not in the particulate or membrane fraction (7). FP21 accounted for Ͼ80% of the acceptor activity in crude S100 extracts of the fucosylation mutant, as if it were the primary acceptor substrate for this enzyme. This activity was also able to attach Fuc to lacto-N-biose (Gal␤1,3GlcNAc␤-) attached to a hydrophobic aglycone moiety, as determined by cross-competition studies. This enzyme activity appeared to mediate attachment of a peripheral Fuc, and was novel in terms of its apparent cytosolic, rather than Golgi, compartmentalization, and its submicromolar K m for GDP-Fuc. As shown here, a 1.2 million-fold purified preparation of the enzyme activity retains its high affinity for GDP-Fuc, displays a high affinity for afucosyl FP21, links Fuc in a ␣1,2 linkage to a ␤1,3-linked Gal on type 1 and 3 acceptors, and copurifies with a polypeptide which can be photoaffinity-labeled with a donor substrate analog.

Cells
Strains Ax3 and HL250 were grown at 22°C in HL-5 medium on a gyratory shaker. In 6-liter flasks, cells grew to a density of 6 -10 ϫ 10 7 per ml, which is referred to as stationary phase. Ax3 is a normal strain and HL250, derived from Ax3 by chemical mutagenesis, is unable to synthesize GDP-Fuc from GDP-Man and thus incorporates negligible Fuc into protein (7,8). Cells were developed by washing logarithmically growing cells in 17 mM KH 2 PO 4 neutralized to pH 6.0 with NaOH (KP) and shaking in 6-liter flasks at 2 ϫ 10 7 cells/ml in KP for 12 h. Cells were synchronized with respect to the cell cycle by diluting stationary phase cells in a 500-ml flask to 10 6 cells/ml of fresh HL-5 in a 6-liter flask, as described previously (19).
Fucosyltransferase Assay-cFTase activity was usually assayed as the transfer of [ 3 H]Fuc from GDP-[ 3 H]Fuc to pNP-LNB, which was determined by adsorption onto a C 18 Sep-Pak cartridge followed by elution with methanol (20). GDP-Fuc was at a concentration 3-fold above its K m value, and pNP-LNB was at slightly below its K m value. Each assay tube routinely contained 9 g of pNP-LNB (equivalent ot a final concentration of 0.36 mM) dried from a 1 mg/ml aqueous solution in a 1.5-ml polypropylene microcentrifuge tube by vacuum centrifugation, 5 l of 20 mg/ml BSA in buffer D, and 40 l of enzyme sample diluted, if appropriate, in buffer D. During pilot studies, column fractions were assayed after desalting on PD-10 columns (Pharmacia) equilibrated in buffer D. These columns contain 9.1 ml of Sephadex G-25 medium. The reaction was initiated by the addition of 5 l of a mixture prepared in buffer H and containing, after dilution to 50 l, 1.0 M GDP-Fuc (1-8 ϫ 10 5 dpm), prepared from a mixture of GDP-[ 3 H]Fuc and unlabeled GDP-Fuc, 15 mM MgCl 2 , 2.0 mM MnCl 2 , 0.10 mM EDTA, 1.0 mM DTT, and 15% glycerol. The mixture is stable for Ն3 days at 0°C, and thus divalent cation-promoted degradation of GDP-Fuc (21) did not occur. When assaying crude cell extracts and DEAE column fractions, the mixture was supplemented with sodium fluoride and ATP, at final concentrations in the assay mixture of 2.0 mM and 1.0 mM, respectively. Assays of crude extracts containing sucrose lacked glycerol. Reactions containing enzyme preparation A were supplemented with 0.05% Tween 20. The reaction mixture was incubated for 10 -120 min at 30°C, and stopped by the addition of 1 ml of cold water. The reaction mixture was either immediately filtered over the C 18 Sep-Pak cartridges, or frozen at Ϫ80°C until filtration. Cartridges were reused until flow rates became unacceptable. Samples were filtered in groups of 10 on a Visiprep manifold (Supelco), which was alternated between a waste receptacle and a second receptacle containing 20-ml scintillation vials. Incorporation was determined as described previously (7).
Reactions involving protein acceptor substrates were conducted similarly, except that protein acceptors were introduced from concentrated stock solutions, Tween 20 was present at 0.05% (v/v), BSA was reduced to 0.5 mg/ml, and the reaction was initiated by the addition of the cFTase. Incorporation into protein substrates was determined after SDS-PAGE of the samples, within 1 day of fixation with Coomassie Blue and destaining, by excising the appropriate region of the gel and scintillation counting, as described (7).
Protein Determination-Protein concentration was determined by a commercial modification of a Coomassie Blue dye binding method (22), or from A 280 values, assuming an extinction coefficient of 1 ml/mg protein for a 1-cm path length.
Cell Lysis-1-2 ϫ 10 11 cells were grown to near maximum cell density. After washing once in ice-cold water, cells were resuspended in buffer A, suctioned through a bed of glass wool, and lysed by forced passage through a 47-mm diameter Nuclepore filter with 5-m diameter pores mounted on the end of a 60-ml syringe. The lysate was centrifuged at 4,000 ϫ g for 2 min, and the supernatant was centrifuged at 100,000 ϫ g for 60 min. All procedures were performed at 0 -4°C.
DEAE-Sepharose Fast Flow Chromatography-The final S100 supernatant was pumped onto a 450-ml DEAE-Sepahrose Fast Flow column (4.8 ϫ 28 cm) pre-equilibrated in buffer B at 250 ml/h, and the column was washed with buffer B until the A 280 dropped below 1% of its maximum. cFTase was eluted in a linear gradient of 0 -0.25 M NaCl, consisting of 1.25 liter of starting buffer B and 1.25 liter of limit buffer C. Protein was determined based on A 280 or the Coomassie Blue dye binding assay. Fractions from the main activity peak were pooled and frozen at Ϫ80°C.
Phenyl-Sepharose 6 Fast Flow Chromatography-DEAE pools from 1.5 ϫ 10 12 cells were thawed and pumped onto a 175-ml phenyl-Sepharose 6 Fast Flow (high-sub) column (2.6 ϫ 31 cm) at a flow rate of 275 ml/h. The column was washed with buffer D until the A 280 dropped below 1% of its maximum. cFTase was eluted by application of a linear 0 -60% gradient of ethylene glycol, consisting of 400 ml of starting buffer D and 400 ml of limit buffer E. Protein was determined based on A 280 and the Coomassie Blue dye binding assay. cFTase activity was determined after 5-fold dilution of fractions with buffer D. Fractions from the activity peak were pooled and concentrated 20-fold by ultrafiltration using a PM-30 membrane, and then diluted 5-fold in buffer D to reduce ethylene glycol to Ͻ10% (v/v). The DEAE and phenyl columns were reused after in situ washing with 8 M urea and 0.1 N NaOH.
Dye Column Affinity Chromatography-The phenyl-Sepharose pool was diluted to 10% (v/v) ethylene glycol with buffer D and pumped onto a 25-ml column (1.6 ϫ 25 cm) containing Reactive Blue-4 coupled to cross-linked 4% agarose beads (5.2 mg dye/ml), which had been sequentially precycled with 0.5 mg/ml bovine serum albumin in buffer D, buffer F, buffer D, and 10% (v/v) ethylene glycol in buffer D. After sample loading, the column was washed in buffer D until the A 280 dropped below 1% of its maximum, and eluted with a linear gradient of 0.1-1.5 M NaCl, consisting of 60 ml of starting buffer D and 60 ml of limit buffer F. Protein was determined based on A 280 and the Coomassie Blue dye binding assay. cFTase activity was determined after 5-fold dilution with buffer D. Fractions from the activity peak were pooled, concentrated 2-3-fold on a Centriplus-30 centrifugal concentrator, and diluted with buffer B to a final calculated concentration of 0.1 M NaCl.
GDP Affinity Chromatography-The Reactive Blue-4 pool was passed over a 1.2-ml GDP-adipate-agarose (1.6 mol GDP/ml packed resin) column connected in series with a 4.0-ml GDP-hexanolamineagarose (2.8 mol of GDP/ml of packed resin) column. The GDP moiety of GDP-adipate is linked via the 2Ј-or 3Ј-hydroxyl of the ribose, whereas the GDP moiety of GDP-hexanolamine is attached via its ␤-PO 4 , which corresponds to the linkage with Fuc in the native donor substrate GDP-Fuc. After washing in buffer D, the GDP-hexanolamine column was disconnected and eluted with a 0 -2 mM gradient of GDP, formed from buffers D and G. 97% of the activity was eluted before 1 mM GDP as determined by HPLC gel filtration, and was pooled and concentrated to Ͻ1 ml in a Centriplus-30 concentrator.
The final Superdex 200 activity pool, purified 1.2 ϫ 10 6 -fold, is referred to as preparation A. Activity purified 317-2200-fold through DEAE, phenyl, and Superdex resins is referred to as preparation B, activity purified 106-fold through DEAE, phenyl, and Reactive Blue-4 resins is referred to as preparation C, and activity purified 27-fold through the DEAE-resin is referred to as preparation D.

Fucosyltransferase Characterization
Photoaffinity Labeling-GDP-hex-ASA, generously provided by Eric Holmes, was iodinated in its salicylate moiety as described (23). Superdex 200 column fractions, 30 -300 l, were preincubated in a quartz 1-ml cuvette for 15 min in dim red fluorescent illumination in the presence of 30 M GDP-hex-125-I-ASA, Ϯ 700 M GDP-Fuc, in buffer D. Photolysis was effected by illumination at 254 nm with a 30-watt Mineralight for 30 s on opposite sides, as described (23). Each sample was centrifuged through two concentration/dilution cycles in a Microcon-10 centrifugal ultrafiltration concentrator. Laemmli sample electrophoresis buffer at 70°C. was added directly to the concentrated sample in the Microcon-10 device, and transferred together with a rinse aliquot to the SDS gel sample well.
SDS-PAGE-SDS-PAGE was performed on 7-20% linear acrylamide gradient gels using the Laemmli discontinuous buffer system, as described previously (8). 80-l aliquots from the Superdex 200 fractions were reduced in 5 mM DTT, treated with with 40 mM iodoacetamide to reduce artifactual bands (24), and then diluted with 4 ϫ concentrated Laemmli sample buffer. Gels were silver-stained according to Ref. 25.
pH Variation, Divalent Cations, and Inhibitors-Effects of pH variation, divalent cations, and potential inhibitors were determined on preparation B by diluting the enzyme 10-fold from buffer D into appropriate solutions. MES was used to buffer the pH range from 5.5 through 6.5; Bis-Tris propane was used from pH 6.5 to 8.5; and CAPSO was used from pH 8.5 to 9.5. Activities did not vary more than 10% between buffers at the crossover pH values 6.5 and 8.5. Inhibitors were preincubated with the enzyme for 15 min prior to the start of the assay.
Kinetic Studies-Reactions were conducted as described for FTase assays. Unless otherwise indicated, reaction mixtures contained 125 pg of preparation A protein as enzyme, 1 M GDP-Fuc, and the indicated amount of acceptor substrate, and were incubated for 1 h at 30°C.

cFTase Reaction Product Characterization
Synthesis of Fucosyl-pNP-LNB-The cFTase assay reaction was scaled up to synthesize suitable quantities of the reaction product for fucosidase digestion. To remove salts and proteins, the reaction product was adsorbed to a C 18 Sep-Pak, eluted with MeOH, dried, redissolved in water, and passed over a PD10 gel filtration column equilibrated in water. The radioactive peak was collected and the concentration of reaction product was calculated from the specific activity of GDP- Exoglycosidase Digestions-Reaction mixtures contained 120 pmol of [ 3 H]fucosyl-pNP-LNB (140,000 dpm), and the indicated amount of enzyme, in a final volume of 50 l of buffer composed according to the supplier's recommendation. Reactions were incubated at 37°C, and aliquots were assayed by determining the percent of total disintegrations/min which did not adsorb to a C 18 Sep-Pak cartridge.

FP21 Isolation
Purification of Dictyostelium FP21-FP21 isoforms-I and -II were purified over DEAE, phenyl, and monoclonal antibody 3F9 affinity columns from strain Ax3 S100 extracts, as described (8). Purification of FP21 from strain HL250 also yielded two isoforms similar in chromatographic behavior to the Ax3 isoforms, and are referred to as GP21-I and GP21-II. Protein was dialyzed against 50 mM HEPES-NaOH (pH 7.4), and concentrated in Centriprep centrifugal ultrafiltration concentrators (10-kDa molecular mass cut-off). Protein concentration was determined by amino acid composition analysis of phenylisothiocyanatederivatives after acid hydrolysis, using norleucine as an internal standard (26).
Preparation of recombinant Dictyostelium FP21-The open reading frame of FP21 was amplified using Dictyostelium DNA as the template and fp15 and fp14 as primers (see Table IV in Ref. 8) in a polymerase chain reaction. These primers contained BamHI restriction sites which were used to clone the amplified DNA into the BamHI restriction site of the inducible (27, 28) expression vector pET19b (Novagen, Madison, WI), downstream of the T7 RNA polymerase transcription element, such that an oligo-His tag was introduced at the NH 2 terminus. The deduced NH 2 -terminal sequence of rP21 is MGHHHHHHHHHHSS-GHIDDDDKHMLEDP followed by the natural FP21 sequence, which was verified by sequencing of the plasmid DNA (8). Expression host Escherichia coli BL21 (DE3) cells carrying a lysogen with a copy of the T7 RNA polymerase gene under lacUV5 control were transformed under carbenicillin selection. After induction with isopropyl-1-thio-␤-Dgalactopyranoside, expressing colonies were examined. Inclusion bodies were not observed. rP21 was isolated from clone rP21A under nondenaturing conditions using an affinity column consisting of nickel cations immobilized on Sepharose 6B, essentially as described (29). Purified protein was exhaustively dialyzed against 50 mM HEPES-NaOH (pH 7.4), and concentrated in a centrifugal ultrafiltration concentrator (10-kDa molecular mass cut-off). Purified protein (Fig. 7) was shown to be recognized in a Western blot by monoclonal antibodies 3F9 and 4E1, and anti-FP21(68 -82)/L97 (8), confirming its identity as rP21. Protein concentration was determined by amino acid composition analysis (see above).

Pilot Studies on Partially Purified
Enzyme-Pilot studies on partially purified enzyme established the conditions of the standard assay. After DEAE-ion exchange chromatography (27-fold purified; preparation D) or a combination of ion ex-change chromatography, phenyl-Sepharose hydrophobic interaction chromatrography, and gel filtration (2200-fold purified; preparation B) enzyme activities using pNP-LNB as substrate were linear with respect to time (0 -48 h) and protein concentration (data not shown). 70% of the activity maximum at pH 7.2 was retained over a broad pH range of 5.5 to 9.0, beyond which activity dropped rapidly. The activity was similar in Tris, Bis-Tris propane, MES, HEPES, or CAPSO buffers. Activity was optimal at concentrations of NaCl from 100 to 300 mM, but was stable for short times at concentrations up to at least 1.5 M NaCl, or (NH 4 ) 2 SO 4 at 20%. The enzyme was significantly less active or less stable in KCl. Of the divalent cations tested, as chlorides of Mg 2ϩ , Mn 2ϩ , Ca 2ϩ , Zn 2ϩ , Fe 2ϩ , and Co 2ϩ , only MgCl 2 and MnCl 2 supported activity. 0.5-2.0 mM Zn 2ϩ , Fe 2ϩ , and Co 2ϩ inhibited activity from 50 to 95% in the presence of 0.5 mM MgCl 2 . Optimal activity occurred in 15 mM MgCl 2 and 2.0 mM MnCl 2 . Sodium fluoride and ATP stimulated activity 20 -40% in crude extracts but inhibited activity 10 -30% in purified preparations, and thus were included only in the former preparations. Activity was inhibited 47% by 1.0 mM N-ethylmaleimide, and only 18% by 1.0 mM pyridoxal-5phosphate, using 0.27 M GDP-Fuc and 0.36 mM pNP-LNB as substrates, suggesting that the active site may contain an essential cysteine (36). Glycerol (up to 30% (v/v)), 1 mM DTT, and 0.1 mM EDTA (in the presence of a concentration excess of Mg 2ϩ ) did not inhibit activity, and thus were included as potential stabilizing agents. NaN 3 at 0.02% (w/v) led to a 25% reduction in activity and so was not used. Optimal temperature was 30°C, with lower activities detected at 37 and 24°C. These properties of the partially purified enzyme activity were similar to those previously determined for the unfractionated activity (7).
To test the effect of detergents, enzyme purified through the Reactive Blue-4 step (preparation C) was diluted 100-fold in buffer D and incubated in varying concentrations of 16 detergents as described under "Experimental Procedures." Only Tween 20 and Tween 80 sustained activity at all concentrations tested. Activity losses ranged from 30 to 99% for the other detergents; in general, greater losses were observed as detergent concentrations diminished below the critical micelle concentration. At later stages of purification, 0.05% Tween 20 was superior to BSA in preserving enzyme activity at 4°C.
Enzyme Purification-cFTase present in the S100 fraction was remarkably stable, with 90% activity remaining after 10 days at 4°C, or after 10 freeze (Ϫ80°C)/thaw cycles. Although activity could be recovered from desalted (NH 4 ) 2 SO 4 precipitates, the broad range (30 -70%) over which activity precipitated was not useful (data not shown). Activity quantitatively adsorbed to DEAE resins from the crude extract, and Ͼ98% of the eluted activity emerged at about 30 mM NaCl as a monodisperse peak ahead of the major protein peaks (Fig. 1A), thus resulting in a 25-30-fold purification ( Table I). The remaining activity emerged later as a small peak. The Fast Flow resin was superior to DEAE-Sephadex and DEAE-cellulose (fibrous or granular) because of rapid flow, resistance to clogging, and resistance to shrinkage at higher ionic strengths. The column was stable to multiple cycles of use and washing in 0.1 N NaOH. Enzyme also adsorbed at pH 6.8 but enrichment was reduced 2-fold. Enrichment was also diminished by Ͼ2-fold if the cell lysate was brought to 5 mM MgCl 2 and centrifuged for 1 h at one-half the g-force, compared to the standard method. As determined by HPLC gel filtration (see below), activity exhibited a single apparent M r of 95 ϫ 10 3 . 5 The pool of the DEAEpurified enzyme was stable to storage at 4°C and tolerated repeated freeze/thawing.
The DEAE pool adsorbed quantitatively to phenyl-Sepharose 6 Fast Flow (high sub) and Ͼ99% of the eluted activity emerged as a single peak with slight trailing close to the end of a 0 -60% gradient of ethylene glycol (Fig. 1B). cFTase activity was assayed in eluted fractions after dilution in buffer D, and recovery ranged from 50 to 90%, or 53% for the trial reported (Table  I). Quantitative adsoprtion to phenyl-Sepharose 6 Fast Flow (low sub), or to phenyl Sepharose CL-4B, required addition of 10 -20% (NH 4 ) 2 SO 4 to the DEAE pool, and thus these resins were not used.
The pool of phenyl-purified enzyme was concentrated by ultrafiltration. Recovery of activity after concentration varied from 10 to 90%, or 27% in the reported trial (Table I), and activity was not detected in the ultrafiltrate. At this stage of purification, the activity typically displayed an apparent M r value of 95 ϫ 10 3 after HPLC gel filtration (see below), and thus poor recovery appeared to be associated with insolubility or denaturation. In early trials with the other phenyl resins involving long column residence times, substantial activity with an M r value of 40 ϫ 10 3 was correlated with poor recovery of activity, 5 as discussed further below.
The phenyl-pool activity adsorbed efficiently to the the Reactive Blue-4 dye resin, and 97% of the eluted activity emerged as a monodisperse peak early in a gradient of 0.1-1.5 M NaCl (Fig. 1C). The remaining activity emerged later as a small peak, and overall activity recovery was 92%. cFTase activity could also be adsorbed to RB-1, RB-72, RG-5, RR-120, RG-19, and CB dye resins, but the Reactive Blue-4 resin was selected on the basis of more selective binding relative to BSA. RY-3 and RY-86 dye resins did not adsorb activity. Concentration of activity by ultrafiltration resulted in a 33% loss of activity.
The Reactive Blue-4-pool activity was unretarded by GDPadipate Sepharose CL-4B. 290 g of protein was recovered from the column after elution with 2 mM GDP, but was not examined further. Activity in the flow-through fraction was, however, quantitatively adsorbed to GDP-hexanolamine Sepharose. The selective binding to GDP-hexanolamine was predicted based on the inhibitor characteristics of a series of GDP-Fuc analogues (see below), and is a behavior typical of mammalian ␣1,2-FTases (30 -33). The GDP-hexanolamine column was eluted with a 0 -2 mM GDP gradient, and the activity eluted in the 0 -1 mM range of the gradient. Ͻ5% additional activity was eluted by various combinations of high GDP and NaCl concentrations, Tween 20 and ethylene glycol. After ultrafiltration, the enzyme activity was stable at 4°C for at least 1 days.
The GDP affinity pool was applied to a Superdex 200 gel filtration column by HPLC. 95% of the recovered activity eluted as a monodisperse peak (peak I) centered at a M r position of 95 ϫ 10 3 (Fig. 1D). An additional 3.7% eluted at M r position 40 ϫ 10 3 (peak II), and the remaining activity eluted near the void volume. In pilot studies, Ͼ90% of activity applied to the GDPhexanolamine column was recovered after gel filtration. The 35% recovery in the reported trial (Table I) did not seem to be due to retention on the column (see above), and may have been due to 3 days storage of the GDP-hexanolamine fractions at 4°C prior to further processing. After gel filtration, activity decayed with a 12-h half-life at 4°C, but was stable for 3 days in 0.05% (v/v) Tween 20 at 4°C. Activity was only partially stabilized by 2.0 mg/ml BSA, and strongly destabilized by purified cytochrome c (data not shown). Finally, activity was stabilized by freezing at Ϫ80°C.
HPLC gel filtration fractions were analyzed by SDS-PAGE and silver staining (Fig. 2). Three major silver-stained bands were visible in the fractions of highest activity (peak I), centered at 30.4-ml elution volume. Only the amount of the most 5 T. Scott-Ward and C. M. West, unpublished data. heavily stained band, at M r position of 85 ϫ 10 3 , correlated precisely with the activity of the fractions. To obtain further evidence that this polypeptide, FT85, was equivalent to the enzyme, selected HPLC gel filtration fractions were UV-irradiated in the presence of the photoactive donor substrate analogue GDP-hex-125 I-ASA, at 30 M, in the presence or absence of 690 M GDP-Fuc. Photolabeling under similar conditions has previously been useful in identifying fucosyl-and mannosyltransferases (23,35). SDS-PAGE followed by autoradiographic analysis of a fraction from peak I (30.4-ml elution volume) revealed that the most intensely photolabeled polypeptide, PL85, migrated at the position of FT85 (Fig. 3, lane e), and labeling of this band was inhibited by GDP-Fuc (lane f). PL85 was not detected in the fraction eluting at 28.6 ml (lane c), which possessed Ͻ0.2% of the peak activity, but was faintly detectable in the fraction eluting at 33.1 ml, which possessed 3.1% of the peak activity (not shown). Two additional, much less intensely photolabeled bands were also detected in the major activity peak at M r positions 40 and 29 ϫ 10 3 , and are discussed below. The relative intensities of these bands is clearer in Fig. 3, lane i, which is a shorter autoradiographic exposure. Multiple minor bands were also photoreactive in lane e, but in each case labeling was only slightly inhibited by GDP-Fuc (lane f), showing that their reactivity was nonspecific and that photoprotection by GDP-Fuc was not due to general UV absorbance by the nucleotide moiety.
The minor cFTase activity peak eluting at 37.7 ml (Fig. 1D, peak II; M r position of 40 ϫ 10 3 ) contained a single, specifically-  Table I. Protein, which was monitored by absorbance at 280 nm using a 0.5-cm path length flow cell, and cFTase activity, which was assayed as dpm in the presence of 1 M GDP-Fuc and 0.36 mM pNP-LNB, are plotted as a function of elution volume. Panel A, typical example of an S100 fraction which was pumped onto the DEAE-Sepharose Fast Flow resin. After washing, the column was eluted with a 0 -0.25 M gradient of NaCl, whose concentration at the entry point of the 450-ml column is plotted. Fractions from the main DEAE column activity peak which contained Ͼ500 dpm activity/aliquot were pooled and frozen. Panel B, activity pools from 10 DEAE columns were pooled and pumped onto the phenyl-Sepharose 6 Fast Flow (high sub) resin. After washing, the column was eluted with a 0 -60% (v/v) gradient of ethylene glycol, whose concentration at the entry point of the 175-ml column is plotted. Panel C, fractions from the main phenyl column activity peak which contained Ͼ4000 dpm/aliquot were pooled, concentrated, diluted to reduce the ethylene glycol concentration, and pumped onto the Reactive Blue-4 agarose dye resin. After washing, the column was eluted with a 0.1-1.5 M NaCl gradient, whose concentration at the entry point of the 25-ml column is plotted. Fractions from the main peak which contained Ͼ2000 dpm/aliquot were pooled, concentrated, and diluted to 0.1 M NaCl prior to application to the GDP-adipate and GDP-hexanolamine columns (not shown). Panel D, the GDP-eluate from the GDP-hexanolamine column was concentrated and applied to the Superdex 200 HPLC gel filtration column and eluted isocratically. 95.0% of the activity emerged as peak I centered at 30.4 ml, and 3.7% emerged as a minor peak centered at 37.7 ml, as peak II. Calibration of the Superdex column with M r standards (see "Experimental Procedures") indicated that peaks I and II had apparent M r values of 95 ϫ 10 3 and 40 ϫ 10 3 , respectively. labeled photoreactive species, referred to as PL40, at an M r position of 40 ϫ 10 3 (Fig. 3, lane g). Negligible PL85 was detectable in this fraction. The amount of PL40 varied with the level of cFTase activity in neighboring fractions (data not shown). PL40 was not detectable by silver staining (Fig. 2, 37.7 ml), as expected based on its low apparent abundance relative to PL85 (Fig. 3, compare lanes e and g). PL40, together with PL29, were also observed in peak I (Fig. 3, lane e). PL40 appears to be a proteolytic fragment of PL85 based on the following observations. 1) The ratio of PL40 to PL85 in the major peak I was time dependent. After 7 days of storage at 4°C, during which activity diminished by Ͼ90%, only PL40 could be detected (Fig. 3, lanes k-o).
2) The original photolabeling of peak I described above in Fig. 3 had been performed after 4 days of storage at 4°C, during which activity had decreased by 50%. SDS-PAGE analysis of the pooled peak I fractions at this time revealed that the level of FT85 had decreased concomitant with the appearance of the new bands labeled at FT40 and FT29 (compare Fig. 3, lane a, with Fig. 2, 30.4 ml). Thus the appearance of PL40 in the main activity peak is correlated with the appearance of FT40. 3) Pilot studies on phenyl-purified enzyme showed that cFTase activity became smaller as determined by HPLC gel filtration as activity decreased, 5 and that the M r 95 ϫ 10 3 activity peak I was less stable than the M r 40 ϫ 10 3 activity peak II. 6 We conclude that PL85 is equivalent to FT85, and that FT85 is susceptible to proteolytic degradation to FT40, which is equivalent to PL40. FT40 possesses similar K m values for its substrates and similar substrate specificity, but has a reduced specific activity compared to FT85 (see below). This model explains both the time-dependent ap-6 P. Teng-umnuay and C. M. West, unpublished data. The samples in lanes a-j had lost about 50% of their cFTase activity prior to photolabeling and SDS-PAGE, which explains the reduced abundance of FT85 and increased abundance of FT40 and FT29 in lane a relative to Fig. 2. Lanes k-o, peak I fractions from a separate gel filtration of the same cFTase preparation A were photolabeled after Ͼ90% loss of cFTase activity. The amount of original activity which was photolabeled corresponded to 2% of the original activity which was photolabeled in lanes c-j. The relative intensity of labeling of PL40 in the sequential fractions is approximately proportionate to the original activity in the fractions. 2 Together with the results shown in Fig. 2 and other data, this figure shows that FT40, and possibly FT29, are breakdown products of the intact cFTase polypeptide, FT85. These and other results (see text) suggest that reduction of cFTase activity occurs as FT85 is proteolytically degraded to FT40 and possibly FT29.

TABLE I Purification of cFTase activity
Purification of cytosolic GDP-Fuc:lacto-N-bioside fucosyltransferase activity from Dictyostelium cells. S100 extracts were prepared and chromatographed on DEAE-Sepharose Fast Flow. Active DEAE fractions were pooled and frozen at Ϫ80°C. Ten DEAE preparations were thawed and pooled for the subsequent purification steps. The final Superdex 200 activity pool is referred to as preparation A. Activity purified 317-2200-fold through DEAE, phenyl, and Superdex resins is referred to as preparation B, activity purified 106-fold through DEAE, phenyl, and Reactive Blue-4 resins is referred to as preparation C, and activity purified 27-fold through the DEAE-resin is referred to as preparation D.
c Minimum estimate derived from the next stage of purification.
pearance of FT40 in the M r 95 ϫ 10 3 activity peak after gel filtration, as well as the time-dependent appearance of the M r 40 ϫ 10 3 activity peak II before gel filtration. In summary, the cFTase activity was purified 1.2 million-fold at 4.3% yield. Some of the losses could be attributed to specific proteolysis, which may explain the small proportion of activity (Ͻ5%) which could be resolved from the main peak during DEAE, phenyl, Reactive Blue-4, and gel filtration chromatographies. The FT85 polypeptide represents about 30% of the total protein. The cFTase consisted of a single 95 ϫ 10 3 M r activity after the first chromatography step and, aside from inconsistent generation of the apparently proteolytically-derived 40 ϫ 10 3 M r activity, no other molecular species expressing this activity were detected during the purification.
Properties of the Enzyme-Fucosylation of pNP-LNB by the highly purified preparation A was linearly dependent on time and pNP-LNB concentration (data not shown). This preparation exhibited a K m for pNP-LNB of 0.58 mM at 1 M GDP-Fuc, compared to 0.73 mM for the unfractionated activity (7), and a V max of 620 nmol/min/mg protein (Fig. 4). Preparation A exhibited a K m for GDP-Fuc of 0.34 M (Fig. 5) at 2.0 mM pNP-LNB, about 4-fold lower than the value obtained for the unfractionated activity (7). The M r 40 ϫ 10 3 form which was generated in some trials exhibited similar kinetic properties, with K m values for GDP-Fuc of 0.23 M and for pNP-LNB of 0.77 mM, and was probably a degradation product of the higher M r form (see above).
Donor substrate Analogues-Nucleotides are often potent glycosyltransferase inhibitors and, similarly, GDP was a good inhibitor of the cFTase (Table II). Guanosine, GMP, and GTP were also good, but slightly weaker inhibitors, indicating that the 5Ј-phosphates are not important. Substituents on the 5Јphosphates are also not important, because GDP-Man and GDP-hex-ASA were good inhibitors. The ribose moiety is significant, because guanine was not a good inhibitor. The 2Јdeoxy group of ribose is not required for recognition, but bulky substituents are not tolerated, as 2Ј,3Ј-isopropylidene guanosine and 3Ј,5Ј-cGMP were poor inhibitors. Integrity of the ribose ring is essential as the periodate-oxidized 2Ј,3Ј-dialcohol was ineffective. The identity of the guanine base is important because 5Ј-phosphates of adenosine were not significantly inhibitory. Furthermore, the guanine moiety does not tolerate substituents at positions 6 -8. Similar results were observed for the M r 40 ϫ 10 3 form detected in some trials of preparation B. 5 These observations suggested that guanine ribonucleosides modified at the 5Ј-position of ribose may be useful for purifying and identifying the enzyme. These predictions were validated by the affinity chromatography and photoaffinity results described above.
Acceptor Substrate Analogues-The pNP-LNB acceptor substrate consists of Gal␤1,3GlcNAc␤ attached to a paranitrophenyl aglycone moiety. pNP was an important recognition determinant as Gal␤1,3GlcNAc at a 12-fold concentration excess inhibited the enzyme by only 50% (Table III). Benzyl-LNB and 8-methoxycarbonyloctyl-LNB exhibited only 15 and 9% of the activity of pNP-LNB at 1 mM (Table IV); these differences were attributable to changes in K m or V max depending on the aglycone (data not shown).
The cFTase preferred a disaccharide acceptor as negligible activity was observed with monosaccharide-pNP substrates. 2-O-methylation or ␣1,2-fucosylation of the nonreducing terminal Gal of pNP-LNB abolished activity (Table IV). The ␤1,3linkage of the Gal was important for enzyme recognition, as Gal␤1,4GlcNAc and Gal␤1,6GlcNAc were at best very weak inhibitors (Table III), and Gal␤1,4GlcNAc␤-pNP exhibited only several percent of the activity of pNP-LNB, even above the K m of pNP-LNB (Table IV). Finally, the enzyme was not specific for the identity or linkage of the penultimate sugar of the disaccharide, as Gal␤1,3GalNAc␣-benzyl was a good substrate (Table IV).
Characterization of the cFTase Reaction Product-The inability of the cFTase to fucosylate the 2-O-methylated derivative of lacto-N-biose suggested that the enzyme catalyzes the formation of a Fuc1,2Gal linkage (37). To test this possibility, the reaction product of pNP-LNB with preparation B of the enzyme, containing 3 H in its fucosyl moiety, was subjected to digestion with fucosidases. As previously observed (7), Fuc was readily released with nonspecific mammalian ␣-fucosidases capable of cleaving Fuc␣1,2Gal and Fuc␣1,3/4/6GlcNAc linkages (Table V). ␣1,2-Fucosidase from A. oxidans released Fuc with similar kinetics, whereas ␣1,3/4-fucosidase from almond emulsin did not show any activity. Although less almond emulsin fucosidase was present in these reactions, the calculated level of activity was still in vast excess of what would be required to hydrolyze an ␣1,4-linkage.
Protein Substrates-To determine whether the cytosolic protein FP21, isolated from mutant strain HL250 in its afucosyl form (GP21), is a substrate for the purified (preparation A) cFTase, GP21 isoforms-I and -II were purified to near homogeneity under nondenaturing conditions (Fig. 7). Since FP21 is not efficiently precipitated by trichloroacetic acid, acetone or ethanol, the reaction mixture was resolved by SDS-PAGE, and the GP21 band was stained and counted. GP21-I and -II were each fucosylated in a time (not shown)-and concentration-dependent fashion (Fig. 8). A Lineweaver-Burk analysis of GP21-II fucosylation (Fig. 8) estimated an apparent K m for GP21-II of 0.21 M, and an apparent V max of 460 nmol/min/mg protein. The apparent V max of the cFTase for GP21-II is about 75% of the apparent V max for pNP-LNB, whereas the apparent K m is over 3 orders of magnitude lower than the corresponding value for pNP-LNB. GP21-I fucosylation appeared to exhibit cooperativity with the Lineweaver-Burk plot tapering to the same y intercept value and slope as GP21-II at higher substrate concentrations.
Fucosylation of the FP21 polypeptide depended on its glycosylation status. Recombinant Dictyostelium FP21 (rP21), purified under nondenaturing conditions from E. coli with a short oligo-His tag at its N terminus (Fig. 7), exhibited no acceptor activity with preparation A (data not shown), consistent with previous evidence that the single Fuc on FP21 is located at the nonreducing terminus of an oligosaccharide (7,8), which is expected to be absent from rP21. FP21-I and FP21-II, isolated from normal Dictyostelium cells, were also not substrates, consistent with previous studies on crude extracts that soluble FP21 is fully fucosylated (7).
pNP-LNB and GP21 would reciprocally inhibit each other's fucosylation if they were substrates of the same enzyme. Both GP21-I and -II inhibited pNP-LNB fucosylation in a concentra-  Table I Table I  tion-dependent manner (Table VI), as expected based on analysis of crude S100 extracts (7). At a pNP-LNB concentration slightly below its K m , a concentration of GP21-I slightly above GP21-II's K m value inhibited pNP-LNB fucosylation approximately 40%, whereas a higher concentration of GP21-II was required for 40% inhibition. Conversely, much higher concentrations of pNP-LNB were required to inhibit GP21-I than GP21-II fucosylation. Thus both GP21 isoforms and pNP-LNB appeared to be fucosylated by the same enzyme, but GP21-I appeared to have a higher affinity than GP21-II for the cFTase. Recognition of GP21 by the cFTase appeared to depend primarily on carbohydrate rather than peptide determinants, based on the inhibitory potential of the different protein analogues. rP21 failed to inhibit the fucosylation of GP21-I, GP21-II, and pNP-LNB (Table VI), and in fact mildly stimulated the fucosylation of GP21-I and GP21-II. Since this stimulatory effect was not observed for pNP-LNB fucosylation, it may reflect substrate-substrate interactions, as 1) rP21 stimulated the fucosylation of GP21-I more than that of GP21-II, 2) GP21-I fucosylation was cooperative with respect to its own concentration (see above), and 3) GP21 and FP21 are known to aggregate in the concentration range examined (8). FP21-I was only a weak inhibitor (Table VI) compared to FP21-II (see below), showing an effect only at the highest inhibitor:substrate ratio tested against 0.028 M GP21-I (Table VI). The absence of inhibitory effects by both rP21 and FP21-I suggested that the enzyme does not recognize the polypeptide backbone of FP21.
In contrast, FP21-II was a good inhibitor of the fucosylation of all three substrates, pNP-LNB, GP21-I, and GP21-II, with Ͼ50% inhibition observed at 10 -100-fold concentration excess over the GP21-II and GP21-I substrate concentrations tested. The effect of FP21-II may represent product inhibition, as the reaction product, Fuc␣1,2Gal␤1,3GlcNAc␤-pNP, was found to be an inhibitor of pNP-LNB fucosylation, when tested over a similar range of inhibitor:substrate ratios (Table VI). The failure of FP21-I to exert inhibition, except at the highest concentrations, may be due to its extra mole of Gal (8) which, if applied following fucosylation, may mask recognition of the FP21-I oligosaccharide. If rP21 and the GP21 and FP21 isoforms differed only in their glycosylation, the simplest interpretation is that the specificity for fucosylation and inhibition of fucosylation resides in the sugar portion of the substrate.
Developmental Regulation-S100 extracts were prepared from proliferating, cell cycle synchronized, stationary, and developing cells (See "Experimental Procedures") and assayed in the presence of 1.0 M GDP-Fuc and 0.36 mM pNP-LNB. The specific activity of the cFTase was found to vary less than 50% in all samples tested, indicating that the enzyme is constitutively expressed.
FIG. 6. Co-chromatography of the pNP-LNB cFTase reaction product with Fuc␣1,2Gal␤1,3GlcNAc␤-pNP. pNP-LNB was fucosylated in the presence of GDP-[ 3 H]Fuc by preparation A of the cFTase and recovered on a C 18 Sep-Pak and by gel filtration. The 3 H-reaction product was combined with 10 g each of pNP-LNB and synthetic Fuc␣1,2Gal␤1,3GlcNAc␤-pNP, and the mixture was applied to a PA-1 column and eluted isocratically with 150 mM NaOH. Peak 1 is unknown material present in the reaction product, peak 2 is at the position of Fuc␣1,2Gal␤1,3GlcNAc␤-pNP, and peak 3 is at the elution position of pNP-LNB. No sugars or radioactivity were detected from 25-60 min. Ͼ95% of the radioactivity coeluted with Fuc␣1,2Gal␤1,3GlcNAc␤-pNP on this column, and also on a PA-100 column eluted with a gradient of NaAc in 0.1 N NaOH (not shown).   (30 -33, 40). Separate ␣1,2-FTases appear to be encoded by the H and Secretor loci in humans (33,39), and DNAs encoding the H and Secretor enzymes have been cloned (40 -43). Like all other known Golgi glycosyltransferases (45,46), the Secretor and H enzymes are synthesized as type 2 membrane proteins consisting of a large C-terminal catalytic ectodomain, attached via a linker region to a transmembrane domain and a small N-terminal endodomain. The soluble, secretory forms of Golgi glycosyltransferases are apparently derived by proteolytic cleavage within the linker region (45). It is not clear whether the secretory form is physiologically meaningful. The K m values of Golgi and Golgi-derived FTases for GDP-Fuc are in the range of 10 -100 M, and apparent K m values for oligosaccharide acceptor substrates range from 0.1 to 20 mM (30 -33, 40). The order of magnitude of these K m values is consistent with the expected concentrations of these substrates in the Golgi apparatus.
The Dictyostelium cFTase, which is also capable of catalyzing the formation of a Fuc␣1,2Gal linkage, shows both similarities to and differences from the mammalian ␣1,2-FTases. The cFTase occurs in soluble form after gentle cell lysis that maintained the latency of Golgi enzymes, and its activity could not be detected in particulate extracts of the cell (7). Its polypeptide could be purified to apparent homogeneity using conventional and affinity chromatography and electrophoresis methods which have been successfully applied to secreted forms of mammalian enzymes. Although the cFTase appears to be fairly hydrophobic when chromatographed on phenyl-Sepharose columns and certain detergents stabilized its activity after purification, detergents were not used during purification. The apparent M r of the cFTase polypeptide (FT85), at 85 ϫ 10 3 , is approximately twice that of known Golgi enzyme cleavage products. The implication that the cFTase diverges from the Golgi type 2 membrane protein structural paradigm and resides in the cytosol must await confirmation from sequencing and cloning studies.

TABLE VI
Inhibition of cFTase Reactions were carried out using preparation A in the presence of 1.0 M GDP-[ 3 H]Fuc and the indicated concentration of acceptor substrate and inhibitors, for 1 h at 30°C. Fucosylation of pNP-LNB was assayed using a C 18 Sep-Pak cartridge, and fucosylation of GP21-I and GP21-II was assayed by SDS-PAGE and counting of gel bands. The Sep-Pak assay did not detect GP21 fucosylation, and the gel assay did not detect pNP-LNB fucosylation. rP21, FP21-I, and FP21-II were not fucosylated under the conditions of the assay. The K m of the purified cFTase for GDP-Fuc, at 0.34 M, is unusually low compared to Golgi FTases, but similar to the value of a cytosolic O-GlcNAc transferase for its donor substrate UDP-GlcNAc (34). The high affinity of the cFTase for GDP-Fuc is consistent with its proposed cytosolic localization, because in the cytosol the cFTase would be in competition with the Golgi GDP-Fuc transporter (2) for this substrate. Although the cFTase differs from mammalian Golgi ␣FTases in its high affinity for GDP-Fuc, its nucleotide binding is more similar to ␣1,2-FTases than ␣1,3-FTases in its preference for GDP-hexanolamine over GDP-adipate (31). The proportionately high sensitivity of the cFTase to inhibition by other guanine nucleotides (Table II) raises interesting regulatory questions about the effects of their intracellular pools.
The K m of the purified cFTase for the type 1 disaccharide conjugate pNP-LNB (Gal␤1,3GlcNAc␤-pNP) is similar to that of mammalian ␣1,2-FTases. The monosaccharide conjugate pNP-Gal, and the type 2 disaccharide conjugate Gal␤1,4Glc-NAc␤-pNP are only very poorly if at all fucosylated, and Gal␤1,6GlcNAc is at best a poor inhibitor. In contrast, the enzyme will fucosylate the type 3 disaccharide Gal␤1,3Gal-NAc␣-pNP, thus tolerating alternate sugars of opposite anomeric linkage penultimate to the terminal, fucosylated Gal. The acceptor specificity preferences of the cFTase are similar to that of highly purified porcine submaxillary ␣1,2-FTase (30,44) and the Secretor-type ␣1,2-FTase purified from human serum or plasma (31)(32)(33), and distinct from the H-type ␣1,2-FTase, which has a broader specificity range (31)(32)(33)40). Speculation that the Secretor-type gene preceded a gene duplication event leading to divergent evolution of the H-type gene (39) is reinforced by the present identification of a Secretor-type enzyme in the cellular slime molds. The specificity of the cFTase is compatible with what is known about the FP21 oligosaccharide, which contains Gal. However, the underlying sugar is not GlcNAc or GalNAc, as amino sugars are not detected in FP21 (8).
The cFTase exhibits varying V max and K m values for different aglycone moieties attached to the disaccharide. Of the three which have been tested, pNP has the highest activity, benzyl is intermediate, and the 8-methoxycarbonyloctyl moiety exhibits an order of magnitude lower activity compared to pNP when present at concentrations near or below their K m values. This suggests that acceptor substrate recognition involves more than the terminal disaccharide, which is consistent with the lower than expected inhibitory potential of free lacto-N-biose and the 3000-fold lower apparent K m of the native substrate GP21. Effects of alternate aglycone moieties have also been observed for mammalian FTases (31,33) and other Golgi glycosyltransferases (47), and polypeptide domains have been shown to be important determinants of recognition by certain glycosyltransferases (48). However, rP21 and FP21-I are not inhibitors of the cFTase, suggesting that the FP21 polypeptide is not an important determinant for cFTase recognition. The higher affinity of the enzyme for GP21-I compared to GP21-II, as determined by cross-inhibition studies (Table VI), in contrast to the higher inhibitory potential of FP21-II compared to FP21-I, may be most easily explained by glycosylation differences which have been described for the two FP21 isoforms (8). Why the different isoforms are differentially glycosylated remains to be determined. Since GP21 is the major acceptor of the reaction in crude extracts from the GDP-Fuc synthesis mutant (7), high affinity recognition of the FP21 carbohydrate implies that the carbohydrate structure is of limited distribution, and may be required because FP21 is not concentrated in a compartment. This is reminiscent of the relationship between UDP-Glc:glycoprotein Glc-1-phosphotransferase, another cyto-solic glycoprotein glycosyltransferase involved in peripheral modifications, and phosphoglucomutase, which appears to be its predominant acceptor substrate (5).
The V max of the purified cFTase with respect to pNP-LNB was 620 nmol/min/mg protein, at 1.0 M GDP-Fuc. The V max would be expected to extrapolate to 830 nmol/min/mg at saturating GDP-Fuc, or about 2.5 mol/min/mg with respect to the content of FT85 in the highly purified preparation A. This value is greater than the estimated specific activity of a highly purified Se-type FTase from serum (33), but less than that of a purified porcine Golgi ␣1,2-FTase (30). Projecting back to the cell, the specific activity suggests that there are on the order of 200 copies/cell of the cFTase in the cytosol. If the approximately 2 ϫ 10 5 copies/cell of the acceptor substrate FP21 2 were dissolved in the full volume of the cell, FP21 would be at a concentration of about an order of magnitude above its K m for fucosylation. If the cFTase were operating at its V max , there would be enough enzyme to be able to fucosylate all FP21 in about 1 h.