Engineering of CHO cells for the production of vertebrate recombinant sialyltransferases

Materials

All materials were purchased from Sigma-Aldrich Merck (Gillingham, UK) unless otherwise stated. Vectors pCR8/GW/TOPO and pSecTag/FRT/V5-His-TOPO, and the Gateway (GW) vector conversion system were obtained from Invitrogen (Carlsbad, CA, USA). LR Clonase II and the Flp-In-CHO cell line were also from Invitrogen (Carlsbad, CA, USA). FITC-labelled lectins SNA-I and MAA were from EY Laboratories (San Mateo, CA, USA) and biotinylated SNA-I, RCA-I, and MAA lectins were from Vector Laboratories (Peterborough, UK). Oligonucleotide primer synthesis and DNA sequencing was carried out by Eurofins. Proof-reading Phusion High-Fidelity DNA Polymerase (New England Biolabs, Ipswich, MA, USA) was used for PCR amplification of cDNA during construction and standard Taq polymerases (Promega, Madison, WI, USA) for screening. PNGase F was from New England Biolabs (Ipswich, MA, USA).

Construct design

Protein sequences (Table 1) were aligned in MEGA-X (Fig. S1). Only the soluble, cytoplasmic catalytic domains of the SIATs were cloned because we wanted these proteins to be secreted into the media. In order to make all the cloned proteins as comparable as possible we defined the catalytic domains of ST6Gal I and ST6Gal II as starting 121 amino acids N-terminal to a Cys that is totally conserved across all vertebrate SIATs, and which is part of a disulfide bridge that links the L- and S-motifs (Cys-184 in human ST6Gal I NP_003023.1) (Fig. 1). In addition there will be linking amino acids as a consequence of the recombination sequences that are part of the GW cloning system (Table S1). Most proteins have identical sequences through their N-terminal fusions to the SIAT catalytic domain (ending SEFAL). Slight variations in ST6Gal I (zebrafish) fusion sequences ending at KLAL, ST6Gal I (stickleback) ending at AGS and ST3Gal IV (zebrafish) with an additional QKKW sequence are a consequence of an evolving cloning approach (Table S1) and are explained further below. For comparative purposes we have also constructed full-length human and zebrafish ST6Gal I expression vectors. These constructs, based on pcDNA5/FRT/V5-His-TOPO (Invitrogen, Carlsbad, CA, USA), do not support active secretion from CHO cells, however, the catalytic domain of ST6Gal I is detected in cell culture supernatants (Donadio et al., 2003; Monaco et al., 1996).

Human ST6GAL1 cDNA was sourced from the HepG2 cell line while human ST6GAL2 was sourced from the RPMI cell line. Other STGAL cDNAs were isolated from species include mammals—rat (Rattus norvegicus); birds—chicken (Gallus gallus); fish—fugu (Takifugu rubripes), stickleback (Gasterosteus aculaeatus), and zebrafish (Danio rerio). Total RNA (2 μg) was isolated from appropriate tissues or cells using the RNeasy Kits (Qiagen, Hilden, Germany) following the manufacturer’s recommended protocol (Table S2). Contaminating genomic DNA was removed using an on-column DNase I treatment using an RNase-free DNase Kit (Qiagen, Hilden, Germany). RNA was reverse-transcribed using Superscript VILO (Invitrogen, Carlsbad, CA, USA) as detailed in the manufacturer’s protocol.

Cloning

Primers were designed to each SIAT cDNA ensuring that, after recombination, the gene would be fused in-frame to the N-terminal Igκ secretion signal sequence as well as to the C-terminal V5-His epitope and tag. SIAT fragments were amplified by PCR from cDNA, gel purified and ligated into a GW entry vector (pCR8/GW/TOPO). The GW vector conversion system (Invitrogen, Carlsbad, CA, USA) was used to convert pSecTag/FRT/V5-His-TOPO to a corresponding pSecTag/FRT/GW destination vector. In brief, a GW cassette containing attR recombination sites flanking a ccdB gene and a chloramphenicol-resistance gene, was amplified by PCR using specific primers and ligated into pSecTag/FRT/V5-His-TOPO vector. Ligations were transformed into ccdB survival competent cells, allowing propagation of the GW destination vector containing the ccdB gene. SIAT cDNA fragments were then transferred to this destination vector (pSecTag/FRT/GW) by LR recombination. LR Clonase (Invitrogen, Carlsbad, CA, USA) II was used to catalyze the in vitro recombination between the entry GW clone (containing SIAT flanked by attL sites) and the GW destination vector (containing attR sites). After recombination these sites become attB1 and attB2 sites. Clones were sequenced to confirm in-frame fusion with the N-terminal leader and C-terminal tags. As noted above, there were some variations to this routine procedure leading to some linking sequence differences. Zebrafish st6gal1 was cloned directly into the pSecTag/FRT/V5-His-TOPO, and stickleback ST6GAL1 was recombined into our pSecTag/GW destination vector from a PCR-generated product containing added recombination sites in a related procedure (Fu et al., 2008). Either this led to the absence of a GW sequence in the zebrafish ST6Gal I fusion protein or a shorter GW sequence in the stickleback ST6Gal I protein. Zebrafish ST3Gal IV, for unknown reasons, included an additional short sequence which fortunately maintained reading frame.

Transfection and culture of Flp-In-CHO cells and generation of stable cell lines

Flp-In-CHO cells (Invitrogen, Carlsbad, CA, USA) were grown in Ham F12, 10% Foetal Bovine Serum (FBS), L-glutamine (2 mM) containing 100 μg/mL zeocin. For stable transfections we used a 9:1 (w/w) ratio of pOG44 (1.8 μg) (Invitrogen, Carlsbad, CA, USA) to vector (0.2 μg) and a 1:3 ratio of total DNA (2 μg) to Fugene 6 (6 μL) (Promega, Madison, WI, USA). The plasmid pOG44 expresses the Flp recombinase that recombines the pSecTag derivative plasmid FLP Recombination Target (FRT) sites with the corresponding sites in the Flp-In-CHO cell chromosome. Transfections were carried out in six-well plates (Nunc, Roskilde, Denmark) containing respectively 3 × 10⁵ cells. Media was removed from adherent, low passage number Flp-In-CHO cells at approximate 60% confluence and replaced with fresh Ham F12, 10% FBS, 2 mM L-glutamine in the absence of zeocin. After 24 h incubation at 37 °C media was replaced with fresh Ham F12, 10% FBS, L-glutamine (2 mM) containing 600 μg/mL hygromycin (optimized previously). Media was changed every 2 days until foci appeared at approximately day 10 (approximately 99% of non-transfected Flp-In-CHO cells were dead at this stage). At 3 weeks, cells were rinsed with phosphate-buffered saline (PBS), treated with Gibco™ TrypLE (Thermo Fisher Scientific, Waltham, MA, USA) under standard procedures, resuspended with 2 mL Ham F12, 10% FBS, 2 mM L-glutamine, 600 μg/mL hygromycin and transferred to a T25 flask containing 10 mL of media. These cells were subsequently split every 3–4 days to T75 flasks and stocks were frozen (liq. N₂) at passage 5. Once stable lines were established, Flp-In-CHO cells were weaned into ProCHO-AT (Lonza, Basel, Switzerland) containing 150 μg/mL hygromycin for several passages over 4–6 weeks. Cells grew strongly in ProCHO-AT but with both adherent and suspension phenotypes. For upscaling, cells were split and grown in T175 triple flasks (500 cm²) in ProCHO-AT to provide 1 L of media for experiments (5 × 200 mL).

Lectin microscopy

Cells were seeded (3 mL, 1.2 × 10⁶ cells) in six-well plates containing sterile coverslips and incubated under standard CHO culture conditions to approximately 60% confluence. PBS-washed cells were fixed with 4% paraformaldehyde. Cells were washed three times with a modified TBS buffer, mTBS (20 mM Tris–HCl, pH 7.2, 100 mM NaCl, 1 mM MgCl₂, 1 mM CaCl₂), blocked with 1% bovine serum albumin (BSA, ≥99%, globulin-free)(Sigma-Aldrich Merck, Gillingham, UK) for 30 min then washed as before with mTBS. Fluorescein isothiocyanate (FITC)-labelled lectins (EY Laboratories, San Mateo, CA, USA) SNA-I and MAA (containing both MAA-1 and MAA-2) were prepared in mTBS (40 μg/mL), added to coverslips (0.5 mL, 20 μg), and incubated for 2 h at 4 °C in the dark in the absence or presence of 1 mL 100 mM lactose. Plates were washed three times with mTBS and then with PBS. The coverslip with the attached cells was mounted on a glass slide with one drop of ProLong Gold antifade mountant containing 4′,6-diamidino-2-phenylindole (DAPI) (Thermo Fisher Scientific, Waltham, MA, USA) and allowed to dry overnight in the dark. Images were acquired using an upright microscope (Olympus BX53) with violet (autofluorescence/DAPI) and green (FITC) filters.

Sialyltransferase assay

The Sialyltransferase Activity Kit (R&D Systems, Abingdon, UK) was used essentially as described by the manufacturer and as described elsewhere (Meng et al., 2013; Wu et al., 2011). Our standard assay conditions, used for comparative purposes across a range of sialyltransferase preparations, were 2-(N-morpholino) ethanesulfonic acid (MES) buffer, pH 6.5 (100 mM), MnCl₂ (10 mM), CD73 (50 ng), CMP-Neu5Ac (donor substrate), acceptor and 2 μL sialyltransferase in a total volume of 25 μL. All assays were carried out in triplicate. Acceptors were lactose (2.6 mM) or N-acetyllactosamine (LacNAc, 2.6 mM). Routinely we used 0.2 mM CMP-Neu5Ac donor for comparative assays and 1 mM CMP-Neu5Ac (with 8 mM lactose) when determining accurate specific activities (4× K_m concentrations). The sialyltransferase was 100 ng commercial of ST6Gal I (Sino Biologicals (mouse), Wayne, PA, USA) or varying amounts of crude and purified sialyltransferase preparations. Negative control was purified prostate specific antigen (PSA) produced in parallel, originally from the pSecTag/FRT/V5-His-TOPO Vector Kit. Reactions were incubated for 20 or 30 min at 37 °C. For specific activity determination a range of enzyme amounts were assayed while keeping both donor and acceptor levels at non-limiting levels.

Purification

Bovine serum albumin present in FBS was a contaminant in early purification protocols therefore Flp-In-CHO cells were weaned onto ProCHO-AT media. Cell-free media (50 mL) in 1× binding buffer (50 mM Tris–HCl, pH 7.5, 500 mM NaCl) and 1 mL Ni Sepharose Excel resin (GE Healthcare, Chicago, IL, USA) were incubated overnight on a roller at 4 °C. The resin was collected after centrifugation (4,000g, 15 min, 4 °C) and packed into columns. Columns were washed with 100 column volumes (CV) binding buffer containing 20 mM imidazole (optimized for protein yield and sialyltransferase activity across a 0–30 mM imidazole range). Binding proteins were eluted with two CV 50 mM Tris–HCl, pH 7.5, 500 mM NaCl containing 500 mM imidazole. Buffer exchange with TBS (20 mM Tris–HCl, pH 7.2, 100 mM NaCl) was carried out using Amicon® 3 kDa spin filters (EMD Millipore, Burlington, MA, USA) by centrifugation at 4,000g for 30 min at 4 °C (5 times). The retentate (approximately 500 μL) was concentrated using a 30 kDa spin filter to approximately 50 μL. Protein concentrations were measured spectrophotometrically by NanoDrop and calibrated against BSA protein standards using the bicinchoninic acid (BCA) protein assay (Pierce, Appleton, WI, USA). Proteins were analyzed in reducing conditions on 12% NuPAGE Novex Bis–Tris gels using the NuPAGE MOPS SDS Buffer Kit (Thermo Fisher Scientific, Waltham, MA, USA). The predicted molecular weights of all secreted sialyltransferase fusion proteins are given in Table S3. Gels were Coomassie or silver stained. A wet transfer protocol to Polyvinylidene difluoride (PVDF) membrane (1 h at 24 V) was used (Invitrogen, Carlsbad, CA, USA). Blocking was with 4% dried skimmed milk in TBS-T (TBS + 0.05 % Tween-20) overnight at 4 °C. Detection of His-tagged proteins used either a two-step protocol with mouse anti-His antibody (Novagen, EMD Millipore, Burlington, MA, USA) (1:1,000 dilution) and a HRP-conjugated goat anti-mouse IgG (1:1,000 dilution), or a one-step protocol using mouse anti-His-HRP (1:1000 dilution) (Sigma-Aldrich Merck, Gillingham, UK). BenchMark His-tagged Protein Standard and PageRuler Prestained Protein Ladder (10–180 kDa) were loaded as molecular weight markers (Thermo Fisher Scientific, Waltham, MA, USA). Detection was either with 3,3′-Diaminobenzidine (DAB) or with Enhanced Chemiluminescence (ECL) (Advansta, Menlo Park, CA, USA). Where applied Ni Sepharose-purified protein was digested with PNGase F (New England Biolabs, Ipswich, MA, USA) for 1 h at 37 °C as recommended by the manufacturer.

HILIC-HPLC

Glycans were labelled with 2-amino benzamide (2-AB) (Bigge et al., 1995) and separated by hydrophilic interaction liquid chromatography (HILIC) using a GlycoSep N-Plus HPLC column (Prozyme, Ballerup, Denmark) on a Waters Alliance 2695 HPLC system with Waters 2475 fluorescence detection. Flow rate was 0.67 mL/min with 20% 50 mM ammonium formate, pH 4.4 (solvent A) and 80% acetonitrile (Romil, Cambridge, UK). Samples were injected in 80% acetonitrile with a linear gradient of 20–58% solvent A over 48 min. Sialyltransferase reactions (50 μL scale) were carried out much as described above using excess of donor (CMP-Neu5Ac) and acceptor (LacNAc, Lac or whey permeate) at approximately 4× K_m concentrations. Whey permeate (provided by Glanbia plc) was 79.6% lactose by our estimate. Approximately 900 ng of all SIATs (controls and experimental samples) were assayed. Duration (4 h or 16 h) and temperature (37 or 20 °C) of incubations were varied though 16 h incubation became routine. After reaction 450 μL ice-cold HPLC-grade water was added to the reaction, followed by centrifugation (10 min, 20,000g, 4 °C) through Amicon® spin filters. Eluants were concentrated to dryness ready for 2-AB labelling. Resulting glycans were fluorescently labelled with 2-AB by reductive amination (Bigge et al., 1995). In brief, an appropriate volume of 0.35M 2-AB in Dimethyl sulfoxide (DMSO) containing 30% acetic acid was dissolved in sodium cyanoborohydride (1M) at 65 °C. A volume of 60 μL of the preparation was used to label dried glycans. Samples were incubated for 2.5 h at 65 °C. A GlycoClean S cartridge (Prozyme, Ballerup, Denmark) was used to separate labelled glycans from unreacted 2-AB following the manufacturer’s instructions. Elution was carried out with 3 × 0.5 mL water and 2-AB labelled glycans were evaporated to dryness using a concentrator. In this HILIC system each sugar has a retention time, stated in glucose units (GU) that is unique for that molecule. Calibration of retention with dextran (20 μg, dextran ladder from Leuconostoc mesenteroides) (Sigma-Aldrich Merck, Gillingham, UK) was carried out three times over the course of a run to insure accurate GU values. Fifth polynomial equation resolution GraphPad Prism (GraphPad Software, San Diego, CA, USA) was used to convert retention times into GU (Marino et al., 2011). Calibration of quantities was determined for a 10-fold dilution (1 μg–0.1 ng) of lactose, 3′ sialyllactose and 6′ sialyllactose using peak height and/or integrated area. Standards were LacNAc (CarboSynth, Compton, UK), Lac (Sigma-Aldrich Merck, Gillingham, UK), 3′ sialyllactose (Sigma-Aldrich Merck, Gillingham, UK), 6′ sialyllactose (Glycom, Esbjerg, Denmark), and Neu5Ac (Sigma-Aldrich Merck, Gillingham, UK).

Enzyme-linked lectin assay

Enzyme-linked lectin assay (ELLA) was based on previous protocols (Couzens et al., 2014; McCoy, Varani & Goldstein, 1983; Thompson et al., 2011). Lectin binding is generally quite weak and requires multivalent binding sites acting together to give strong and specific binding. For this assay, we therefore used asialofetuin (ASF) as the acceptor in preference to the disaccharides Lac or LacNAc. Fetuin is a simple glycoprotein (Fig. S2) with a defined structure offering six potential terminal LacNAc sites and is readily available in its asialylated form. ASF (50 μL per well of a 0.5 μg/mL solution in PBS) was added to Nunc™ MaxiSorp polystyrene 96 well ELISA plates (Thermo Fisher Scientific, Carlsbad, CA, USA) and incubated at 4 °C overnight. Wells were blocked for 2 h at room temperature with 150 μL 0.5% (w/v) polyvinyl alcohol. Wells were washed (four times) with 100 μL mTBS-T (mTBS containing 0.05% Tween 20). Sialyltransferase reactions (50 μL) were prepared containing 50 mM MES pH 6.5, 10 mM MnCl₂, 0.2 mM CMP-Neu5Ac, 2.5 units Antarctic phosphatase (New England Biolabs, Ipswich, MA, USA) and the sialyltransferase preparation to be assayed. We used 100, 200, and 400 ng of commercial mouse ST6Gal I as positive control, 400 ng of our test sialyltransferase or a negative control with no sialyltransferase. Reactions was added to wells and incubated at 37 °C for 16 h. Wells were washed (four times) with 100 μL mTBS-T and biotinylated lectins (Vector Laboratories, Peterborough, UK) were added at 2, 5, or 10 μg/mL of SNA-I, RCA-I or MAA, respectively. After a further 1 h incubation at room temperature wells were again washed (four times) with 100 μL mTBS-T. HRP-Streptavidin (R&D Systems, Abingdon, UK) was added at 1/200 dilution and incubated 1 h at room temperature. Wells were washed (four times) with 100 μL mTBS-T. 3,3′,5,5′-tetramethylbenzidine (TMB) substrate (50 μL) was added followed by incubation at room temperature for 15 min with the reaction stopped by adding 50 μL 2M sulfuric acid. Absorbance was measured at 450 nm on a SpectraMaxR® spectrophotometer (Molecular Devices, San Jose, CA, USA).

Statistics

Student t-tests were used to compare SIAT activities between recombinant PSA produced by the control Flp-In-CHO cell line and all other recombinant SIATs purified from stable cell lines. Mann–Whitney tests were carried out to compare the lectin binding of ASF with all recombinant SIATs. Tests were performed with PASW Statistics 22 (IBM SPSS Statistics, Armonk, NY, USA).

Constructs

All known SIATs (CAZy-family 29) have an approximate 250 amino acid catalytic domain that is naturally orientated to the lumen of the Golgi. The full length protein has a transmembrane domain that attaches the catalytic domain to the Golgi membrane through a stem structure that is not essential for catalytic activity. Our focus in this study was on expressing the catalytic domains of the SIATs, however, for comparative purposes we also constructed two full-length ST6Gal I expressing vectors (see Materials and Methods). All catalytic domain-expressing constructs were based on pSecTag/FRT/V5-His TOPO with inserts amplified by PCR from cDNA. In order to make all the cloned proteins as comparable as possible we aligned all the protein sequences and maintained a consistent length peptide linker (part of the stem) between the Igκ secretion signal sequence and the known functional limits of the catalytic domain. ST6Gal I and ST6Gal II therefore started 121 amino acids N-terminal to a Cys that is totally conserved across all animal SIATs (Cys-184 in human ST6Gal I). The first amino acid in the human ST6Gal I was Val-63. For ST3Gal IV we chose to define the catalytic domain as starting approximately 70 amino acids N-terminal to this same invariant Cys. All ST6Gal I, ST6Gal II, and ST3Gal IV expressing constructs are summarized in Table 1 and Fig. 1, with additional details in Tables S1–S3. Some minor differences that predated our final routine cloning approach are explained in the Materials and Methods section.

Sequences

We cloned and developed stable cell lines of three separate human ST6Gal I sequences. One was the wild-type sequence (NP_003023.1). Clone hST6 B1 had an interesting G327S mutation in a highly conserved region in the S-motif: this Gly is highly conserved across ST6Gal I and ST6Gal II proteins in all species (Harduin-Lepers, 2010; Harduin-Lepers et al., 2005). Clone hST6 B1 subsequently refers to the codon-optimized and synthesized version of this clone. hST6 A5 had two mutations (I295T and K395R) in less highly conserved regions. The zebrafish clone zST6 had both an H111I and a K187R mutation in comparison to the database sequence (NP_001003853.1). Rat had a R105G mutation. All other ST6Gal I protein sequences matched the database sequences (chicken, fugu, and stickleback). Human ST6Gal II (sourced from the RPMI cell line) had an A269T missense mutation. Human ST3Gal IV clones (hST3 A1 and hST3 A2) were different by a six amino acid insertion present in hST3 A1 but were otherwise identical to the database protein sequences. Some of these variant sequences may have been inadvertently introduced in the amplification process (though proof-reading Taq polymerases were used throughout for cloning) or be common variants in laboratory species, but whatever their origins they were useful for the purposes of this study which was to generate a library of enzymes with potentially different physiochemical properties.

Clones were transfected into Flp-In-CHO cells and selected for with hygromycin. Transfected cDNA should recombine at the corresponding FRT sites engineered into the CHO chromosome of these Flp-In cells, therefore no further selection was carried out to identify highly expressing clones and transfected cells were pooled. The transfected cells were investigated initially for lectin binding. CHO cells do not express ST6Gal I therefore there are minimal NeuAc(α2-6)Gal terminal structures on surface glycans (Jenkins, Parekh & James, 1996; Lee, Roth & Paulson, 1989; Monaco et al., 1996). The Sambucus nigra SNA-I lectin has specificity for NeuAc(α2-6)Gal/GalNAc therefore this lectin can be used to identify terminal NeuAc(α2-6) structures present in CHO cells as a result of functional expression of the transfected cDNA (Smith, Song & Cummings, 2010). The Maackia amurensis MAA lectin (undefined mixture of MAA-1 and MAA-2) on the other hand has specificity for NeuAc(α2-3)Gal/GalNAc (Geisler & Jarvis, 2011) therefore this lectin can be used to identify terminal NeuAc(α2-3) which is naturally produced in CHO cells by the activity of ST3Gal enzymes. We have used these two lectins to show that NeuAc(α2-6) structures are much more prominent in transfected CHO cells (ST6GAL1-transfected) and that there is little changes in the NeuAc(α2-3) structures (Fig. 2).

SIAT assays by phosphate linked assay

Sialyltransferase activity was determined indirectly through a phosphate linked assay. LacNAc is considered the best acceptor for human ST6Gal I in this type of assay. Suitable potential acceptors were LacNAc, ASF, or Lac. As production of sialyllactose was the focus of our study, Lac or LacNAc (as disaccharides) were preferred over ASF (a glycoprotein). Samples assayed were predominantly His-purified proteins isolated from the media of stably transfected cells, though initially cell lysates were used instead of secreted protein. We determined significant sialyltransferase activity for purified ST6Gal I proteins of all species which were comparable to the commercial control (mouse ST6Gal I, mST6 in Fig. 3). Both mutant human ST6Gal I proteins (hST6 B1 and hST6 A5) were active. Of the ST3Gal IV preparations, only zebrafish ST3Gal IV (zST3) showed significant activity. The two human ST3Gal IV clones (hST3 A1, hST3 A2) and the human ST6Gal II clone showed marginal activity in this assay. The results may be influenced by the choice of acceptor: it is possible that these constructs, showing marginal activity with LacNAc, would show stronger activity with ASF as acceptor. For several enzymes we also measured specific activity across a range of protein amounts. In a direct comparison of activity in cellular lysates human ST6Gal I clones hST6 A5, hST6 B1, and zebrafish clone zST6 gave activities of 0.6 pmol/min/μg, 7.1 pmol/min/μg, and 6.2 pmol/min/μg, respectively. In these samples activity may be influenced by the efficiency of secretion with a poor secretor giving a proportionally higher level of protein in the lysates.

Protein purification

Proteins were purified as described in Materials and Methods. Because of purification issues around bovine serum albumin present in FBS, cells were weaned onto a serum-free media (ProCHO-AT). A consequence of the switch to ProCHO-AT was that the CHO cells were not fully adherent to culture plates. This was irrespective of whether the Flp-In-CHO cells were transfected or not. Although we did not determine whether the two phenotypes (adherent and non-adherent) differed in their recombinant sialyltransferase expression characteristics, we did show that the phenotype was not fixed—non-adherent cells returned to a mix of adherent and non-adherent cells when reseeded.

Typical protein yields from 200 mL media ranged from 1 to 33 μg across clones. This corresponded to a typical recovery of 2–5% though estimates of protein concentration in crude media were inconsistent. For comparative purposes sialyltransferase assays were carried out with a fixed volume of recombinant enzyme and corrected for protein concentration. Specific activities for select preparations of cell lysates, media or IMAC-purified proteins were determined across a range of protein amounts. Specific activities of 1,697 pmol/min/μg (for commercial mouse ST6Gal I) and 3,147 pmol/min/μg (for human ST6Gal I, hST6 A5, His-purified from a cellular lysate) corresponded to activity measurements for both of approximately 450 pmol/min/μg in the comparative assay (fixed protein) of Fig. 3. Purified proteins were silver-stained on SDS/polyacrylamide gels and were analyzed on Western blots using anti-His antibodies (Fig. 4A). Typically bands on gels were weak and varied in size depending on the species. Glycosylation (heterogeneous) also led to diffuse bands for several proteins and the suggestion of different prominent glycoforms. PNGase-F digestion of purified human ST6Gal I showed a shift downward in molecular mass from 55 kDa (for a prominent glycoform) to 41 kDa (Fig. 4B). There were two prominent glycoforms in this preparation of 55 and 43 kDa, with two weaker glycoforms of 60 and 50 kDa.

HILIC

Indirect evidence from the phosphate linked assay suggested the presence of sialyltransferase activity. Direct evidence was obtained by incubating the enzyme with lactose and detecting the expected products (3′ sialyllactoses for ST3Gal IV and 6′ sialyllactose for ST6Gal I). The products were labelled with 2-AB and separated by HILIC. Each sugar has a retention time, which when stated in GU calibrated against a dextran ladder, is unique for that molecule. Sialyllactose standards generated distinctive GU values of 2.88 and 3.35 for 3′ and 6′ sialyllactose, respectively.

Reactions were carried out routinely with 8 mM lactose (or whey permeate equivalent) and 2 mM CMP-Neu5Ac which approximate to 4× K_m concentrations (for human ST6Gal I), ensuring that both acceptor and donor were in excess. Yields were checked at 1, 4, and 16 h at 37 °C but were subsequently set at 16 h. Both 3′ and 6′ sialyllactoses peaks were evident at the expected positions depending on the purified enzyme used, ST3Gal IV or ST6Gal I, respectively (Fig. 5). The human ST6Gal I clone (hST6 B1) yielded 11.5 ng 6′ sialyllactose considerably higher than the 3.9 ng of 6′ sialyllactose achieved with ST6Gal II. Chicken and stickleback ST6Gal I were also shown to yield 6′ sialyllactose. The efficiencies of these reactions were generally low, never achieving more than a 6% conversion of lactose to sialyllactose, though better yields could have been achieved with more optimization.

Enzyme-linked lectin assays

At the cellular level (Fig. 2) we demonstrated increased SNA-I labelling of CHO cells expressing ST6Gal I. At the molecular level and in vitro this can be demonstrated by the de novo sialylation of ASF. ASF in a microplate assay should show limited binding to SNA-I but after reaction with purified ST6Gal enzymes there should be increased labelling by SNA-I. There should also be reduced binding to the exposed Gal groups in ASF now hidden as a result of sialylation. Gal groups (Galα1-4GlcNAc and other Galα1-related oligosaccharides) typically bind the RCA I lectin. In two separate experiments we observed clear binding of SNA-I by ELLA after reacting ASF with the different purified SIATs (Fig. 6A). Chicken and rat ST6Gal I were as effective as the commercial (mouse) ST6Gal I for adding on SNA-I recognized sugars (probably Neu5Ac(α2-6) linked to Gal. The other ST6Gal I preparations were less effective and the ST6Gal II preparation showed no activity in this assay. As expected the ST3Gal IV preparations did not affect SNA-I binding as Neu5Ac(α2-3) linked to Gal is not an epitope for SNA-I. Reduced binding of RCA I was also observed for the strongest ST6Gal I preparations (Fig. 6B). The most effective enzyme in this assay was chicken ST6Gal I with approximately the same activity as the commercial mouse ST6Gal I.