Finnish Red Cross Blood Service and Laboratory of Biochemistry and Microbiology, Department of Chemical Technology, Helsinki University of Technology, Finland
 

A PHARMACEUTICAL HUMAN APOTRANSFERRIN

PRODUCT FOR IRON BINDING THERAPY

Leni von Bonsdorff

Dissertation for the degree of Doctor of Science in Technology to be presented with due permission of the Department of Chemical Technology for public examination and debate in the Auditorium KE 2 (Komppa Auditorium) at Helsinki University of

Technology (Espoo, Finland) on the 14th of November, 2003, at 12 noon.

Helsinki 2003

ACADEMIC DISSERTATIONS FROM

THE FINNISH RED CROSS BLOOD SERVICE

NUMBER 48

SUPERVISOR

Docent Jaakko Parkkinen, MD, PhD

Department for Research and Development

Finnish Red Cross Blood Service

REVIEWERS

Professor Maija Tenkanen, DrSc (Tech)

Department of Applied Chemistry and Microbiology

University of Helsinki

Helsinki, Finland

Docent Tapani Tuomi, DrSc (Tech)

Finnish Institute of Occupational Health

Helsinki, Finland

OPPONENT

Jan Over, PhD

Sanquin Plasma Products

Amsterdam, The Netherlands

ISBN 952-5457-05-2 (print)

ISBN 952-5457-06-0 (pdf)

ISSN 1236-0341

Helsinki 2003

Yliopistopaino

ABSTRACT

Transferrin is the major iron binding protein in human plasma. It binds iron with high affinity in a redox inactive form and delivers it to growing cells. Each molecule is capable of binding two molecules of ferric iron. Normally, transferrin is only about 30% saturated with iron. In certain clinical conditions, the iron concentration in serum is increased so that the iron binding capacity is exceeded and non-transferrin-bound iron (NTBI) is formed in serum. NTBI is potentially toxic because it generates free radical formation and can be taken up by tissues, leading to excess deposits that can potentiate tissue damage. It is also known that iron enhances the growth of bacteria and fungi, and can predispose patients to septic infections.

This thesis describes the development of an efficient process for producing pharmaceutical grade iron-free apotransferrin. The biochemical efficacy of apotransferrin for iron binding therapy was studied in early phase clinical trials in haematological stem cell transplant (SCT) patients. The scope of this work did not include studying the clinical efficacy of apotransferrin.

The manufacturing method used fraction IV of the Cohn cold ethanol human plasma fractionation process as starting material. Apotransferrin was purified in two ion exchange chromatography steps and ultrafiltration with over 90% recovery. In order to obtain a virus-safe product, the process comprised solvent detergent treatment as the main virus inactivation step and virus filtration and polyethylene glycol precipitation to remove physico-chemically resistant infectious agents. The purity of the product was at least 98%, main impurities being IgG, IgA and hemopexin. Methods for studying the iron binding capacity, the transferrin conformation and its iron forms, and the glycosylation variants were developed and used to study the quality of the finished product batches. The product had intact iron binding capacity and a native conformation. The results of several production batches indicated that the manufacturing could be carried out reproducibly. Product characterisation by electrospray and MALDI-TOF mass spectrometry indicated no other chemical modifications than N-linked glycan chains and disulphide bonds, except minor oxidation. A stable liquid formulation suitable for intravenous infusion was developed.

The biochemical binding of NTBI to apotransferrin in vivo was studied by several methods. The bleomycin method for NTBI determination was modified for microwell measurement and evaluated. The bleomycin assay was reproducible and NTBI was found in serum samples only when transferrin saturation was >80% and haemolysed samples were excluded. The bleomycin assay that measures redox-active iron underestimated the true concentration of NTBI. The concentration of NTBI could be calculated from the shift of transferrin iron forms found in vivo after intravenous infusion of apotransferrin to patients. It could also be determined with a chelation based method, which, however, had a lower specificity than the bleomycin method. In haematological SCT patients, the concentration of NTBI could be as high as 20 µmol/l.

Apotransferrin given in single intravenous doses to six patients bound NTBI effectively, although in most cases temporarily. With repeated high dose regimens, the appearance of NTBI was prevented in 5 of 8 patients.

The influence of NTBI on the growth of the opportunistic pathogen Staphylococcus epidermidis was studied both with purified transferrin and in serum milieu. In both cases, growth was critically dependent on NTBI and on a high transferrin saturation.

Only at high initial bacterial concentrations could growth be detected with partially saturated transferrin. Apotransferrin administered to SCT patients bound NTBI and restored the growth inhibitory effect of serum. Exogenous apotransferrin might protect the patients against infections by S. epidermidis and other opportunistic pathogens whose growth is dependent on NTBI.

In conclusion, the apotransferrin was pure and safe and showed in vivo the biochemical effects that could be expected of a functional human apotransferrin product. In SCT patients it was possible to prevent the appearance of NTBI and maintain the bacterial growth inhibitory effect in serum.

PREFACE

This work was carried out at the Department of Research and Development, Division of Plasma Products of the Finnish Red Cross Blood Service in Helsinki. I am grateful to the Director of the Institute Docent Jukka Rautonen and the former Director Professor Juhani Leikola as well as the Division Director Visa Vanamo for providing the research facilities and the possibility to carry out this work. Former Division Director Professor Gunnar Myllylä is thanked for long lasting encouragement to complete my scientific studies.

My sincerest thanks go to my supervisor Docent Jaakko Parkkinen, who with great scientific knowledge has guided me throughout this work. I am deeply grateful for the time and effort he has given to the research, for bringing in new ideas and for over again encouraging me to check the details. I truly admire his skill for always finding a simple way of expressing complex problems in a few words.

I thank Professor Simo Laakso, head of the Laboratory of Biochemistry and Microbiology at the Helsinki University of Technology, for encouragement, help and advice. I also thank the reviewers Professor Maija Tenkanen and Docent Tapani Tuomi for careful reading of the thesis and for constructive criticism and valuable discussions.

Of my co-authors of the original publications at FRC, Hannele Tölö is especially thanked for her quick mind and action, and for many supportive discussions, not only in the field of science. My thanks go to Enni Lindeberg for pondering the problems with iron methodology and to Esa Törmä for fruitful collaboration with the virus steps. Sanna Matinaho, Ari Rouhiainen, and Maarit Lönnroth are thanked for providing valuable contributions to the bacterial work.

I express my deepest gratitude to the clinicians at the Helsinki University Central Hospital who carried out the clinical studies. Leila Sahlsted is warmly thanked for a most pleasant collaboration and for meticulously collecting patient samples and data.

Doctor Freja Ebeling is thanked for her invaluable advice and strong support. The advice of Professor Tapani Ruutu is warmly acknowledged.

I am indebted to Doctor Tuula Nyman for the contributions to the protein characterisation work carried out at the University of Helsinki, Institute of Biotechnology.

At the Department of R&D at FRC, I wish to thank all my colleagues for great support.

Especially the excellent technical assistance and joyful times in the laboratory with Teija Kupari, Anne Remes, Sisko Lehmonen, Sinikka Laazizi, Anne Rahola and Hanna Laitinen is warmly remembered. Helena Santala is thanked for true help and support in secretarial as well as other matters.

The "production guys" Mauri Asplund, Pekka Nikkilä, and Lauri Aho are thanked for keen interest in the project and help throughout test runs and optimisation work. Doctor Gunnel Sievers and the chemists at the Department of Chemistry are warmly acknowledged for advice and support. My thanks to Doctor Ritva Soininen for teaching me the basics of GMP, and to the personnel at the Quality Assurance Department for good collaboration. Marja-Leena Hyvönen and Maija Ekholm are thanked for excellent librarian help and Helena Hurri for help with many practicalities.

My former bosses from the Provivo days, Ralf Lundell and Doctor Andrea Holmberg, are thanked for starting up the transferrin project, and colleagues Leena Rahkamo and

Arto Forsberg for fun and rewarding collaboration.

My dear friends (including Dino), thank you for being there!

I warmly thank my parents, Tina and Magnus, for always caring and supporting me in my studies, and for helping my family in many ways. My brothers Jan and Petter and their families are thanked for support and keen interest in my work.

My children Johan, Mona and Mats, thank you for putting up with an absent or absentminded mother. Anders, my dear husband, I cannot thank you enough.

I acknowledge the financial support of Anna and Signe von Bonsdorffs släktfond.

Helsinki, October 2003

Leni von Bonsdorff

TABLE OF CONTENTS

ABSTRACT ......................................................................................................................... 3

PREFACE ............................................................................................................................ 5

LIST OF ORIGINAL PUBLICATIONS....................................................................................... 9

ABBREVIATIONS .............................................................................................................. 10

INTRODUCTION ................................................................................................................ 11

1 Transferrin and serum iron .................................................................................... 11

1.1 Transferrin characteristics........................................................................... 11

1.2 Iron and role of transferrin.......................................................................... 13

1.3 Serum non-transferrin-bound iron (NTBI) ................................................. 15

1.4 Measurement of NTBI in serum ................................................................. 17

2 Iron and infection .................................................................................................. 20

2.1 NTBI, microbial growth, and risk of infection ........................................... 20

2.2 Iron acquisition of Staphylococcus epidermidis ......................................... 22

3 Methods to produce apotransferrin........................................................................ 22

AIMS OF THE STUDY......................................................................................................... 24

MATERIALS AND METHODS.............................................................................................. 25

1 Methods for production, identification and characterisation of apotransferrin (I) 25

1.1 Apotransferrin production method.............................................................. 25

1.2 Identification of transferrin ......................................................................... 25

1.3 Apotransferrin characteristics and purity.................................................... 25

2 Serum iron parameters (II-V) ................................................................................ 27

2.1 Patient serum samples................................................................................. 27

2.2 Serum iron and transferrin saturation ......................................................... 27

2.3 Determination of serum NTBI .................................................................... 27

2.4 Determination of transferrin iron forms in serum....................................... 28

3 GrowthofStaphylococcus epidermidis (IV, V).................................................... 29

3.1 Bacterial strains and preparation of inoculum ............................................ 29

3.2 Monitoring of growth with pure transferrin................................................ 29

3.3 Monitoring of growth in serum samples..................................................... 29

RESULTS .......................................................................................................................... 31

1 Pharmaceutical apotransferrin preparation (I)....................................................... 31

1.1 Large scale purification of apotransferrin................................................... 31

1.2 Identification and characterisation of the apotransferrin product ............... 32

2 Binding of NTBI with apotransferrin .................................................................... 34

2.1 Evaluation of assay for measuring NTBI (II) ............................................. 34

2.2 Binding of NTBI in patient serum (III, V).................................................. 35

3 GrowthofStaphylococcus epidermidis (IV, V).................................................... 36

DISCUSSION ..................................................................................................................... 38

1 Apotransferrin product (I) ..................................................................................... 38

2 Determination of serum NTBI (II, III) .................................................................. 39

3 Binding of NTBI by apotransferrin administration (III, V) .................................. 41

4 Transferrin, NTBI and the growth of Staphylococcus epidermidis (IV,V)........... 42

CONCLUSIONS.................................................................................................................. 45

REFERENCES .................................................................................................................... 46

LIST OF ORIGINAL PUBLICATIONS

This thesis is based on the following original publications which are referred to in the text by their Roman numerals.

I von Bonsdorff L, Tölö H, Lindeberg E, Nyman T, Harju A, Parkkinen J.

Development of a pharmaceutical apotransferrin product for iron binding therapy. Biologicals 2001;29:27-37.

II von Bonsdorff L, Lindeberg E, Sahlstedt L, Lehto J, Parkkinen J. Bleomycindetectable iron assay for non-transferrin-bound iron in hematologic malignancies. Clin Chem 2002;48:307-314.

III Sahlstedt L, von Bonsdorff L, Ebeling F, Ruutu T, Parkkinen J. Effective binding of free iron by a single intravenous dose of human apotransferrin in haematological stem cell transplant patients. Br J Haematol 2002;119:547-553.

IV Matinaho S, von Bonsdorff L, Rouhiainen A, Lönnroth M, Parkkinen J.

Dependence of Staphylococcus epidermidis on non-transferrin-bound iron for growth. FEMS Microbiol Lett 2001;196:177-182.

V von Bonsdorff L, Sahlstedt L, Ebeling F, Ruutu T, Parkkinen J. Apotransferrin administration prevents growth of Staphylococcus epidermidis in serum of stem cell transplant patients by binding of free iron. FEMS Immunol Med Microbiol 2003;37:45-51.

In publication III, the authors L. Sahlstedt and L. von Bonsdorff contributed equally.

ABBREVIATIONS

AAS Atomic absorption spectrophotometry

ATCC American Type Culture Collection

BPT Bathophenantroline

BSA Bovine serum albumin

BVDV Bovine viral diarrhoea virus

CFU Colony forming units

CRP C-reactive protein

CV Coefficient of variation

DNA Deoxyribonucleic acid

EDTA Ethylenediamine tetraacetic acid

FeNTA Ferric nitrilotriacetate

GAPDH Glyceraldehyde-3-phosphate dehydrogenase

HAV Hepatitis A virus

HIV Human immunodeficiency virus

IEX-HPLC Ion exchange high performance liquid chromatography

IgA Immunoglobulin A

IgG Immunoglobulin G

MALDI-TOF Matrix assisted laser desorption ionization-time of flight

NTA Nitrilotriacetic acid

NTBI Non-transferrin-bound iron

OD Optical density

PAGE Polyacrylamide gel electrophoresis

PBS Phosphate buffered saline

PCR Polymerase chain reaction

PEG Polyethylene glycol

RP-HPLC Reversed phase high performance liquid chromatography

SCT Stem cell transplant

SD Solvent detergent

SDS Sodium dodecyl sulphate

SEC-HPLC Size exclusion high performance liquid chromatography

TBA Thiobarbituric acid

TBE Tris-borate-EDTA

INTRODUCTION

1 Transferrin and serum iron

1.1 Transferrin characteristics

Transferrins represent a class of proteins found in biological fluids of all vertebrates and invertebrates with the property of reversibly binding iron. The name transferrin was suggested by Holmberg and Laurell (1947) and is used as the name for the iron-binding β-globulin found in blood serum and other extracellular fluids of vertebrates (de Jong et al., 1990). Other proteins of the same group are lactoferrin in milk and in neutrophil granules (Metz-Boutigue et al., 1984), ovotransferrin of egg white and melanotransferrin (antigen p97) produced by melanoma cells (Brown et al., 1982). Transferrin was first detected and isolated in works described by Holmberg and Laurell (1945) and Schade and Caroline (1946). The latter group used the newly developed Cohn fractionation procedure to prepare the iron-binding fraction of plasma. Since then, the protein has been thoroughly studied and characterised (de Jong et al., 1990; Evans et al., 1999; Harris and Aisen, 1989).

Human transferrin is a single polypeptide chain containing 679 amino acid residues (MacGillivray et al., 1983). The protein has 19 disulphide bridges. Crystallographic studies have shown that the transferrin molecule is organised into two homologous lobes of about 330 amino acid residues, the N- and the C-lobe. The lobes are linked by a short flexible spacer peptide and each lobe contains two dissimilar domains divided by a cleft which is the binding site for Fe3+ (Bailey et al., 1988; Wang et al., 1992) (Figure 1). At the iron binding site, four of the six Fe3+ co-ordination sites are occupied by the protein ligands (2 tyrosine, 1 histidine and 1 aspartate residue) and two by the bidentate carbonate anion (Bailey et al., 1988; Hirose, 2000). The synergistic binding of an anion, preferentially the carbonate molecule, is essential for the iron binding (Harris and Aisen, 1989).

Two N-linked oligosaccharides are found in the C-lobe at aspargine residues Asn413 and Asn611. The carbohydrates constitute about 6% of the mass of transferrin that has a molecular weight of 79750. The glycan chains are mainly biantennary (85%) and triantennary (15%) complex-type glycans (Fu and van Halbeek, 1992; Spik et al., 1985). The number of sialic acid residues per transferrin molecule is between 4 and 6. Glycosylation variants occur in different conditions. In patients suffering from alcoholism, carbohydrates lacking two to four of the terminal trisaccharides, comprising the negatively charged sialic acid and the neutral N-acetylglucosamine and galactose have been found, and disialotransferrin and to a lesser degree mono- and asialotransferrin are present (Stibler, 1991). Such carbohydrate-deficient transferrin can be used as a marker for alcoholism (Turpeinen et al., 2001). Increase of branching has been found during pregnancy (Leger et al., 1989) and in congenital glycosylation disorder, when there also is an increase in fucosylation (Mills et al., 2001; Yamashita et al., 1993). The variation of glycosylation is a determinant of the microheterogeneity of transferrin. Another determinant is the genetic polymorphism. Genetic variants were first detected by starch electrophoresis. The most common was designated TfC, the more anodal TfB and cathodal TfD. Variants within the subgroups have been identified with isoelectric focusing (Kamboh and Ferrell, 1987). Of at least 38 variants, only 4 occur with a frequency over 1% (de Jong et al., 1990). Among Caucasians, the C variant, having several subtypes including C1, C2 and C3, is found almost exclusively

(Tenkanen et al., 1989). A single amino acid substitution determines the TfC1 and TfC2 variants (Namekata et al., 1997). The highest frequencies of the D variant is found among African blacks, although hardly ever as homozygous D phenotypes (Kasvosve et al., 2000). Attempts to establish a relationship between the transferrin variant, serum transferrin concentration, and iron-binding capacity in Caucasians have been inconclusive (Sikstrom et al., 1993). A functional difference was demonstrated between CD and CC phenotypes in blacks, with a lower in vitro binding of iron to transferrin for the CD individuals and differences in serum iron parameters between the phenotypes (Kasvosve et al., 2000). In a transferrin variant where an amino acid substitution in the C-lobe leads to an open structure of the iron saturated molecule, a reduced affinity for the transferrin receptor was found (Evans et al., 1994b).

Each molecule of transferrin can bind two Fe3+ ions. The four iron forms of transferrin are the iron-free apotransferrin, the monoferric transferrins with iron in the C- or the Nlobe, respectively, and the diferric holotransferrin. The affinity for iron at physiological pH 7.4 is high, with a binding constant of about 1022 (Harris and Aisen, 1989). Upon binding of iron, the lobes undergo a conformational transition from the apo-structure with an open interdomain cleft to a closed holo-structure (Hirose, 2000). The conformational change can be studied by urea polyacrylamide gel electrophoresis first described by (Makey and Seal, 1976), whereby the different iron forms of transferrin can be resolved. The presence of urea causes the lobes that do not contain iron to unfold and decreases their mobility (Evans and Williams, 1980). It is believed that the change in conformation is important for the binding of the molecule to the transferrin receptor, and diferric transferrin has a 10-100 times higher affinity for the receptor than apotransferrin (Richardson and Ponka, 1997). The binding mechanism of transferrin to its receptor is not fully elucidated, but it is believed that the recognition site requires the presence of both transferrin lobes (Mason et al., 1997; Mason et al., 2002; Zak et al., 2002) and that the glycans are of less importance (Hoefkens et al., 1997). In circulation, where normally only a third of the transferrin iron binding sites are occupied, iron preferentially binds to the N-terminal lobe which is also called the acid labile lobe (Evans et al., 1999). The metal release from transferrin is mainly accomplished by a decrease in pH, where protonation of the iron ligands causes release of the iron. The iron release can also be accelerated by other chemical compounds capable of complexing iron such as pyrophosphates (Morgan, 1979) and citrate (Gumerov et al., 2003). Transferrin is capable of binding several other metals, but with a lower affinity (Harris and Aisen, 1989).

Figure 1. A ribbon diagram of a diferric rabbit serum transferrin molecule. The arrow indicates the position of the Fe3+ molecule in the inter-domain cleft in the N-lobe. The model is based on the PDB atomic coordinate file 1JNF (Hall et al., 2002).

1.2 Iron and role of transferrin

Iron is essential for virtually all living organisms. It is needed in proteins for oxygen transportation and for enzymes that can catalyse redox reactions and DNA synthesis. In biological systems and aqueous solutions, the transition metal iron can exist in two oxidative states as ferrous (Fe2+) or ferric iron (Fe3+), capable of respectively donating or accepting electrons. In aerobic conditions, Fe2+ is oxidised to Fe3+, which is the most common form found in the environment. The properties of iron are associated with two types of problems. First, Fe3+ is highly insoluble at physiological pH. With increasing pH, Fe3+ readily hydrolyses to hydroxy-forms, followed by irreversible formation of hydroxy-iron polymers that precipitate. At physiological pH 7, the solubility of Fe3+ is only about 10^-18 mol/l (Spiro and Saltman, 1974). Secondly, the redox active iron can be toxic as it is capable of forming harmful free radicals (Halliwell and Gutteridge, 1985). Ferrous iron catalyses the formation of the highly oxidising hydroxyl radical (OH⋅) from hydrogen peroxide, known as the Fenton reaction (1). Ferric iron can be reduced to ferrous iron by superoxide (O2 ^-). This net reaction (2) is referred to as the Haber-Weiss reaction. However, the term "Fenton-catalysed Haber-Weiss reaction" often used to describe these reactions, has been criticised. In vivo, the superoxide is mainly considered to be the source of hydrogen peroxide for the Fenton reaction, and other agents as well as superoxide can reduce iron. Toxicity therefore arises from the Fenton reaction alone, and the Haber-Weiss reaction probably does not take place (Koppenol, 2001).

Fe2++ H2O2 -> Fe3+ + OH- + OH∙ (1)

O2

- + H+ + H2O2 ->O2 + OH∙ + H2O (2)

The hydroxyl radical is, although short-lived, highly reactive and able to cause secondary radicals and initiate the lipid peroxidation chain reaction. It also causes DNA strand breaks, inactivates enzymes and can depolymerise polysaccharides (McCord, 1998).

In the blood stream, iron is transported to the cells bound to transferrin. This type of iron transportation brings several benefits. First, the transferrin is only about 30% saturated with iron to ensure capacity to sequester available iron. Second, iron is kept in a redox-inactive form. It has been shown that transferrin-bound iron is not capable of promoting hydroxyl radical formation (Baldwin et al., 1984). Third, iron is not hydrolysed as it would be if were in the form of a free salt and can be targeted directly to cells requiring iron. Transferrin-bound iron is taken up by the cells by receptormediated endocytosis (Richardson and Ponka, 1997) whereafter apotransferrin is recirculated to the blood stream.

Iron is used for the synthesis of proteins requiring iron for biological activity, or stored in intracellular ferritin. Several other iron-binding proteins are involved in the human iron metabolism. Although transferrin is of utmost importance in the transportation of iron, the ∼3-4 mg of transferrin-bound iron is only ∼0.1% of the total body iron of ∼4000 mg. However, the transferrin iron turnover is significant and about 30 mg of iron is transported daily to the cells. Of this, about 80% is transported to the bone marrow for haemoglobin synthesis in developing erythroid cells. Haemoglobin iron comprises 60-70% of body iron, followed by the intracellular storage protein ferritin that holds 10-20% of iron. The rest of the iron is found in myoglobin, cytochromes and ironcontaining enzymes. Body iron stores are highly conserved, only 1-2 mg is lost daily and the iron balance is maintained by dietary iron absorption (Conrad and Umbreit, 2000; Fairbanks and Beutler, 1995).

1.3 Serum non-transferrin-bound iron (NTBI)

In certain clinical conditions, the iron concentration in serum is increased. If the iron binding capacity of transferrin is exceeded, non-transferrin-bound iron (NTBI) can form in serum. The reasons for the increase of iron are diverse (Table 1). Although the causes of NTBI vary, it is generally considered that the presence of NTBI is a pathological manifestation and is never found in healthy individuals (Breuer et al., 2000a). NTBI was first detected in chronic iron overload disorders such as haemochromatosis. It can also be detected in acute events such as for haematological patients undergoing myeloablative conditioning therapy, and especially those who receive stem cell transplants (SCT) (Bradley et al., 1997; Sahlstedt et al., 2001). The myeloablative conditioning therapy results in a halt of erythropoiesis, the only effective way of transferrin iron utilisation. If iron release to circulation continues from macrophages taking up senescent erythrocytes, this results in a remarkable increase in the total serum iron level.

The form and nature in which NTBI exists in serum is not fully elucidated (Esposito et al., 2002; Hider, 2002). Primarily, the formation of NTBI was considered a spillover phenomenon when the transferrin saturation was exceeded. According to (Grootveld et al., 1989), a significant portion of NTBI in haemochromatosis is in the form of citrate and, possibly, acetate complexes. The tendency of iron to form large iron-polymers at physiological conditions could lead to different types of iron-citrate complexes, varying in size from monomeric to large oligomer complexes with even 17-19 iron molecules (Hider, 2002). Albumin could also act as an iron binding ligand due to the presence of large number of negatively charged carboxylate sites (He and Carter, 1992). In a study using pure albumin, it was found that albumin had capacity to bind added iron which resulted in inhibition of lipid peroxidation (Loban et al., 1997). Studies with hypotransferrinaemic mice have indicated a mixture of iron species with differing reactivity to NTBI assays (Simpson et al., 1992). The concentrations varied in the range of 1-20 µmol/l depending on the assay used to measure NTBI. No detectable mononuclear iron was measured by electron paramagnetic resonance spectrometry, indicating lack of citrate-bound iron in monomeric form. Thus, NTBI is probably a mixture of different forms of iron bound to extracellular components, and the proportions of the iron forms may vary between patient groups and clinical conditions.

Table 1. Examples of clinical conditions when NTBI has been detected in serum.

Formula (3) is based on the maximal binding of 2 mol Fe3+ per 1 mol transferrin and a molecular weight of 79570 for transferrin (Morgan, 1992).

2.3 Determination of serum NTBI

The bleomycin-detectable iron assay by Evans and Halliwell (1994) was the primary assay for NTBI determination in publications II, III, and V, and it was adopted for a halved sample size. Additions were done in the following order: 250 µl 1 mg/ml DNA (Type I DNA from calf thymus, Sigma), 25 µl 1.5 IU/ml bleomycin sulphate (Sigma), 50 µl 50 mmol/l MgCl2, 25 µl of sample, standard or blank and finally 50 µl 8 mmol/l ascorbic acid (Merck, Germany). A predetermined amount of about 7 µl 25 mmol/l HCl was added before the sample to adjust the pH of reagent mixture to 7.4. All reagents except the bleomycin were treated with Chelex (Bio-Rad) to remove excess iron. After incubation for 1 h at 37° C, 50 µl 0.1 mol/l EDTA, 250 µl 1% (w/v) thiobarbituric acid (Sigma), and 250 µl 25% (v/v) HCl were added and the mixture was incubated at 80°C for 20 min for chromogen formation. The pink chromogen complex was extracted with 1.5 ml butanol and the phases were separated by centrifugation. 350 µl of the clear top phase was transferred to microwell plates and the absorbance was measured at 535 nm using a microplate reader. The samples were measured in parallel with a corresponding blank without the addition of bleomycin. The absorbance value of the blank was subtracted from each sample absorbance value. The reagent blank value was subtracted from the absorbance values of the standards, and a standard curve with 0.1 to 3 µmol/l of iron (Iron atomic absorption standard solution in 1% HCl, Sigma Aldrich) was calculated for each series. Data was transformed to logarithmic values before calculation of the standard curve by linear regression in order to give weight to the standards with a low iron concentration (0.1-1 µmol/l) before fitting the curve.

The chelation method with NTA (Gosriwatana et al., 1999) was used in publications II and IV. 100 µl 800 mmol/l NTA with pH 7 was added to 900 µl serum and the NTAiron complex was recovered by ultrafiltration after 30 min at room temperature. The iron in the ultrafiltrate was quantified with a colorimetric assay with bathophenantroline chromogen or by atomic absorption spectrophotometry.

2.4 Determination of transferrin iron forms in serum

To study the transferrin iron forms in serum samples, the serum samples were either precipitated with rivanol before urea-PAGE, or the transferrin bands were visualised by immunoblotting after electrophoresis of serum. In both cases, the electrophoresis was carried out as described for the pure transferrin samples.

To obtain a transferrin fraction, 400 µl of 0.375% (w/v) rivanol (6,9-diamino-2- ethoxyacridine lactate, Sigma) with 10% (w/v) glycerol in TBE buffer was added to 50 µl of serum and the precipitate was removed by centrifugation (Williams and Moreton, 1980). Samples corresponding to 2 µl of serum were applied to the gel. Proteins were visualised with Coomassie Brilliant Blue staining.

For immunoblotting, serum samples with about 0.15 µg of transferrin were separated in 10x10 cm gels. Proteins were electroblotted from the gel onto polyvinylidene fluoride membrane (Immobilon-P, Millipore) in a transfer buffer containing 25 mmol/l Tris, 192 mmol/l glycine and 20% (v/v) methanol. The membrane was treated with 0.5% (v/v) Tween in PBS over night. Transferrin bands were visualised by immunostaining using rabbit anti-human transferrin IgG (Dako A/S) as the primary antibody in 1% (w/v) BSA, 0.05% (v/v) Tween 20 in PBS for 2 h at room temperature. The blots were washed three times with PBS containing 0.05% Tween 20 and incubated with anti-rabbit IgG conjugated with alkaline phosphatase (Jackson Immuno Research Laboratories Inc) in 1% BSA, 0.05% Tween 20 in PBS for 1 h at room temperature. Following three additional washes with 0.05% Tween 20 in PBS, the blots were stained with 5-bromo-4- chloro-3-indolyl phosphate-nitroblue tetrazolium colour development solution (Immuno-Blot Alkaline Phosphatase Assay kit, Bio-Rad). The reaction was stopped with 100 mmol/l Na-acetate, pH 5, containing 5 mmol/l Na-EDTA for 5 min. The blots were washed with distilled water and dried.

The proportions of the different transferrin iron forms were determined by scanning with a laser densitometer.

3 GrowthofStaphylococcus epidermidis (IV, V)

3.1 Bacterial strains and preparation of inoculum

Two multi-resistant clinical isolates of S. epidermidis (16779 and 19435) from neutropenic patients with a haematological malignancy and septicaemia and the ATCC 12228 strain (Manassas, VA, USA) were used. Stock cultures were kept frozen at -70°C and subcultures were made on trypcase-soy agar dishes (bioMérieux, France). The growth medium was RPMI-1640 (R-7509, Sigma) with 50 mmol/l HEPES buffer, pH 7.4 and 2 mmol/l L-glutamine. The medium was depleted of iron using Chelex-100 resin (Bio-Rad) and supplemented with 10 µmol/l CaCl2 and 100 µmol/l MgCl2. Ferric nitrilotriacetate (FeNTA) with 3 mmol/l iron was used for iron addition in precultivation of the strains and for saturating normal serum with iron.

To prepare the inoculum, the strains were cultivated in the iron-depleted RPMI with or without iron supplementation. The iron-depleted RPMI was used to induce the siderophore production. The siderophore index of the bacterial cultures was determined with the Chrome Azurol S liquid assay (Payne, 1994). 10-fold dilutions of the bacteria in saline were used as inocula.

3.2 Monitoring of growth with pure transferrin

Inocula with and without siderophore induction were used. 50 µl was used to inoculate 250 µl of RPMI containing 2.5 g/l of apotransferrin and different amounts of FeNTA. The growth was monitored by turbidity measurements every 30 min in microwell plates for 96 h at 37°C with periodic shaking in a Bioscreen C Microbiology Reader (Labsystems, Finland). Viable counts were determined by plating on trypcase-soy agar plates (bioMérieux, France).

3.3 Monitoring of growth in serum samples

The inoculum was cultivated in two passages, the first one in iron replete RPMI and the second one in iron-depleted RPMI. Serum was buffered with 50 mmol/l HEPES, pH 7.4. 50 µl inoculum was added to 200 µl buffered serum. The growth was monitored by optical density measurements every 30 min for 24 h at 37°C with periodic shaking in the Bioscreen C Microbiology Reader. The apparatus was placed in an air-tight chamber filled with 5% CO2 to retain the pH and bicarbonate levels of serum. For patient samples, the extent of growth was calculated from the change in OD during growth by subtracting the baseline OD from the peak level reached at the end of the growth, and was expressed as ΔOD. Viable counts were determined by plating on trypcase-soy agar plates (bioMérieux, France).

RESULTS

1 Pharmaceutical apotransferrin preparation (I)

1.1 Large scale purification of apotransferrin

In fraction IV, transferrin comprises 20-30% of the total protein and the main impurity is albumin (approx. 50%). The bulk of the albumin and other impurities were removed in the flow-through fraction of the first chromatography step, the cation exchange (Fig. 3). By using a pH-dependent elution in this chromatography, no buffer exchange was needed before the second chromatography, the anion exchange. This chromatography step was the final purification step. It was preceded by the solvent-detergent (SD) virus inactivation step and the SD chemicals were removed in this step by extensive column flushing before elution. In the eluate recovered from the anion exchange step, apotransferrin with a protein purity of at least 98% was obtained (Fig. 3). Later steps, including the diafiltration, concentration and the virus removal filtration, did not significantly increase the protein purity. In the overall process, over 90% of apotransferrin could be recovered from the dissolved fraction IV solution.

Figure 3. Flow scheme of the apotransferrin manufacturing process and SDS-PAGE purity analysis gel

By dissolving the fraction IV paste in water, the pH of the solution remained at the same level of pH 5.7 as used in the ethanol precipitation. Evidently due to the low pH and citrate buffer used for fraction IV precipitation, the iron was mostly dissociated from transferrin in the dissolved fraction IV. By adding EDTA to the solution and applying the solution to the first ion exchange at the low pH, iron was effectively removed by the wash fractions and apotransferrin with a low iron content corresponding to about 0.3% saturation could be recovered.

The specific virus inactivation and removal steps in the process comprised the SD treatment, the virus filtration and the PEG precipitation. Virus validation studies showed that the SD treatment rapidly inactivated the enveloped viruses HIV and BVDV with ≥4.7 log10 and ≥5.1 log10 reduction factors, respectively. The removal of the nonenveloped parvovirus B19 was studied in spiking experiments using PCR for virus detection. The results indicated that 3.7 log10 was removed in the PEG precipitation step and ≥4.5 log10 in the virus filtration. The virus filtration step could be carried out with high recovery at a high protein load of 4 kg/m2.

1.2 Identification and characterisation of the apotransferrin product

The first 25 amino acids in the N-terminal sequence of the purified product showed identity with human transferrin. The identity was also confirmed with mass mapping with MALDI-TOF of the peptides after cleavage with endoproteinase-LysC, and revealed over 90% of the expected apotransferrin peptides. Electrospray mass spectrometry of the protein indicated a mass of the protein of 79556, which is the same as the calculated mass of 79555.4 of the apotransferrin molecule with two biantennary disialylated glycans. The mass was calculated from the polypeptide chain (75181.4 Da) with 19 disulphide bonds (-38 Da), two glycans (2x2224 Da) and glycan linkage (-36 Da). IEX-HPLC was used to separate the transferrin into four peaks. The IEX-HPLC gave reproducible results for different transferrin batches. The proportions of the main peaks (peaks 2-4) were 7-9%, 79-83% and 9-12% respectively (Fig. 4). The masses of the peaks and the glycan structures to which they resembled closely to are presented in Fig. 4.

For more detailed glycan analysis, glycan chains were enzymatically removed from apotransferrin and characterised by MALDI-TOF mass spectrometry. Glycan analysis confirmed the presence of the glycan structures in the IEX-HPLC peaks as predicted by electrospray mass spectrometry. Glycan analysis also revealed traces of structures carrying fucose in all IEX-HPLC peaks. Minor oxidation of single amino acids were found for some of the mass structures. The mass spectrometric studies revealed no chemical modifications other than glycosylation and the presence of the predicted disulphide bonds, except minor oxidation.

The product purity and the protein integrity and conformation was studied for several production batches. The results indicated here are mean results of four batches. In SECHPLC the product was separated into a major peak of 98.5% corresponding to transferrin monomer and a minor peak, which corresponded to a molecular size of 180-290 kDa. The non-monomeric peak contained mainly IgA and IgG and a small amount of transferrin dimer. No polymers or fragments were found. The product showed a major band of 76 kDa in SDS-PAGE. The iron binding capacity was between 92% and 96% for different batches. The assay proved reproducible with imprecision CVs of 3% and 1%, inter- and intra-assay, respectively. The method was used in combination with urea-PAGE, which showed that the iron-free apotransferrin was fully converted into the iron-saturated form when iron was added. The iron content of the product was low, corresponding to approximately 0.3% transferrin saturation. No zinc or aluminium could be detected. The major plasma-derived impurities were hemopexin (0.6%), IgA (0.7%) or IgG (0.3%). Protease contaminants (plasmin/plasminogen and prekallikrein activator) were all below the detection limits of the assays.

Figure 4. Apotransferrin separated into four glycoform peaks by IEX-HPLC. The molecular weight of apotransferrin in the peaks was measured by electrospray mass spectrometry and the proposed glycan chains are shown.

The stability of the 50 g/l liquid formulation was studied. The most informative assays to detect changes in the product proved to be the iron binding capacity, the transferrin iron forms by urea-PAGE, IEX-HPLC and SEC-HPLC. The liquid product proved to be stable for up to 2 years at refrigerator (2-8°C) and 12 months at room temperature.

Lyophilised formulations were stable for four years at refrigerator temperature. The finished product was sterile, had a low endotoxin content (<2 IU/ml) and was pyrogen free according to the European Pharmacopoeia methods.

2 Binding of NTBI with apotransferrin

2.1 Evaluation of assay for measuring NTBI (II)

In the bleomycin assay, the sample and reagent volumes were halved and absorbance measurements were done in microwell plates. The bleomycin assay proved reproducible. The imprecision coefficients of variation (CVs) were 7.7% and 8.2% intraassay and 18.4% and 9.8% interassay for a low (0.2 µmol/l) and a high (1.5 µmol/l) control, respectively. To study the assay accuracy, ferric nitrilotriacetate (FeNTA) was added to saturate transferrin in serum and the full saturation point was confirmed by urea-PAGE. The NTBI was ≥0.1 µmol/l in samples with full saturation. The recovery of the added iron was only ~33% in samples with full transferrin saturation. The interference of sample haemolysis was studied. The haemolysis had a clear raising effect on NTBI, but had no effect on the transferrin saturation or on the total iron. The degree of haemolysis was studied visually and by haemoglobin determination. In samples with non-visible haemolysis (haemoglobin 290-645 mg/l) up to 0.07 µmol/l NTBI could be measured. No interference of ferritin was found when patient samples containing high ferritin concentrations were studied, nor was the treatment drug cyclophosphamide found to interfere with the assay. A detection limit of 0.1 µmol/l could be established based on accuracy studies and on the interference of non-visible haemolysis.

399 patient serum samples were studied and the NTBI levels were compared with transferrin saturation values. NTBI (≥0.1 µmol/l) was found only when transferrin saturation was >80%. When samples with visible haemolysis were excluded, there were no positive samples with <80% transferrin saturation. The NTBI varied between 0 and 0.97 µmol/l in patient samples without visible haemolysis. The highest NTBI concentration measured with the bleomycin assay in a haemolysed patient sample was 1.6 µmol/l.

The bleomycin assay was compared with the chelation method for measuring NTBI. The detection limit of the chelation assay was 1.5 µmol/l and the recovery of iron in serum was about 64%. The chelation method gave clearly higher NTBI concentrations (up to 10-14 µmol/l) than the bleomycin assay (up to 1.27 µmol/l) in 61 samples studied in parallel. There was no clear correlation between the quantitative results of the two assays in samples above the detection limit of the respective assays. Due to a low sensitivity (0.01 AU/µmol/l) of the colorimetric determination of the chelated iron, iron quantification in the ultrafiltrate was also studied by AAS. The AAS gave similar results as the colorimetric assay, and verified the higher NTBI concentration compared with the bleomycin assay. However, the chelation method gave positive results for some samples also at <80% transferrin saturation. In these samples the partial transferrin saturation was confirmed by urea-PAGE.

2.2 Binding of NTBI in patient serum (III, V)

Apotransferrin was administered to SCT patients in clinical trials. In the first trial (III), a single dose of 100 mg/kg was given to six patients three days after the SCT. Serum samples taken before and after the administration and the serum iron parameters including NTBI and urea-PAGE were studied. Initially, all patients had serum transferrin saturation >80% and NTBI in their serum. 15 min after the apotransferrin injection, serum NTBI had disappeared in all patients and transferrin saturation decreased to 30-50%. The mean increase in the serum transferrin concentration was 1.95 g/l.

Before apotransferrin administration, only fully saturated diferric transferrin was found in serum, shown by urea-PAGE. At 15 min after the injection, iron-free and monoferric forms appeared in the sera of all patients. The amount of the diferric transferrin remained about the same in five patients but was reduced in one patient. This suggested that most of the iron bound by the administered apotransferrin represented NTBI and was not shuttled from the fully saturated endogenous transferrin. At the later time points the amount of the diferric form increased, while the iron-free form disappeared first and the monoferric form remained detectable somewhat longer. This indicated that the administered apotransferrin was first converted into the monoferric form and further into the diferric form. The conversion occurred at different rates in the individual patients and coincided with the increase in the calculated transferrin saturation. The amount of iron bound by the administered apotransferrin during the first 15 min was on the average 16 µmol/l (range 11-22 µmol/l), when calculated from the amount of formed monoferric transferrin. This was much higher than the NTBI concentrations measured by the bleomycin-assay before the administration (0.3 ± 0.1µmol/l, mean ± SD ) (II).

NTBI reappeared and the transferrin saturation exceeded 80% 12-48 h after the injection in four patients and after six days in one patient. NTBI remained non-detectable for the whole 12-day follow-up period in one patient.

Apotransferrin was also given in repeated doses to twenty patients, starting six days before the SCT, before the onset of myeloablative conditioning which leads to appearance of NTBI in patient sera. Three different dosage regimens were studied during a three-week follow-up period, the total doses per patient were 0.3, 0.6 and 1 g/kg, respectively. With the repeated doses, the occurrence of NTBI could be completely avoided in 5 of 8 patients who received 1 g/kg of apotransferrin and in one patient who received 0.3 g/kg. In the other patients, the apotransferrin dosing was not sufficient to completely prevent full transferrin saturation and NTBI appearance during the follow-up period.

3 GrowthofStaphylococcus epidermidis (IV, V)

The growth of S. epidermidis in serum-free conditions in the presence of pure transferrin at different saturation levels was studied. The transferrin saturation was monitored by absorbance measurements and the transferrin iron forms by urea-PAGE. At low initial bacterial cell densities (102 - 104 CFU/ml), the growth of S. epidermidis was critically dependent on NTBI and a high transferrin saturation of at least 90%. At higher initial densities, growth was detected also in the presence of partially saturated transferrin. The siderophore induction measured as the siderophore index was highest when the inoculum was precultivated in iron-depleted media. Addition of FeNTA up to 1 µmol/l resulted in a linear decrease of the siderophore index. Induction of siderophore production during precultivation did not change the growth dependence on NTBI. All three S. epidermidis strains showed the same dependence on full transferrin saturation for growth.

In serum, a similar influence of NTBI on growth was detected. In normal serum samples with iron added to form NTBI and in patient samples with NTBI, S. epidermidis grew consistently at all inoculated (102 - 105 CFU/ml) cell concentrations. In normal serum, no growth was detected at inoculated cell densities up to 103 CFU/ml. At higher initial densities, growth was detected but with a slower rate than in the samples with NTBI. Addition of apotransferrin to bind NTBI restored the growth inhibitory effect, whereas addition of holotransferrin did not. Again, the three S. epidermidis strains behaved similarly.

The ability of S. epidermidis to grow in samples from SCT patients was studied. Using one of the multiresistant strains, bacteria (103 CFU/ml) were inoculated to serum samples and the growth was measured as the increase in OD from baseline to its maximal level, expressed as ΔOD. The sensitivity of the OD assay to detect growth in serum samples was verified by viable count measurements. The detection limit for significant growth corresponded to a ΔOD value of 0.05, which was calculated as 10 SD above the background ΔOD value of serum samples (n=18) without growth curve. The threshold density of S. epidermidis which could be detected by the OD determination was about 106 CFU/ml and thus a positive ΔOD value indicated growth of at least 3 log10 in 24 h.

The bacterial growth studied in 132 patient samples was found to have a close relation to the presence of NTBI and a high transferrin saturation level (>80%). In samples from two patients who received no apotransferrin, the bacterial growth was promoted at the same time as NTBI appeared in serum due to myeolobalative treatment. In three patients who received a single dose of apotransferrin the growth was prevented simultaneously as the NTBI was bound by transferrin. The bacteriostatic effect typically disappeared when NTBI reappeared. A patient who received repeated apotransferrin in high enough doses to prevent formation of NTBI also had restored bacteriostatic effect in serum throughout the follow-up period.

Markers for septic infections (fever >38°C, clearly elevated CRP >30 mg/l) were compared in the patients who did not receive apotransferrin or received repeated doses of apotransferrin at two different dose levels, 0.3-0.6 g/kg and 1.0 g/kg respectively. There was a trend, however not statistically significant, towards a lower incidence of suspected septic infections among the patients who received high apotransferrin doses compared with the patients who did not receive apotransferrin. The apotransferrin administrations reduced in a dose-dependent manner the number of days with detectable NTBI in the patient serum. When the total number of days with positive infection markers was evaluated as a proportion of all study days, there was a significant reduction in the number of days with fever and days with elevated CRP in the patient groups who received low and high doses of apotransferrin, respectively. When both markers were evaluated together, there was a significant reduction in the high dose group.

DISCUSSION

1 Apotransferrin product (I)

The developed manufacturing process enabled production of pure and virus-safe apotransferrin with high yield and only a few process steps. Transferrin was maintained in the iron-free apoform from the beginning of the process, and excess iron was removed in the first ion exchange step. It was therefore not necessary to have a separate iron-removal step. Apotransferrin was purified by a combination of a cation and a anion exchange chromatography. Only one ultrafiltration step was needed for concentration and buffer exchange of the pure apotransferrin solution.

The viral safety of the apotransferrin product was based on reduction of potential virus load in the starting material by donor selection and testing for infection markers and by effective virus inactivation and removal during the manufacturing. The major virus inactivation step in the process was SD treatment, which very reliably inactivates all enveloped viruses (Horowitz et al., 1993; Pehta, 1996). Virus filtration with a 15 nm pore size hollow fibre filter effectively removes even the small non-enveloped viruses, such as HAV and parvovirus (Burnouf-Radosevich et al., 1994). Parvovirus B19 currently remains a problem with plasma products because its level in the starting plasma may be very high (up to109 gen. eq./ml) and a single virus filtration step may not be capable of removing the potential virus load from the plasma pools (Siegl and Cassinotti, 1998). Although clinically significant parvovirus transmissions by plasma products are apparently very rare, parvovirus transmission might be harmful to immunocompromised patients receiving high-dose chemotherapy (Schmidt et al., 2001), who participated in this study. Therefore, the starting plasma was tested for parvovirus B 19 by PCR as a complementary measure on the overall virus safety. Both PEG precipitation and 15 nm virus filtration were implemented to the process in order to achieve two effective removal steps for parvovirus. Considering the potential level of parvovirus in the plasma pools after PCR screening and the clearance of parvovirus observed in virus removal studies of the process, it could be concluded that the finished product was with high confidence safe also with respect to parvovirus transmission.

The structural characterisation indicated that the pure apotransferrin product was identical with human plasma iron-free transferrin. Electrospray mass spectrometry did not indicate chemical modifications in the apotransferrin molecule other than the Nlinked glycans and the predicted 19 disulphide bonds, except minor oxidation most probably of single methionine residues. The identified glycan chains were structurally closely similar to the structures described before for human plasma transferrin (Fu and van Halbeek, 1992). The minor transferrin variant with only one biantennary glycan chain has not been identified in the earlier studies.

The process yielded a pharmaceutical composition of apotransferrin, which had a purity of at least 98%, contained mainly monomeric transferrin and had no detectable polymers or aggregates. Assays were developed for the assessment of the iron binding capacity and integrity of the pure apotransferrin product, which proved to be sensitive indicators of the stability of the product. Both the lyophilised formulations and the liquid formulation proved to have good stability at refrigerator temperature, four and two years respectively. The developed transferrin assays were also used to study the quality of the finished product. Results of products from several batches confirmed that the production method could be carried out with consistent results for the finished product, which is a prerequisite for a production method of a medicinal product. The formulation with sodium chloride at pH 6 was suitable for intravenous administration and the product sterility and process hygiene controls were confirmed by the pharmacopoeia methods.

Potentially harmful impurities described in other transferrin products include aggregates and polymers formed during pasteurisation (Haupt, 1993). Presence of high amounts of process derived aluminium (Rigal et al., 1989) and zinc (Rolf et al., 1997) has been reported. The product with a high aluminium content was used in a study with three patients undergoing myeloablative chemotherapy and bone marrow transplantation.

Two of the three patients developed acute renal failure, which may have been associated with the administration of the high aluminium content of the transferrin preparation.

The study did not demonstrate efficacy of the transferrin preparation in the binding of free iron in the patients or other benefits, but rather suggested that the therapeutic use of transferrin may not be safe. As with other plasma derivatives, contaminating plasma proteases present a further possible cause for adverse effects in patients. However, the level of all these plasma and process derived impurities was very low in the present apotransferrin product.

2 Determination of serum NTBI (II, III)

The bleomycin assay evaluation demonstrated that the microwell modification of the assay could be carried out reproducibly using one-half the amount of reagents and serum compared with the original assay (Evans and Halliwell, 1994), which made it possible to study large numbers of patient samples in microwell plates. Sample haemolysis interfered with the assay, and there was a clear correlation between the NTBI and the haemoglobin concentrations. An earlier report indicated that pure haemoglobin does not interfere with the assay (Gutteridge et al., 1981). It is possible, that with the mechanical haemolysis studied here other degradation products from haemoglobin were formed which could have caused the interference. However, the nonvisible haemolysis did not result in false positives when using a detection limit of 0.1 µmol/l, which was also the detection limit which was measured in the accuracy studies.

When the bleomycin assay was compared with the chelation method for measuring NTBI, it was found that the latter one gave NTBI concentrations that were clearly higher (max 14 µmol/l) than with the bleomycin assay (max 1.6 µmol/l). Similar NTBI concentrations in patient samples obtained by the bleomycin assay have been found by (Carmine et al., 1995), whereas levels of 20-28 µmol/l have been measured by others (Halliwell et al., 1988; Harrison et al., 1994). The higher total levels were found with an earlier version of the bleomycin assay (Gutteridge et al., 1981) without the final butanol extraction step, which apparently reduces the risk of possible interfering compounds in the spectrophotometric measurement. Using the chelation method the NTBI concentrations were in a similar range of 0-10 µmol/l as found by others using the same type of assay (Bradley et al., 1997; Gosriwatana et al., 1999).

It was also found that unlike the bleomycin assay in which positive samples were detected only when transferrin saturation was >80%, the chelation method detected NTBI in some serum samples also when the transferrin saturation was <80%. The use of a chelator to mobilize NTBI in serum samples evidently is a reason for the higher NTBI levels gained by the chelation-based assay. The chelator probably mobilises iron from complexes in which it is not available for the bleomycin reagent. 80 mmol/l NTA was able to mobilise iron from albumin, but also in small amounts from other proteins such as transferrin (~0.5 µmol/l) and ferritin (~0.3 µmol/l) (Gosriwatana et al., 1999). The use of the chelating agent could also explain our results of detectable NTBI in serum samples with <80% transferrin saturation, if the chelating agent had mobilised a small amount of transferrin-bound iron.

In the patients who received a single apotransferrin dose, the level of NTBI was also calculated from the amount of iron bound to the administered apotransferrin, based on the shift of apotransferrin into the monoferric form measured by urea-PAGE 15 min after the injection. The average level (16 µmol/l) represented about 50% of total serum iron in the patients, which was slightly higher but in the same order of magnitude as found by the chelation method. It is possible that the intravenously administered apotransferrin had the capacity to mobilise NTBI from complexes in the vessel wall.

Such loosely bound iron cannot be measured in serum samples. The vascular walls can theoretically provide polyanionic sites for NTBI, such as glycosaminoglycan chains of endothelial surface proteoglycans. Another speculation for the high concentration of NTBI calculated in the 15 min samples, is that the apotransferrin was able to immediately sequester low-molecular-weight iron which is released by macrophages from degrading erythrocytes (Fillet et al., 1989; Moura et al., 1998). Instead of being at least partly taken up by the transferrin independent uptake mechanism in parenchymal and especially liver cells (Craven et al., 1987; Kaplan, 2002), the iron could be directly bound by the apotransferrin which was available after the infusion.

The results of the different methods to measure NTBI indicate that the bleomycin assay does not measure the total amount of NTBI. The true concentration of NTBI in serum of the SCT patients can most likely be in the order of 10 µmol/l, even up to 20 µmol/l, as measured by the chelation method and calculated as the iron bound to injected apotransferrin in vivo. However, the bleomycin assay has high specificity and does not measure ferritin iron, which makes it a useful method to detect NTBI. The NTBI measured by the bleomycin assay is capable of binding to bleomycin and generating free radicals and is therefore considered to be a measure of the redox activity of NTBI.

The lower concentrations obtained with the bleomycin assay reflects the heterogeneous nature of NTBI being a mixture of different iron forms with different affinities for different chelators (Hider, 2002; Petrat et al., 2002). It could mean that only a portion of the loosely bound NTBI measured in serum is redox-active and capable of binding to bleomycin whereas a larger portion can be chelated by NTA.

3 Binding of NTBI by apotransferrin administration (III, V)

In the patients who received single dose injections of apotransferrin, NTBI measured by the bleomycin assay disappeared from the sera of all patients indicating that it was effectively bound by the administered apotransferrin. NTBI remained undetectable for a variable time ranging from a few hours up to several days. In most patients, NTBI reappeared 12-48 h after the apotransferrin injection. Another indication of iron binding by the administered apotransferrin was the conversion of transferrin into monoferric and diferric forms in the sequential serum samples after the apotransferrin injection.

Complete prevention of NTBI during the whole studied SCT period could be achieved with repeated administrations starting before the onset of the myeloablative treatment.

Apparently no other iron-chelating agents have been studied in SCT patients, but desferrioxamine has been extensively studied in chronic iron overload diseases such as hereditary haemochromatosis and thalassemia major. It has been shown that part of the NTBI in haemachromatosis patients is not effectively chelated with desferrioxamine (Breuer et al., 2001) and the presence of NTBI in patient sera without full transferrin saturation has been reported (Aruoma et al., 1988; Gosriwatana et al., 1999; Loreal et al., 2000). However, our results with the bleomycin assay did not indicate the presence of NTBI unless transferrin saturation exceeded 80%. The disappearance of bleomycindetectable NTBI after the apotransferrin injection indicated that the redox-active iron in the serum of SCT patients was in a form that was effectively bound by transferrin. It is possible that the underlying disease is the reason for the differences in NTBI forms found in the different patients groups. In haemochromatosis, the NTBI is formed after a slow accumulation of body iron stores over a long time period of several, even decades of years. In the haematological patients studied in this work, a halt of erythropoiesis and release of iron from senescent erythrocytes form the NTBI within hours. It may be, that the gradual formation