Ammonium alters N-glycan structures of recombinant TNFR-IgG: degradative versus biosynthetic mechanisms.

The effect of ammonium on the glycosylation pattern of the recombinant immunoadhesin tumor necrosis factor-IgG (TNFR-IgG) produced by Chinese hamster ovary cells is elucidated in this study. TNFR-IgG is a chimeric IgG fusion protein bearing one N-linked glycosylation site in the Fc region and three complex-type N-glycans in the TNF-receptor portion of each monomer. The ammonium concentration of batch suspension cultures was adjusted with glutamine and/or NH(4)Cl. The amount of galactose (Gal) and N-acetylneuraminic acid (NANA) residues on TNFR-IgG correlated in a dose-dependent manner with the ammonium concentration under which the N-linked oligosaccharides were synthesized. As ammonium increased from 1 to 15 mM, a concomitant decrease of up to 40% was observed in terminal galactosylation and sialylation of the molecule. Cell culture supernatants contained measurable beta-galactosidase and sialidase activity, which increased throughout the culture. The beta-galactosidase, but not the sialidase, level was proportional to the ammonium concentration. No loss of N-glycans was observed in incubation studies using beta-galactosidase and sialidase containing cell culture supernatants, suggesting that the ammonium effect was biosynthetic and not degradative. Several biosynthetic mechanisms were investigated. Ammonium (a weak base) is known to affect the pH of acidic intracellular compartments (e.g., trans-Golgi) as well as intracellular nucleotide sugar pools (increases UDP-N-acetylglucosamine and UDP-N-acetylgalactosamine). Ammonium might also affect the expression rates of beta1, 4-galactosyltransferase (beta1,4-GT) and alpha2,3-sialyltransferase (alpha2,3-ST). To separate these mechanisms, experiments were designed using chloroquine (changes intracellular pH) and glucosamine (increases UDP-GNAc pool [sum of UDP-GlcNAc and UDP-GalNAc]). The ammonium effect on TNFR-IgG oligosaccharide structures could be mimicked only by chloroquine, another weak base. No differences in N-glycosylation were found in the product synthesized in the presence of glucosamine. No differences in beta1, 4-galactosyltransferase (beta1,4-GT) and alpha2,3-sialyltransferase (alpha2,3-ST) messenger RNA (mRNA) and enzyme levels were observed in cells cultivated in the presence or absence of 13 mM NH(4)Cl. pH titration of endogenous CHO alpha2,3-ST and beta-1,4-GT revealed a sharp optimum at pH 6.5, the reported trans-Golgi pH. Thus, at pH 7.0 to 7.2, a likely trans-Golgi pH range in the presence of 10 to 15 mM ammonium, activities for both enzymes are reduced to 50% to 60%. Consequently, ammonium seems to alter the carbohydrate biosynthesis of TNFR-IgG by a pH-mediated effect on glycosyltransferase activity.

Incorporation of 15N from ammonium into the N-linked oligosaccharides of an immunoadhesin glycoprotein expressed in Chinese hamster ovary cells.

Elevated ammonium concentrations in the medium of cultivated cells have been shown to increase the intracellular levels of uridine-5'-diphospho-N-acetylglucosamine (UDP-GlcNAc) and uridine-5'-diphospho-N-acetylgalactosamine (UDP-GalNAc; Ryll et al., 1994). These sugar nucleotides are substrates for glycosyltransferases in the glycosylation pathway. In our experiments, recombinant Chinese hamster ovary cells producing an immunoadhesin glycoprotein (GP1-IgG) have been cultivated under controlled cell culture conditions in the presence of different ammonium concentrations.15N-Labeled ammonium chloride (15NH4Cl) was added exogenously to the cell culture media to determine if ammonium was incorporated into UDP-GlcNAc and cytidine-5'-monophospho-N-acetylneuraminic acid (CMP-NANA) pools, and subsequently incorporated into GP1-IgG as N-linked glycans. The intracellular pools of UDP-activated hexosamines (UDP-GNAc) were followed during the time course of the experiment. To assess the extent of15NH4+incorporation into the glycans of GP1-IgG, the glycoprotein was first purified to homogeneity by protein A chromatography. Enzymatically released N-glycans were then analyzed by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. N-Glycans synthesized in the presence of15NH4Cl revealed an N-glycan-dependent increase in mass-to-charge of 2.5-4.8 Da. These results indicate that 60-70% of the total nitrogen containing monosaccharides had incorporated15N. Presumably,15NH4+was incorporated into GlcNAc and N-acetylneuraminic acid as proposed earlier (Ryll et al., 1994). This might be a universal and previously not described reaction in mammalian cells when exposed to nonphysiological but in cell culture commonly found concentrations of ammonium. The data presented here are of significance for glycoprotein production in mammalian cell culture, since it has been shown previously that elevated levels of UDP-activated hexosamines affect N-glycan characteristics such as branching and degree of amino sugar incorporation. In addition, our results demonstrate that isotope labeling in combination with MALDI-TOF-MS can be used as an alternate tool to radioactive labeling of sugar substrates in metabolic studies.

Removal of sialic acid from a glycoprotein in CHO cell culture supernatant by action of an extracellular CHO cell sialidase.

We have directly tested the hypothesis that Chinese hamster ovary (CHO) cell-produced glycoproteins are subject to extracellular degradation by a sialidase endogenous to the CHO cell line. Factors important to understanding the potential for extracellular degradation are addressed including the glycoprotein specificity, subcellular source, mechanism of release, and stability of the sialidase activity. The extracellular CHO cell sialidase apparently originates from the cytosol of the cells, and is released to the cell culture supernatant as a result of damage to the cellular membrane. The extracellular sialidase is active toward a variety of CHO cell-produced glycoproteins, and can hydrolyze sialic acid from the recombinant glycoprotein gp120 in the culture supernatant. While measuring the actual degradation of a glycoprotein by extracellular CHO cell sialidase can be difficult, data presented here suggest that the level of degradation can be estimated indirectly by using a more convenient fluorescent substrate, 4-methylumbelliferyl-alpha-D-N-acetylneuraminic acid, to quantify sialidase activity. Degradation by sialidase is minimized through addition of the sialidase inhibitor 2,3-dehydro-2-deoxy-N-acetylneuraminic acid to the culture supernatant. The results in this study suggest additional potential approaches for minimizing degradation by sialidase, including isolation of a sialidase-deficient CHO cell line.

Purification and characterization of alpha-L-fucosidase from Chinese hamster ovary cell culture supernatant.

In this study, alpha-L-fucosidase from Chinese hamster ovary (CHO) cell culture supernatant was purified 11 200-fold to apparent homogeneity to assess the rate of fucose hydrolysis from oligosaccharide and glycoprotein substrates. The fucosidase migrated as a single band of 51 kDa on SDS-PAGE and is a glycoprotein, as determined by retention on concanavalin A-Sepharose, and by lectin blotting with concanavalin A. Hydrolysis of the artificial substrate 4-methyl-umbelliferyl-alpha-L-fucoside (4MU-Fuc) followed simple Michaelis-Menten kinetics, and was competitively inhibited by free fucose and by two known fucosidase inhibitors, fucosylamine and deoxyfuconojirimycin. Hydrolysis of fucose from oligosaccharides including 2'-fucosyllactose, 3-fucosyllactose, Fuc alpha(1,6)GlcNAc and pooled gp120 oligosaccharides with the Fuc alpha(1,6)GlcNAc linkage also followed simple Michaelis-Menten kinetics. However, activity toward 4MU-Fuc was optimal near pH 7, while activities toward the oligosaccharide substrates were optimal near pH 5. No fucose was released from the recombinant CHO cell-produced glycoproteins gp120 or soluble CD4 with the Fuc alpha(1,6)GlcNAc linkage, or from human serum alpha 1-acid glycoprotein with the Fuc alpha(1,3)GlcNAc linkage. Enzymatic removal of sialic acid and galactose from gp120 oligosaccharides did not alter the susceptibility of gp120 to fucosidase attack. These data suggest that released CHO cell fucosidase does not contribute to the heterogeneity of fucosylation that has been observed in CHO cell culture-produced glycoproteins.

Assessing genetic heterogeneity in production cell lines: detection by peptide mapping of a low level Tyr to Gln sequence variant in a recombinant antibody.

A recombinant antibody directed against the human epidermal growth factor receptor-2 extracellular domain was subjected to detailed structural characterization. Heterogeneity in the heavy chain was demonstrated by recovery of two forms of a tryptic peptide, with either glutamine or the expected tyrosine at residue 376. Subsequent experiments indicated that the Y376Q variant developed during transfection of the antibody heavy and light chain genes into Chinese hamster ovary cells. Levels of the Y376Q variant (range: 27% to 1%) in the purified antibody were inversely proportional to cell age. The established cell line was subcloned and found to be heterogeneous by polymerase chain reaction analysis of cell extracts and protein analysis of the purified antibody. Ten percent of subclones produced high levels of the Y376Q variant while 90% of the subclones produced antibody with only the expected heavy chain sequence. This report demonstrates the utility of peptide mapping as a sensitive tool for assessing genetic heterogeneity of recombinant cell lines.

Isolation and properties of a soluble sialidase from the culture fluid of Chinese hamster ovary cells.

A soluble sialidase that can degrade recombinant glycoproteins expressed in Chinese hamster ovary (CHO) cells has been isolated and purified to near homogeneity from the cell culture fluid of this host. Purification of approximately 34,000-fold was carried out using conventional purification techniques including sequential DEAE-Sepharose and S-Sepharose ion-exchange chromatography, followed by hydrophobic interaction chromatography with Phenyl-Toyopearl. Final purification was achieved by heparin-agarose and chromatofocusing chromatography. The minimum molecular weight of the sialidase on SDS-PAGE was approximately 43,000 Da. When the final preparation was examined under non-denaturing conditions, two major (pI = 6.8 and 7.0) and five minor electrophoretic forms with different isoelectric points were identified. The basis for the electrophoretic heterogeneity is not known, but it was not due to carbohydrate diversity since no carbohydrates were detected on the purified protein. The enzyme degraded a variety of sialyl-conjugate substrates, at a pH optimum of 5.9, including intact glycoproteins, oligosaccharides and gangliosides with a 4-fold preference for 2,3- versus 2,6-linked sialic acid residues. With ganglioside substrates, internally linked sialic acid residues were not cleaved by the enzyme. Delineation of this enzyme from the lysosomal and plasma membrane sialidases was made using inhibition studies with C-9 substituted 5-acetamido-2,6-anhydro-3,5-dideoxy-D-glycero-D-galacto-non-2- enonic acid derivatives. The enzyme was identified in several CHO cell lines by immunoblotting using antiserum raised against a synthetic peptide based on amino acid sequence of a fragment derived by trypsin digestion of the purified sialidase.(ABSTRACT TRUNCATED AT 250 WORDS)

Protein sorting by high-performance liquid chromatography. I. Biomimetic interaction chromatography of recombinant human deoxyribonuclease I on polyionic stationary phases.

Chromatographic separations can be tailored to exploit specific interactions between a stationary phase ligand and a protein structural feature of interest. Variations in this feature then form the basis for sorting a mixture of closely related proteins into defined subpopulations. This report describes the sorting of variants of recombinant human deoxyribonuclease I (rhDNase) that differ in the occurrence of deamidation at a single residue. rhDNase, an enzyme that non-specifically hydrolyzes DNA, is glycosylated and exhibits considerable charge heterogeneity owing to the sialylation and phosphorylation of its N-linked oligosaccharides. This heterogeneity obscures the relatively subtle differences between deamidated and intact rhDNase, preventing separation on this basis in conventional ion-exchange HPLC. Published structural information on bovine DNase reveals that the analogous labile asparagine residue is involved in DNA binding, so stationary phases containing polyanionic ligands mimicking nucleic acids were employed to separate the deamidation variants of rhDNase. Electrostatically immobilized DNA, a "tentacle" cation exchanger (TCX) and immobilized heparin columns all resolved the deamidated and intact forms of rhDNase when operated at pH 4.5. The ligands of the TCX and heparin columns are sufficiently long, flexible and polyanionic to interact with rhDNase in a manner similar to DNA and to sort rhDNase variants on the basis of the charge difference of a single residue involved in that interaction. A non-hydrolyzable double-stranded oligonucleotide analogue of DNA was also synthesized and immobilized to an HPLC support. This column, operated at pH 6, where rhDNase is active, resolved the two isomeric products of deamidation of rhDNase, i.e., variants of the enzyme containing either aspartate or isoaspartate in lieu of asparagine at the deamidation site in rhDNase. This is the first reported separation of intact variants of a glycoprotein differing on the basis of these isomeric products of deamidation through the common cyclic imide mechanism.

Biochemical characterization of recombinant human nerve growth factor.

Recombinant human nerve growth factor (rhNGF) was expressed and secreted by Chinese hamster ovary cells and purified to homogeneity using ion-exchange and reversed-phase (RP) chromatography. The isolated product was shown to be consistent with a 120-amino-acid residue polypeptide chain by amino acid composition, sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), RP-HPLC, and mass spectrometry and with an N-terminal sequence consistent with that expected from the cDNA for human nerve growth factor. By size-exclusion chromatography, rhNGF behaves like a noncovalent dimer. Limited enzymatic digests of the 120-residue monomer produced additional species of 118 (trypsin, removal of the C-terminal Arg119-Ala120 sequence) and 117 (trypsin plus carboxypeptidase B, removal of the C-terminal Arg118-Arg119-Ala120 sequence) residues. Each of these species was isolated by high-performance ion-exchange chromatography and characterized by amino acid and N-terminal sequence analyses, SDS-PAGE, RP-HPLC, and mass spectrometry. All three species were present in the digests as both homodimeric and heterodimeric combinations and found to be equipotent in both the chick dorsal root ganglion cell survival and rat pheochromocytoma neurite extension assays.

Assay requirements for cell culture process development.

Because the structural parameters which influence product "quality" will vary from protein to protein, the screening methods used during development of each production process will also differ. Thus, not all of the methods described above will be needed for each product. One of the challenges of cell culture process development is choosing the correct structural features to monitor. The next level of process development may involve on-line control of many of the parameters discussed above. This would allow product quality-driven manipulations of culture conditions. The major factor limiting this advance may not be the techniques for real-time analysis, but rather a more thorough understanding of the structural features which describe a high-quality pharmaceutical protein.

Rat liver and small intestine produce proapolipoprotein A-I which is slowly processed to apolipoprotein A-I in the circulation.

Two-dimensional electrophoretic analysis of plasma lipoproteins from male Osborne-Mendel rats consistently reveals three isoforms of apolipoprotein A-I (apo-A-I) with the following apparent pI values and quantitative distribution: isoform 3, pI = 5.68, 69%; isoform 4, pI = 5.55, 29%; isoform 5, pI = 5.44, 2%. The two major isoforms were obtained by preparative isoelectric focusing and subjected to NH2-terminal amino acid sequence analysis with the following results: isoform 3, (Asp)-Glu-Pro-Gln-Ser-Gln-Trp-Asp-Arg-Val; isoform 4, X-Glu-Phe-X-Gln-Gln-Asp-Glu-Pro-Gln-Ser. By comparison with the amino acid sequence previously reported for the primary translation product of rat intestinal apo-A-I mRNA (Gordon et al. (1982) J. Biol. Chem. 257, 971-978), isoform 3, the more basic isoform, is identified as mature apo-A-I and isoform 4 as its proform ( proapo -A-I). The proform differs from mature apo-A-I by a 6-amino acid extension at the NH2 terminus. Isoform 5 was not identified further. The plasma steady state distribution of the apo-A-I forms indicates that proapo -A-I is relatively stable in the circulation. Virtually all plasma proapo -A-I is lipoprotein-associated. No significant differences in the steady state proportions of plasma apo-A-I forms were observed between male and female rats, or among various subfractions of plasma high density lipoproteins obtained by heparin-Sepharose affinity chromatography or by density gradient ultracentrifugation. Rats fed a high fat, high cholesterol diet, however, showed an increase in the proportion of circulating proapo -A-I. The relative increase in proform was even more pronounced in rats fed a fat-free diet containing orotic acid. The biosynthesis, secretion, and metabolism of the various apo-A-I forms were also studied. In liver and intestine, the only known sites of apo-A-I synthesis in the rat, approximately 85% of the newly synthesized intracellular apo-A-I, was the proform . Proapo -A-I was also the predominant form (approximately 80%) released into the circulation by isolated, perfused livers and by autoperfused intestinal segments in vivo. Gradual processing of circulating proapo -A-I to mature apo-A-I was observed in vivo following pulse-labeling of apo-A-I with [3H]leucine. Processing in vivo was approximately 80% complete in 10 h.(ABSTRACT TRUNCATED AT 400 WORDS)

Resolution of sulphydryl oxidase from gamma-glutamyltransferase in bovine milk by covalent chromatography on cysteinylsuccinamidopropyl-glass.

1. Sulphydryl oxidase from bovine milk was purified by covalent affinity chromatography on cysteinylsuccinamidopropyl-glass. Selective immobilization of the oxidase occurs through formation of a mixed disulphide between the enzyme and the substrate cysteinyl-glass matrix. Reductive elution of the bound protein can be accomplished with small thiols such as reduced glutathione (GSH), dithiothreitol or cysteine. This method leads to approx. 4000-fold purification of the enzyme from whey. Furthermore, complete resolution of sulphydryl oxidase from gamma-glutamyltransferase was achieved with this procedure. 2. Antibodies prepared against this purified enzyme quantitatively precipitated 95% of the GSH-oxidative activity from detergent-solubilized skim-milk membranes, whereas 100% of the transferase activity remained in the supernatant fraction; these findings confirmed the distinction between these two enzymes. 3. Reverse-phase high-pressure-liquid-chromatographic analyses of assay mixtures containing both enzymes revealed an array of GSH derivatives generated by a combination of the oxidative and hydrolytic activities. However, purified sulphydryl oxidase yielded only GSSG with concomitant stoichiometric loss of GSH. 4. The chromatographic method described is simple and reproducible, and may be applicable to isolation of sulphydryl oxidase from other tissues.