Isoform number I--a new tool to evaluate the quality of erythropoietin.

The overall isoform number I, calculated from capillary zone electrophoresis (CZE) data, is introduced as a new parameter that enables the assessment of the quality of erythropoietin (EPO) in a simple, yet efficient manner. I is defined as the sum of the products of the individual CZE peak area percent shares (pn) of the EPO isoforms (n = 1-8) and the corresponding individual isoform numbers (n): I=p1x1+p2x2+p3x3+ p4x4+p5x5+p6x6+p7x7+p8x8 The results from an internal collaborative study were used to calculate I-numbers for 2 successive batches of EPO Biological Reference Preparations (BRPs) established by the European Pharmacopoeia (Ph. Eur.) Commission. Based on the results of 6 participating laboratories each, the I-numbers calculated for batch 1 (BRP1) and candidate batch 2 (cBRP2) were I = 518.7 +/- 3.5 (coefficient of variation (CV) = 0.7 %) and I = 542.4 +/- 2.2 (CV = 0.4 %) respectively. Thus, the I-numbers of the 2 EPO preparations clearly indicated the difference between BRP1 and cBRP2 described in the literature, but in a much more apparent and simple manner. Notably, one of the 7 laboratories participating in the study with cBRP2 yielded an outlier value of I = 525.6, pointing towards a suboptimal CZE analysis, and was therefore excluded from the cBRP2 mean value calculation. It is suggested that the overall isoform number I of erythropoietin provides a new and highly valid quality control measure that makes it possible to ensure the batch-to-batch consistency of the active pharmaceutical ingredient of EPO market products.

N-glycan charge assay--an alternative for potency assays of therapeutic glycoproteins?

A new in vitro potency assay for erythropoietin has been developed and compared to the exhypoxic polycythemic mouse bioassay for rhuEPO batch release, required according to Ph. Eur. 2001. Evaluation of 14 batches of EPO has shown that the new assay is highly sensitive and able to reveal subtle changes in N-linked carbohydrates. In combination with regular EPO batch release assays (such as sialic acid determination, isoelectric focusing, RP-HPLC, and peptide mapping), this assay may enable to replace the highly variable exhypoxic mouse bioassay for EPO batch release and stability studies and thus reduce the use of laboratory mice.

The patterns of the complex- and oligomannose-type glycans of uromodulin (Tamm-Horsfall glycoprotein) in the course of pregnancy.

Uromodulin was isolated from urine of three pregnant women. Urine of each donor was collected at subsequent stages of their pregnancy and at one month after gestation. Each batch of uromodulin was enzymatically N-deglycosylated and the released N-glycans were isolated, quantified and profiled by high-pH anion-exchange chromatography. In the course of pregnancy no significant changes were detected in the negative charge distribution stemming from sialic acid and sulfate residues on the complex-type carbohydrate chains of uromodulin. Furthermore, no significant changes in the molar ratio between Man6GlcNAc2 and Man7GlcNAc2 were found in the course of pregnancy, only uromodulin from non-pregnant periods showed small differences.

Preparation and partial structural characterization of alpha1T-glycoprotein from normal human plasma.

alpha1T-glycoprotein (alpha1T) was isolated from normal human plasma in the immunochemically homogeneous state. The partial amino acid sequence and carbohydrate chains of this glycoprotein were determined. To achieve this, the carboxymethylated alpha1T was analyzed by sequencing some of the lysylendoprotease, V8 protease, tryptic, and cyanogen bromide peptides as well as the N-terminal sequence of the protein. A large number of amino acid residues (460 amino acids) was determined by chemical procedure. The peptide sequences were compared with that of other proteins. A high degree of homology was found for proteins of the albumin family. Further, human alpha-albumin, a new member of this protein family, showed an amino acid sequence identical to that of alpha1T indicating that these two proteins are very similar in amino acid sequence and composition. These proteins are closely related to alpha-fetoprotein; however, five carbohydrate chains were found on alpha1T at Asn12, Asn88, Asn362, Asn381, and Asn467 as biantennary complex type chains and the chain on Asn362 possessed a rare consensus sequence of Asn-X-Cys. Thus, alpha1T distinguishes itself by possessing five N-glycans, a finding reported here for the first time for the ALB family.

Characterization of factor VIII/von Willebrand factor concentrates using a modified method of von Willebrand factor multimer analysis.

In order to provide patients with von Willebrand disease a factor VIII (FVIII)/von Willebrand factor (vWF) concentrate of reproducible quality, an SDS-agarose gel electrophoresis method has been established to determine the content of the high molecular weight multimers (band 11 and higher) of vWF. This method has been used to characterize the content of high molecular weight vWF multimers in Humate P/Haemate P, a commercial FVIII/vWF concentrate. The average content of high molecular weight vWF multimers of 47 batches of Humate P/Haemate P has been determined to be 84.1% of the corresponding bands in normal human plasma. Use of this multimer analysis method for the characterization of five further commercial products revealed clear differences with respect to the high molecular weight vWF multimer content. Furthermore, there is a linear correlation (r2 = 0.73) between the content of high molecular weight vWF multimers and the specific activity of vWF (determined as vWF:RCoF/vWF:Ag). The method described here for analysis of the content of high molecular weight vWF multimers is a reliable and reproducible method to characterize this class of factor concentrates with respect to vWF multimer composition.

The hypothetical N-glycan charge: a number that characterizes protein glycosylation.

The production of recombinant glycoprotein therapeutics requires characterization of glycosylation with respect to the lot-to-lot consistency. Here we introduce the ¿hypothetical N-glycan charge Z' as a parameter that allows to characterize the protein glycosylation in a simple, however, efficient manner. The hypothetical N-glycan charge of a given glycoprotein is deduced from the N-glycan mapping profile obtained via HPAE-PAD. In HPAEC, N-glycans are clearly separated according to their charge, i.e., their number of sialic acid residues, providing distinct regions for neutral structures as well as for the mono- di-, tri, and tetrasialylated N-glycans (Hermentin et al., 1992a). Z is defined as the sum of the products of the respective areas (A) in the asialo, monosialo, disialo, trisialo, tetrasialo, and pentasialo region, each multiplied by the corresponding charge: [formula: see text] Thus, a glycoprotein with mostly C4-4* structures will provide Z approximately equal to 400 (e.g., rhu EPO (CHO), Z = 361), a glycoprotein carrying largely C3-3* structures will amount to Z approximately equal to 300 (e.g., bovine fetuin, Z = 290), a glycoprotein with mostly C2-2* structures will have Z approximately equal to 200 (e.g., human serum transferrin, Z = 207, or human plasma AT III, Z = 180), and a glycoprotein carrying only high-mannose type or trunkated structures will provide Z approximately equal to 0 (e.g., bovine pancreas ribonuclease B, Z = 15, and hen ovomucoid, Z = 15, respectively). The determination of Z was validated in multiple repetitive experiments and proved to be highly accurate and reliable. Z may therefore be regarded as a new and characteristic parameter for protein N-glycosylation.

N- and O-glycosylation muteins of recombinant human erythropoietin secreted from BHK-21 cells.

Single-site glycomuteins of recombinant human erythropoietin (rhuEpo) were constructed and transiently and stably expressed in BHK-21 cells. The transient expression levels varied among muteins, being highest for mutein rhuEpoGln24 followed by wild-type rhuEpo (rhuEpowt). All other glycomuteins, including rhuEpoGln38, rhuEpoGln83, rhuEpoThr126, and rhuEpoGly126, were secreted at lower levels than rhuEpowt. Muteins expressed in stable cell lines showed similar differences in expression levels. Also each mutein could be affinity-purified from culture supernatants, and was biologically active in vivo. Based on secretion rates from BHK-21 cells, the most potent erythropoietin was rhuEpoGln24. This mutein is also considered to have biologic activities that are superior to rhuEpowt.

A strategy for the mapping of N-glycans by high-performance capillary electrophoresis.

We have evaluated high-performance capillary electrophoresis (HPCE) with respect to its suitability for use in establishing a carbohydrate-mapping database that would enable a carbohydrate structural analysis by mere comparison of migration times. The suitability of HPCE for carbohydrate structural assignments was ascertained by validation experiments. The migration times of distinct N-glycans, prepared and measured on different days, were shown to be highly reproducible, with a coefficient of variation of usually less than 0.20%, requiring only femtomoles of N-glycan per injection for reliable measurements. By including mesityl oxide and sialic acid as internal standards and a triple-correction method, HPCE fulfills the analytical requirements with respect to accuracy, precision, reproducibility, and sensitivity. The N-glycan-mapping database was established using a newly developed and optimized buffer system containing 1,5-diaminopentane as an organic modifier. Approximately 80 different sialylated N-glycans of known structure, which have thus far been measured and characterized, have been entered into our Lotus 1-2-3 mapping database. The database for structural determinations was tested using the N-linked carbohydrates released from recombinant human urinary erythropoietin (baby hamster kidney) by PNGase F treatment and from bovine serum fetuin and alpha 1-acid glycoprotein by automated and manual (large-scale) hydrazinolysis, respectively. The efficiency of the database and of the triple-correction method was further confirmed by HPCE measurements performed in a different laboratory and by a different analyst who used the HPCE system of a different manufacturer.

The mapping by high-pH anion-exchange chromatography with pulsed amperometric detection and capillary electrophoresis of the carbohydrate moieties of human plasma alpha 1-acid glycoprotein.

The reducing oligosaccharides released from alpha 1-acid glycoprotein (AGP) by conventional hydrazinolysis have been analyzed by two different mapping techniques, using high-pH anion-exchange chromatography with pulsed amperometric detection (HPAE-PAD) and capillary electrophoresis (CE) with uv detection at 190 nm. The CE measurements proved about 4000 times more sensitive than the measurements by HPAE-PAD. The N-glycan pool was fractionated by Mono Q anion-exchange chromatography, and individual fractions so obtained were desialylated using Vibrio cholerae neuraminidase. The resulting asialo-N-glycans were further analyzed by HPAE-PAD, revealing 2 major, 4 intermediate, and 4 small peaks and at least 3 spikes, which counted for at least 13 different asialo-N-glycans. The carbohydrate structures were tentatively assigned by comparison of the Mono Q-separated N-glycans with the known AGP carbohydrate structures and known structures contained in a mapping database that allows structural assignment of N-glycans by mere comparison of retention times. In addition to the hitherto known AGP carbohydrate structures, we have tentatively identified a number of sulfated N-glycans that are currently being analyzed in more detail. We have also compared the glycan pools recovered from AGP using hydrazinolysis and glycopeptidase F (PNGase F). Approximately 40 distinct peaks could be detected in the hydrazinolysis-derived N-glycan pool by either technique (HPAE-PAD and CE), while about 30 distinct peaks were detected in the N-glycan pool derived by PNGase F digestion of the tryptic AGP digest of the same batch of AGP. These differences were attributed to an increased desialylation (approximately 3 mol%) during hydrazinolysis, based on the detection by HPAE-PAD and CE of free sialic acid and monosialylated oligosaccharides in the glycan pool derived by conventional hydrazinolysis. The integrity of the N-glycans' chitobiose core was examined by 500-MHz 1H NMR spectoscopy. The hydrazinolysis procedure could be optimized such that the hydrazinolysis-derived N-glycan pool was chromatographically essentially identical to the PNGase F-derived N-glycan pool. Hydrazinolysis proved best, with practically no loss of N-acetlylneuraminic acid and the closest resemblance to the PNGase F-derived N-glycan pool, using an automated apparatus. Notably, it was recognized that, in our hands, PNGase F digestion in the presence of sodium dodecyl sulfate resulted in partial desialylation of the liberated N-glycans.

Decreased affinity of recombinant antithrombin for heparin due to increased glycosylation.

Recombinant antithrombin produced by baby hamster kidney (BHK) or Chinese hamster ovary (CHO) cells was separated into two fractions, containing comparable amounts of protein, by affinity chromatography on matrix-linked heparin. Fluorescence titrations showed that the more tightly binding fraction had a heparin affinity similar to that of plasma antithrombin (Kd approximately 20 nM), whereas the affinity of the more weakly binding fraction was nearly 10-fold lower (Kd approximately 175 nM). Analyses of the heparin-catalysed rate of inhibition of thrombin further showed that the fractions differed only in their affinity for heparin and not in the intrinsic rate constant of either the uncatalysed or the heparin-catalysed inactivation of thrombin. The recombinant antithrombin fraction with lower heparin affinity migrated more slowly than both the fraction with higher affinity and plasma antithrombin in SDS/PAGE under reducing conditions, consistent with a slightly higher apparent relative molecular mass. This apparent size difference was abolished by the enzymic removal of the carbohydrate side chains from the proteins. Such removal also increased the heparin affinity of the weakly binding fraction, so that it eluted from matrix-linked heparin at a similar position to the deglycosylated tightly binding fraction or plasma antithrombin. Analyses of N-linked carbohydrate side chains showed that the weakly binding fraction from CHO cells had a higher proportion of tetra-antennary and a lower proportion of biantennary oligosaccharides than the tightly binding fraction. We conclude that the recombinant antithrombin produced by the two cell lines is heterogeneously glycosylated and that the increased carbohydrate content of a large proportion of the molecules results in a substantial decrease in the affinity of these molecules for heparin. These findings are of particular relevance for studies aimed at characterizing the heparin-binding site of recombinant antithrombin by site-directed mutagenesis.

A strategy for the mapping of N-glycans by high-pH anion-exchange chromatography with pulsed amperometric detection.

We have evaluated the high-pH anion-exchange chromatography with pulsed amperometric detection (HPAE-PAD) with respect to its suitability to establish a carbohydrate mapping database that would enable carbohydrate structural analysis by mere comparison of retention times. The suitability of HPAE-PAD for carbohydrate structural analysis was ascertained by validation experiments. The retention times of distinct N-glycans, prepared and measured on different days, were shown to be highly reproducible, with a coefficient of variation (CV) of less than 0.5%, requiring less than 100 pmol of N-glycan per injection for reliable measurements. Including appropriate internal chromatographic standards, such as (Neu5Ac)1, (Neu5Ac)2, (Neu5Ac)3, and Neu5Gc, the HPAE-PAD method fulfills the analytical requirements with respect to accuracy, precision, reproducibility, and sensitivity. The N-glycan mapping database was established, using two optimized linear gradients "S" and "A" for sialylated and asialo N-glycans, respectively. Approximately 100 different N-glycans of known structure, which have thus far been measured and characterized, have entered our Lotus 1-2-3 mapping database. The efficiency of the database for structural determinations was tested, using the N-linked carbohydrates isolated from rhuEPO, expressed in BHK cells. Nine different sialylated N-glycans of rhuEPO (BHK) could be assigned with a deviation of less than +/- 0.5%, using gradient S, and six of the eight asialo N-glycans of rhuEPO (BHK) detected with gradient A could be assigned with an accuracy of less than +/- 1%, three of them even with an accuracy of less than 0.1%, providing the reliability of the established HPAE-PAD mapping database.

Generation of bispecific monoclonal antibodies for two phase radioimmunotherapy.

A two phase radioimmunotherapy based on bispecific MAbs in which one arm recognises a tumour antigen and the other a radiolabelled chelate, may prove more effective in the treatment of carcinomas than currently available immunotherapies. To establish this system we first showed that penetration into human carcinoma xenografts as well as long term retention of intact MAb outside the carcinoma cells can be obtained. Epitope saturation was not obtained however, despite the large MAb doses injected i.v. for 10 days. We then generated hybridomas producing high avidity anti-metal chelate MAbs (anti-DTPA-Y). These hybridomas were fused with hybridomas producing MAbs against CEA or GIT-mucin, and stable bispecific MAb producing quadromas were obtained. For the anti-GIT-mucin x anti-chelate MAb a purification procedure based on double anti-idiotype affinity chromatography was shown to result in greater than 95% pure bispecific immunoreactive MAb. Comparative in vivo stability studies profiled DTPA-Y as the chelate of choice for in vivo application.

Percolation and binding of MAB BW 494 to pancreatic carcinoma tissues during high dose immunotherapy and consequences for future therapy modalities.

The distribution of the monoclonal antibody (MAb) BW494 in human pancreatic carcinoma biopsies during high dose i.v. immunotherapy was investigated. Using immunohistochemical techniques combined with anti-idiotypic, endothelial cell specific and bispecific MAbs it was shown that 3 days after onset of immunotherapy, MAb BW494 was bivalently bound to tumor cells in some highly vascualized areas near capillaries. No binding was observed in other highly vascularized tumor cell areas although the epitope detected by MAb BW494 was present. In contrast to our expectation the majority of the tumor cells was not yet saturated by the antibody, probably due to diffusion barriers in the solid tumor tissue.

Hinge-thiol coupling of monoclonal antibody to silanized iron oxide particles and evaluation of magnetic cell depletion.

Iron oxide particles of average size 0.5-1.5 microns, covered by a silane coat carrying amino groups (Bio-Mag, Advanced Magnetics, Boston), were derivatized by reaction with N-[(gamma-maleimidobutyryl)oxy]-succinimide (GMBS), N-hydroxysuccinimidyl iodoacetate (NHIA), 2-iminothiolane (2-It), or N-succinimidyl 3-(2-pyridyldithio)propionate (SPDP). The derivatized particles were suitable for the reaction with sulfhydryl groups and subsequently coated with monoclonal antibodies (MoAbs) of different classes and isotypes (IgM, IgG1, IgG2a, IgG2b, IgG3) as well as polyclonal rabbit anti-mouse IgG (RAM). The antibodies were reduced by dithiothreitol (DTT) and covalently conjugated to the BioMag derivatives via liberated sulfhydryls of the hinge region. The observed conjugation ratios, expressed as protein/iron (micrograms/mg), could be reproducibly varied for optimization. These ratios were dependent on the type and amount of antibody offered for coupling to the derivatized particles, decreasing as follows: polyclonal = IgM greater than IgG2b greater than IgG2a = IgG3 greater IgG1. The conjugation ratios were also dependent on the type and amount of the spacer used to derivatize the BioMag particles, decreasing as follows: GMBS greater than NHIA greater than 2-It greater than SPDP. The magnetically responsive magnetite-antibody conjugates ("magneto-beads"), carrying MoAb BMA 081 (anti-CD8; IgG2a), MoAb BB10 (anti-CD10/CALLA; IgG2b), MoAb VIL-A1 (anti-CD10; IgM), and polyclonal RAM, coupled similarly via 3.6 mumol of GMBS spacer per mg of Fe, were further investigated with respect to a depletion effect on specific cell subsets. The rates of cell depletion were found to be strongly dependent on the individual characteristics of the antibody used.(ABSTRACT TRUNCATED AT 250 WORDS)

Percolation and binding of monoclonal antibody BW494 to pancreatic carcinoma tissues during high dose immunotherapy and consequences for future therapy modalities.

The distribution of the monoclonal antibody (MAb) BW494 in human pancreatic carcinoma biopsies during high dose intravenous immunotherapy was investigated. Using immunohistochemical techniques combined with anti-idiotypic, endothelial cell-specific and bispecific MAbs, it was shown that 3 days after onset of immunotherapy, MAb BW494 was bivalently bound to tumour cells in some highly vascularised areas near capillaries. No binding was observed in other highly vascularised tumour cell areas although the epitope detected by MAb BW494 was present. In contrast to our expectation the majority of the tumour cells were not yet saturated by the antibody, probably due to diffusion barriers in the solid tumour tissue.

Attachment of rhodosaminylanthracyclinone-type anthracyclines to the hinge region of monoclonal antibodies.

We have found that a maleimidobenzoyl spacer attached to OH-4' of the rhodosamine moiety of rhodosaminylanthracyclinone-type anthracyclines is most suitable for the attachment of these drugs to carriers, providing important advantages: The spacer is selectively and most readily introduced into the rhodosamine moiety of the drugs, is stable enough for proper handling of the derivatives, and can easily be attached to thiol groups of carrier systems such as reduced monoclonal antibodies. The anthracyclines can be liberated from the conjugates by mere hydrolysis, requiring neither hydrolytic enzymes nor acidic pH. Liberation of the drugs can, moreover, be affected by the presence of the appropriate substituents Z on the phenylene ring of the spacer, thus allowing slowed or enhanced liberation of the cytostatically active drug. The corresponding p-maleimidobenzoyl derivatives of beta-rhodomycin I, N,N-dimethyldaunorubicin, and rodorubicin have been attached to thiol groups of the hinge region of reduced monoclonal antibody BW 494/32, directed against a pancreatic cancer associated glycoprotein antigen, resulting in MoAb BW 494/32 conjugates, carrying 4.8-6.8 mol of cytotoxic residues/mol of MoAb. Rodorubicin was similarly attached to MoAb BW 575/931/2, directed against a small cell lung cancer associated antigen and to MoAb BW 431/26, recognizing an epitope detectable on carcinoembryonic antigen. The results provide evidence that the newly developed method of coupling of anthracyclines to the hinge region of monoclonal antibodies may be of broader use.

Structure-activity relationship of anthracyclines in vitro.

The cytotoxic activities of several natural and semisynthetic anthracyclines against L1210 leukemia and two human colon tumor cells (Colon 4, HT 29) in vitro were examined after short (1 h) and long (7 days) incubation times and correlated with the water/octanol partition coefficients and the DNA-binding affinity of the compounds. Analysis of equation in which cytotoxicity against L1210 (1-h incubation) was parabolically related to the partition coefficient revealed an almost exclusive correlation (r = 0.80) between the cytotoxicity and the parameters, and this correlation was only slightly improved by addition of DNA-binding affinity (r = 0.85). On the other hand, cytotoxic activities displayed after continuous incubation were partially related to both partition coefficients (parabolic dependence) and DNA-binding affinities (linear dependence). In this case the correlation between the activity and partition coefficient (r = 0.67) was significantly improved by addition of DNA-binding affinity (r = 0.90). Similar results were also obtained for human colon tumor cells although the corresponding correlation coefficients were generally of lower value, indicating that cytotoxic activity of anthracyclines against these primary resistant cells may be influenced by additional factors not yet determined.

Investigations with monoclonal antibody drug (anthracycline) conjugates.

The development of monoclonal antibody-drug (anthracycline) conjugates (immunocytostatics) that has emerged during the last decade is briefly reviewed. Stress is layed on the various procedures that have been employed for the chemical attachment of daunorubicin (daunomycin) and doxorubicin (adriamycin) to spacers, high-molecular weight carrier molecules, and immunoglobulins. Anthracycline conjugates that have proved effective in mice with respect to the treatment of cancer are especially evaluated. Also a new method developed for the conjugation of rhodosaminyl anthracyclinones via a hydrolysable spacer, developed in our laboratories, is briefly communicated. Finally, import aspects for further developments of monoclonal antibody-drug conjugates are discussed.