From laboratory to clinic: the development of an immunological reagent Susan Bright, John Adair and David Secher Access to a wide range of high quality and increasingly sophisticated reagents and equipment has underpinned the great surge of knowledge in basic immunology and the growing interest in clinical immunointervention. In this article, the first in an occasional series on immunological research and development in industry, Sue Bright and colleagues outline the key steps in a development programme to take a humanized monoclonal antibody into the clinic. The procedures involved in developing such reagents, particularly for clinical use, are long and require considerable ingenuity and scientific creativity. Genetic engineering offers a new approach to overcoming the limitations of monoclonal antibody therapy. The inherent immunogenicity of rodent monoclonals in humans might be reduced by humanization of the molecule1,2; development of a human anti-mouse antibody (HAMA) response in patients that have received murine monoclonals is a significant limitation to the use of these molecules 3,4. Pharmacokinetic properties of antibodies could be altered by adjusting the molecular form 5,6 and useful effector functions such as toxins can be added which would be beneficial once targeted to the site of action 7. Thus recombinant DNA procedures can contribute to the improvement of the clinical use of monoclonal antibodies. However, in the development of a product, molecular biology is simply the first step in a long programme, the final aim of which is a reliably manufactured molecule that has been tested for safety and efficacy and which can be routinely used by physicians. From the outset it is important to assess market requirements and the competitive environment before deciding whether development of an antibody as a therapeutic product is appropriate for a particular indication and before evaluating the alternative molecular forms. As an illustration of the time scale, the first paper on the humanization of potentially therapeutic monoclonals was published in 1986 (Ref. 8). However, it was not until 1989 that the results of the first clinical trials with such molecules were reported 9a° and the likely launch date of the first product is not before the mid-1990s. This gap reflects the time it takes to develop processes, manufacture suitable clinical material and demonstrate to the satisfaction of the regulatory authorities the safety and efficacy of the product. This review outlines the key steps in a development programme and the time required to take a humanized monoclonal antibody into the clinic. The main example used is chimeric B72.3, a recombinant antibody to a glycoprotein expressed on human adenocarcinoma cells that is" being developed as a radiolabelled product for tumour imaging and for therapy 1H3.
Molecular biology The first step is to identify exactly what kind of antibody is required as options available to the molecular biologists are varied. The answer depends on the target
indication, the likely route of administration and the predicted dose regime. The simplest type of humanized antibody to make is a chimeric molecule in which the original mouse variable region is kept intact and attached to a human constant region1,12. A refinement of this process is complementarity determining region (CDR) grafting, in which only those parts of the variable region that are thought to contribute to antigen binding are retained as mouse residues in a human framework 2a4. CDR grafting might be expected to give a less immunogenic product; however, reproducing the exact affinity of the original mouse hybridoma in the CDR-grafted molecule has proved not to be straightforward and many humanization programmes are developing chimeric products because these currently have better binding properties. Another choice that has to be made is the class and isotype of the antibody. This might very much depend on the target indication. For example, if the antibody is expected to exert its therapeutic effect by blocking or by acting as a delivery system for a cytotoxic agent then an 'inert' isotype such as IgG2 or IgG4 would be a good choice, whereas if the therapeutic role is cell damage, as in an anti-tumour antibody, then IgG1 or IgG3 may be more useful. There is also the option of developing an antibody fragment as the product, such as Fab'2, Fab or even an Fv s,6. Smaller molecules may penetrate diseased tissues more effectively, but they generally have a shorter half-life in vivo and may have to be administered more frequently. It is not the purpose of this review to describe the techniques involved in the molecular biology stage of the programme. These procedures have been well documented and the time scales for achieving the reconstruction of antibody genes to fit the tasks required have been reduced such that this area is no longer a major timeconsuming step in the overall process. B72.3, the example discussed here, is a mouse hybridoma cell line producing a mouse IgG1 monoclonal antibody u. It was decided to make the first humanized version of this antibody as an IgG4 chimeric molecule ~2a3. IgG4 was chosen principally to avoid Fc binding in vivo but also because there are no allotypes on this heavy chain, eliminating a potential source of immunogenicity.
© 1991, ElsevierSciencePublishersLtd,UK. 0167-4919/91/$02.00
Immunology Today
130
Vol. 12 No. 4 1991
Establishing a stable cell line It is possible to evaluate new genes in transient expression systems such as the COS cell system. However, to produce material in quantity, a stable cell line needs to be created. This process may take over a year before a line of the required stability, productivity and clonality is achieved. Various host systems can be used to produce stabte cell lines. Eukaryotic expression systems such as myeloma or Chinese hamster ovary (CHO) cells are suitable for whole antibody 1,13 and bacterial systems, such as the one developed using Escherichia coli that allows secretion of the product into the culture medium, are becoming more reliable and useful for the production of antibody fragments 6. Chimeric B72.3 was made from a stable CHO cell line. The heavy and light chain chimeric genes were inserted separately into an expression plasmid containing the strong promoter/enhancer transcriptional control element from the human cytomegalovirus (HCMV). In addition, a simian virus 40 (SV40) origin of replication was present, provided by the SV40 early promoter fragment, to drive a selectable marker gene, either G418 resistance (neo gene) for the light chain gene or mycophenolic acid resistance (gpt gene) for the heavy chain gene. The chimeric genes were inserted into CHO cells by electroporation; the light chain plasmid was transfected first and positive clones were selected by G418 resistance. The heavy chain plasmid was then electroporated into positive light-chain-producing clones 13. One of the most crucial aspects of this stage of the programme is selection of the most productive cell lines with which to progress further. Small differences in productivity can dramatically affect production costs and capabilities at the large scale. To choose clones well at this early stage requires accurate assay, careful records and good assessment of product secretion profiles. Hasmids that can be amplified within the cell line after transfection have been developed. These use dominant markers, for example resistance to methionine sulphoximine by the amplification of glutamine synthetase (GS), for selection is. The amplifiable option was not open to the B72.3 programme when it began and it was desirable to begin clinical studies as soon as possible. This meant that the first production line does not produce as much antibody as can now be achieved and production costs are higher than they might otherwise be - the usual trade off between time and money. After positive cell lines have been selected they are adapted to growth in suspension, to enable large scale airlift fermentation, and they are adapted to growth in low serum or serum-flee defined media, so that purification to high specification can be achieved. The lines need to be cloned, typically three times, to meet regulatory requirements for 99% clonality. They must also be investigated for stability, and production parameters need to be established 16. The end product of this process of cell line development is a master cell bank, which will last for the lifetime of the product, from which a manufacturer's working cell bank is derived 16. The cells in this bank should be stable high producers. Large-scale manufacture Recombinant antibody made from suspensionadapted CHO cells is produced in airlift fermenters Immunology Today
Fig. 1. A 20001 airlift fermenter with associatedprocess equipment (Celltech). using fed batch technology to increase productivity 17. Volumes of up to 20001 have been prepared. A 2000 l airlift fermenter with associated process equipment is shown in Fig. 1. The process involves taking a vial of cells from the working cell bank, growing the cells to an appropriate inoculum volume and inoculating the fermentation vessel. The skill in this process is not only to maintain cell viability and sterility but to monitor and control conditions such that good productivity is obtained and the best point of harvest can be achieved. Early work done in establishing stability of cell lines over the required number of generations for the complete production process and in determining production parameters is essential if a successful manufacturing process is to be established. After a process to produce material for the first clinical trials is established, further work can be undertaken to improve the process, particularly to increase productivity. However, if a process is significantly altered during development, the final product is made in a different way from material produced for the first clinical trials. If the process differences are great, some of the work may have to be repeated to satisfy regulatory requirements. Again a decision has to be made: is time or money the more important factor ? If it is time, the work is done with a process that is not perfected and the programme can be locked into less than optimal productivity. If generic high productivity fermentation processes are available, productivity can be maximized early in development.
131
Vol. 12 No. 4 1991
Establishing a purification procedure The final product from the fermenter run is supernatant from antibody-producing cells. A procedure has to be developed for purifying antibody from this supernatant to a high purity, with low levels of contaminants such as DNA and aggregated antibody 16,18,19. Development of this procedure can begin during the cell line construction stage. For antibodies, the first step in purification is often an affinity column containing the bacterial protein, protein A, which binds to the Fc region of the antibody and which can be used to achieve significant antibody purification in a single column step. Diethylaminoethyl (DEAE) Sepharose is used as an additional purification step 13. The product then has to be concentrated, formulated, sterile filtered and put into vials. The product specification is designed to ensure a safe, efficacious and consistent therapeutic product and the purification process has to be developed to do this. It is important to establish that the product is safe and that there are only residual traces of potential contaminants such as CHO DNA, retroviruses and protein A residues 16. A major activity to ensure that this occurs is the development of sensitive assays for the detection of trace quantities of contaminants in the product. These assays have to be documented and validated to ensure their reliability. Test assays for viruses and DNA are not sufficiently sensitive to assure the safety of the product and therefore studies to demonstrate the reliable removal of these contaminants are necessary. In the development of a chimeric B72.3 manufacturing process, demonstration of the appropriate clearance of DNA and retroviruses were some of the most important issues to be dealt with. Generic validation data were not sufficient. Challenge experiments were developed in which crude preparations were 'spiked' with potential contaminant and clearance determined using scaleddown model systems. The reduced-scale column systems used were representative of their full-scale equivalents in terms of relative scale and elution conditions. Clearance of retroviruses by the purification procedure was demonstrated using a spiking experiment with Moloney leukaemia virus. Lack of infectivity of any retrovirus contamination of the working cell bank was shown by plaque and focus assays for the detection of ecotropic and xenotropic retroviruses respectively. The final purified bulk product was tested and found negative for reverse transcriptase and plaque and focus assays. An observation made during the development of the purification process for chimeric B72.3 was that, although the protein could be purified to homogeneity as determined by nonreducing high pressure liquid chromatography (HPLC), when run on sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) under nonreducing conditions, a band of approximately 80 kDa could repeatedly be detected 13. This was eventually shown to be a subpopulation of the chimeric antibody molecules in which the hinge disulphide bridges had not formed. In the presence of SDS the antibody dissociates into (heavy:light) monomers giving an 80 kDa band on SDS-PAGE. To date all other chimeric IgG4 antibodies made using this constant region gene, other IgG4 myelomas and IgG4 from human serum show the
Immunology Today
same phenomenon (D. King, J. Adair et al., unpublished). This observation was unexpected and delayed establishment of a purification process while the nature of the '80kDa' band was investigated and its occurrence in natural molecules was confirmed.
Formulation Scientists need to know how the antibody will be administered to the patient as soon as possible in the development programme. What would be the most convenient form for the antibody to leave the manufacturers as a product? How will it be stored by the physician? When this is known the formulation chemists can investigate the properties of the molecule in different buffers and at different temperatures; a final formulation can be selected and long-term stability trials begun. Results with monoclonal antibodies can be surprising. Some molecules store equally well at +4°C and -20°C; some are unstable at one or other of these temperatures; some are easily freeze dried, some are not. With classical pharmaceutical compounds it is sometimes possible to perform accelerated stability studies at higher temperatures, since a direct relationship between time and temperature can be determined. With proteins, accelerated stability studies have to be used with caution, as product deterioration can obviously be quite different at higher temperatures.
Preclinical safety Any new drug, no matter how theoretically sound its mode of action, could fail to be efficacious in humans. Before marketing, efficacy has to be proved by a series of clinical trials. Before these trials can begin safety has to be established2°. Some of the in vitro safety checks that have to be made on the cell line and to validate the purification process have already been mentioned. Analytical and manufacturing methods have to be validated and formulation and stability studies adequately recorded. The aim of the scientists is to gather together all the data on the potential product and present them to the regulatory authorities who will then evaluate the information and decide whether the clinical trials can begin. To research scientists accustomed to preparing data for publication in scientific journals, the process of assembling and presenting the data to the required format for regulatory authorities can take an unexpectedly long time. Preclinical safety studies have to be done to Good Laboratory Practice (GLP). This is a way of working to standard procedures and study protocols, with quality assurance checks, which in theory could be the way that all scientists operate. In practice those used to research laboratories find GLP cumbersome and time consuming. Matching the research workers' obvious desire to get into the clinic as soon as possible with the need to make sure that the enabling studies are done in the correct way is often a difficult part of the move from the research to the development culture. GLP has to be performed, but the time taken to work in this way should not be underestimated. An important part of the preclinical data package is the animal studies. This is a time consuming and expensive element of the development programme. It can be
132
Vol. 12 No. 4 1991
the most problematical in design and interpretation of results and with antibodies there are extra problems not found with classical pharmaceuticals 2°. It is usual to present animal data showing efficacy in an in vivo model. However, not all monoclonal antibodies to human antigens also recognize an antigen found in another species in which an appropriate model of the target indication can be established. Again scientific ingenuity is often required to devise a model that can be used to convince those developing the drug and the regulatory authorities that the antibody is efficacious in vivo. With B72.3 the animal model used was a mouse xenograft model in which human antibody is introduced into nude mice bearing human colorectal turnout xenografts. Appropriate localization to the tumour was demonstrated 13. The murine form of B72.3 had already been demonstrated to locate to tumours in human patients in vivo 13 and this added to confidence that efficacy of the chimeric product might be established in humans. However, because the humanized form of B72.3 is different, some demonstration of efficacy in animals had to be achieved with this molecule. A second group of animal experiments concerns the pharmacokinetics of the administered antibody. How long will it stay in the animal? In which tissues does it persist? A very short half-life is unlikely to lead to a practical drug, particularly one that will be administered by injection, and this is a potential problem if antibody fragments are developed. Undue persistence in nontarget tissues can also be undesirable. Methods for measuring the antibody in body fluids and tissues and for looking at its in vivo metabolites have to be established so that its pharmacokinetics can be studied in animals. Pharmacokinetic studies in animals must be treated with some caution. The pattern of metabolism of large humanized molecules might be very different in nonhuman species than it is in humans. With B72.3 it was appropriate to use radiolabelled forms of the antibody for pharmacokinetic and biodistribution studies because this was the form that would be administered to patients. However, it is not necessarily appropriate to use radiolabelled antibody for animal studies when unlabelled antibody will be used in humans. In these cases, assays must be devised for detecting the administered humanized monoclonal in the blood to look at half-life, in the urine to look at clearance, and in the tissues to look for persistence. The final group of animal studies examine toxicity. Here, workers in biopharmaceuticals can often encounter an unusual problem. The aim of the toxicology programme is to look for safety problems with the new drug in more than one species. Doses are administered to test animals to such a level that some sign of toxicity must be seen. Pathological or clinical signs are demanded of a toxicology study. Safety can then be assessed by comparing the window between the dose which will be administered to humans with the dose that was necessary to demonstrate toxicity in animals. Another rationale for designing a toxicology study such that adverse signs are seen is that the most sensitive organs can be identified and clinical trials designed to closely monitor safety in those organs. With biopharmaceuticals, the difficulty can sometimes be in demonstrating any kind of toxicity, even at the Immunology Today
highest dose that it is practical to deliver. While this makes the developers of the drug feel comfortable it can sometimes lead to a clash of cultures between the classical toxicologists and the biopharmaceutical companies. With chimeric B72.3, the toxicology studies that were required were limited, first because there was a data base on the murine product in humans which indicated no toxicity due to specificity, and second because the dosing regime that would be given to the target patient population was restricted and therefore extensive toxicology was unnecessary. However, some toxicology trials had to be performed. To illustrate the point made above, in mice given five times the human dose equivalent of chimeric B72.3, no toxicological signs were seen, either clinically or histopathologically. In rabbits, there was only one clinical sign - blood glucose levels showed a marginally significant increase over control levels after administration of a high dose of B72.3. (With a classical new chemical entity such clean toxicology data would be unusual.) These were all single dose studies. Repeat dose toxicology would not be appropriate because the animals would presumably develop anti-human antibody responses, which hopefully would not occur in humans. Again, this makes it difficult to translate classical toxicology practice to biopharmaceuticals such as humanized monoclonals 2°. Finally, an important safety property of monoclonals is their specificity. Chimeric B72.3 was shown not to crossreact with an extensive panel of normal human tissues in an in vitro study 13.
To the clinic and beyond This review stops at the clinic. The first development phase is over, the scientists have collected enough data to satisfy themselves and the regulatory authorities that the drug will be safe to administer to humans. They have also begun to establish a manufacturing process, based on a stable cell line, which will yield consistent product for the duration of the clinical trials and beyond. This development programme has taken several years from the construction of the chimeric antibody genes and has required considerable scientific rigour and ingenuity at all steps along the way. The final stage, which again will take a number of years, is to establish by clinical trials that the molecule is safe in humans and is truly efficacious, if necessary comparing it with existing therapies. Then, it may be that product managers can take it over and sell it! We would like to thank our colleagues in Celltech for helpful discussions, in particular John Birch, David Broad, Chris Hill, Roger Holdsworth and Don MacLean. Susan Bright and John Adair are at Celltech Limited, 216 Bath Road, Slough SL1 4EN, UK and David Secher is at Gonville and Caius College, Cambridge, UK.
References 1 Morrison, S.L., Johnson, M.J., Herzenberg, L.A. and Oi, V.T. (1984) Proc. Natl Acad. Sci. USA 81, 6851-6855 2 Reichman, L., Clark, M., Waldmann, H. and Winter, G. (1988) Nature 332, 325-327 3 Courtney-Luck, N.S., Epenetos, A.A., Moore, R. et al. (1986) Cancer Res. 46, 6489-6493
133
rot 12 No. 4 1991
4 Goldstein, G., Fucello, A.J., Norman, D.J. et al. (1986) Transplantation 42, 507-570 5 Skerra, A. and Pluckthorn, A. (1988) Science 240, 1038-1041 6 Better, M., Chang, C.P., Robinson, R.R. and Horwitz, A.H. (1988) Science 240, 1041-1043 7 Chaudhary, V.K., Queen, C., Junghans, R.P. et al. (1989) Nature 339, 394-397 8 Sahagan, B.G., Dorai, H., Saltzgaber-Muller, J. et al. (1986) J. Immunol. 137, 1066-1074 9 LoBuglio, A.F., Wheeler, R.H., Trang, J. et al. (1989) Proc. Natl Acad. Sci. USA 86, 4220-4224 10 Begent, R.H.J., Ledermann, J.A., Bagshawe, K.D. et al. (1990) Antibody, Immunoconjugates and Radiopharmaceuticals (Vol. 3): Abstracts from the Fifth International Conference on Monoclonal Antibody Immunoconjugates for Cancer p. 86 11 Colcher, D., Horan-Hand, P., Nuti, M. and Schlom, J.
iiiii!i!i/!!i!i
(1~81) Proc. Natl Acad. Sci. USA 78, 3199-3203 12 Whittle, N., Adair, J., Lloyd, C. et al. (1987) Protein Eng. 1,499-505 13 Colcher, D., Milenic, D., Roselli, M. et al. (1989) Cancer Res. 49, 1738-1745 14 Queen, C., Schneider, W.P., Selick, H.E. et al. (1989) Proc. Natl Acad. Sci. USA 86, 10029-10033 15 Cockett, M.L, Bebbington, C.R. and Yarranton, G.T. (1990) Biotechnology 8,662-667 16 Committee for Proprietary Medical Products (1987) Trends Biotechnol. 5, G1-G4 17 Rhodes, M. and Birch, J. (1988) Biotechnology 6, 518-523 18 Duncan, M.E., Charlesworth, F.A. and Griffin, J.P. (1987) Trends Biotechnol. 5,325-328 19 Sherwood, R. (1988) Trends Biotechnol. 6, 135-136 20 Committee for Proprietary Medical Products (1989) Trends Biotechnol. 7, G13-G16
!!ii !3iliii!i!i!i!iiiiii!ii! U! !ii!0
Is selective IgA deficiency associated with central HLA genes or alleles of the DR-DQ region? M.
French
and
R. Dawkins
(Immunol. Today, 1990, 11, 271-
274) proposed that products of the central MHC genes might be involved in the susceptibility to selective IgA deficiency (IgA-D) and various autoimmune diseases associated with IgA-D. Our findings, summarized below, do not support this hypothesis. We have recently described positive associations with three D R - D Q haplotypes: DR1,DQw5; DR7,DQw2; and DRw17,DQw2, as well as a strong negative association with DRw15,DQw6 in IgA-D individuals 1. When looking for a shared feature between the observed D R - D Q associations, we observed that the DQf3 chains of the three 'susceptibility' haplotypes all had a neutral alanine or valine at position 57 of the DQJ3 chain. The 'protective' allele had the negatively charged aspartic acid at this position. Thus, selective IgA-D was shown to be one of the as yet rare HLA-associated diseases that are related to specific amino acids or epitopes of the major histocompatibility complex (MHC) class II molecules. In studies on patients with common variable immunodeficiency or combined IgA-IgG2 deficiency, we have found similar associations
!!iiiiii!iiiiiiii!i!iiii!ii!i!!i!ililiiii!i i!i !iiiii!!iilil!iiii!iiiiiiiili...
(Ref. 2 and L. Hammarstr6m et al., unpublished). In a study of patients with IgA-D or common variable immunodeficiency, Schaffer and co-workers 3 found that three genetic events/ markers of the HLA class III region were shared by the two immunodeficiency disorders: deletion of the C4A gene, deletion of the 21hydroxylase A pseudogene, and rare C2 restriction fragments. We have investigated the occurrence of C4A deletions in a large group of individuals with selective IgA-D (n = 95). IgA-D was found to be strongly positively associated with C4A deletion (P
Molecular biology The first step is to identify exactly what kind of antibody is required as options available to the molecular biologists are varied. The answer depends on the target
indication, the likely route of administration and the predicted dose regime. The simplest type of humanized antibody to make is a chimeric molecule in which the original mouse variable region is kept intact and attached to a human constant region1,12. A refinement of this process is complementarity determining region (CDR) grafting, in which only those parts of the variable region that are thought to contribute to antigen binding are retained as mouse residues in a human framework 2a4. CDR grafting might be expected to give a less immunogenic product; however, reproducing the exact affinity of the original mouse hybridoma in the CDR-grafted molecule has proved not to be straightforward and many humanization programmes are developing chimeric products because these currently have better binding properties. Another choice that has to be made is the class and isotype of the antibody. This might very much depend on the target indication. For example, if the antibody is expected to exert its therapeutic effect by blocking or by acting as a delivery system for a cytotoxic agent then an 'inert' isotype such as IgG2 or IgG4 would be a good choice, whereas if the therapeutic role is cell damage, as in an anti-tumour antibody, then IgG1 or IgG3 may be more useful. There is also the option of developing an antibody fragment as the product, such as Fab'2, Fab or even an Fv s,6. Smaller molecules may penetrate diseased tissues more effectively, but they generally have a shorter half-life in vivo and may have to be administered more frequently. It is not the purpose of this review to describe the techniques involved in the molecular biology stage of the programme. These procedures have been well documented and the time scales for achieving the reconstruction of antibody genes to fit the tasks required have been reduced such that this area is no longer a major timeconsuming step in the overall process. B72.3, the example discussed here, is a mouse hybridoma cell line producing a mouse IgG1 monoclonal antibody u. It was decided to make the first humanized version of this antibody as an IgG4 chimeric molecule ~2a3. IgG4 was chosen principally to avoid Fc binding in vivo but also because there are no allotypes on this heavy chain, eliminating a potential source of immunogenicity.
© 1991, ElsevierSciencePublishersLtd,UK. 0167-4919/91/$02.00
Immunology Today
130
Vol. 12 No. 4 1991
Establishing a stable cell line It is possible to evaluate new genes in transient expression systems such as the COS cell system. However, to produce material in quantity, a stable cell line needs to be created. This process may take over a year before a line of the required stability, productivity and clonality is achieved. Various host systems can be used to produce stabte cell lines. Eukaryotic expression systems such as myeloma or Chinese hamster ovary (CHO) cells are suitable for whole antibody 1,13 and bacterial systems, such as the one developed using Escherichia coli that allows secretion of the product into the culture medium, are becoming more reliable and useful for the production of antibody fragments 6. Chimeric B72.3 was made from a stable CHO cell line. The heavy and light chain chimeric genes were inserted separately into an expression plasmid containing the strong promoter/enhancer transcriptional control element from the human cytomegalovirus (HCMV). In addition, a simian virus 40 (SV40) origin of replication was present, provided by the SV40 early promoter fragment, to drive a selectable marker gene, either G418 resistance (neo gene) for the light chain gene or mycophenolic acid resistance (gpt gene) for the heavy chain gene. The chimeric genes were inserted into CHO cells by electroporation; the light chain plasmid was transfected first and positive clones were selected by G418 resistance. The heavy chain plasmid was then electroporated into positive light-chain-producing clones 13. One of the most crucial aspects of this stage of the programme is selection of the most productive cell lines with which to progress further. Small differences in productivity can dramatically affect production costs and capabilities at the large scale. To choose clones well at this early stage requires accurate assay, careful records and good assessment of product secretion profiles. Hasmids that can be amplified within the cell line after transfection have been developed. These use dominant markers, for example resistance to methionine sulphoximine by the amplification of glutamine synthetase (GS), for selection is. The amplifiable option was not open to the B72.3 programme when it began and it was desirable to begin clinical studies as soon as possible. This meant that the first production line does not produce as much antibody as can now be achieved and production costs are higher than they might otherwise be - the usual trade off between time and money. After positive cell lines have been selected they are adapted to growth in suspension, to enable large scale airlift fermentation, and they are adapted to growth in low serum or serum-flee defined media, so that purification to high specification can be achieved. The lines need to be cloned, typically three times, to meet regulatory requirements for 99% clonality. They must also be investigated for stability, and production parameters need to be established 16. The end product of this process of cell line development is a master cell bank, which will last for the lifetime of the product, from which a manufacturer's working cell bank is derived 16. The cells in this bank should be stable high producers. Large-scale manufacture Recombinant antibody made from suspensionadapted CHO cells is produced in airlift fermenters Immunology Today
Fig. 1. A 20001 airlift fermenter with associatedprocess equipment (Celltech). using fed batch technology to increase productivity 17. Volumes of up to 20001 have been prepared. A 2000 l airlift fermenter with associated process equipment is shown in Fig. 1. The process involves taking a vial of cells from the working cell bank, growing the cells to an appropriate inoculum volume and inoculating the fermentation vessel. The skill in this process is not only to maintain cell viability and sterility but to monitor and control conditions such that good productivity is obtained and the best point of harvest can be achieved. Early work done in establishing stability of cell lines over the required number of generations for the complete production process and in determining production parameters is essential if a successful manufacturing process is to be established. After a process to produce material for the first clinical trials is established, further work can be undertaken to improve the process, particularly to increase productivity. However, if a process is significantly altered during development, the final product is made in a different way from material produced for the first clinical trials. If the process differences are great, some of the work may have to be repeated to satisfy regulatory requirements. Again a decision has to be made: is time or money the more important factor ? If it is time, the work is done with a process that is not perfected and the programme can be locked into less than optimal productivity. If generic high productivity fermentation processes are available, productivity can be maximized early in development.
131
Vol. 12 No. 4 1991
Establishing a purification procedure The final product from the fermenter run is supernatant from antibody-producing cells. A procedure has to be developed for purifying antibody from this supernatant to a high purity, with low levels of contaminants such as DNA and aggregated antibody 16,18,19. Development of this procedure can begin during the cell line construction stage. For antibodies, the first step in purification is often an affinity column containing the bacterial protein, protein A, which binds to the Fc region of the antibody and which can be used to achieve significant antibody purification in a single column step. Diethylaminoethyl (DEAE) Sepharose is used as an additional purification step 13. The product then has to be concentrated, formulated, sterile filtered and put into vials. The product specification is designed to ensure a safe, efficacious and consistent therapeutic product and the purification process has to be developed to do this. It is important to establish that the product is safe and that there are only residual traces of potential contaminants such as CHO DNA, retroviruses and protein A residues 16. A major activity to ensure that this occurs is the development of sensitive assays for the detection of trace quantities of contaminants in the product. These assays have to be documented and validated to ensure their reliability. Test assays for viruses and DNA are not sufficiently sensitive to assure the safety of the product and therefore studies to demonstrate the reliable removal of these contaminants are necessary. In the development of a chimeric B72.3 manufacturing process, demonstration of the appropriate clearance of DNA and retroviruses were some of the most important issues to be dealt with. Generic validation data were not sufficient. Challenge experiments were developed in which crude preparations were 'spiked' with potential contaminant and clearance determined using scaleddown model systems. The reduced-scale column systems used were representative of their full-scale equivalents in terms of relative scale and elution conditions. Clearance of retroviruses by the purification procedure was demonstrated using a spiking experiment with Moloney leukaemia virus. Lack of infectivity of any retrovirus contamination of the working cell bank was shown by plaque and focus assays for the detection of ecotropic and xenotropic retroviruses respectively. The final purified bulk product was tested and found negative for reverse transcriptase and plaque and focus assays. An observation made during the development of the purification process for chimeric B72.3 was that, although the protein could be purified to homogeneity as determined by nonreducing high pressure liquid chromatography (HPLC), when run on sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) under nonreducing conditions, a band of approximately 80 kDa could repeatedly be detected 13. This was eventually shown to be a subpopulation of the chimeric antibody molecules in which the hinge disulphide bridges had not formed. In the presence of SDS the antibody dissociates into (heavy:light) monomers giving an 80 kDa band on SDS-PAGE. To date all other chimeric IgG4 antibodies made using this constant region gene, other IgG4 myelomas and IgG4 from human serum show the
Immunology Today
same phenomenon (D. King, J. Adair et al., unpublished). This observation was unexpected and delayed establishment of a purification process while the nature of the '80kDa' band was investigated and its occurrence in natural molecules was confirmed.
Formulation Scientists need to know how the antibody will be administered to the patient as soon as possible in the development programme. What would be the most convenient form for the antibody to leave the manufacturers as a product? How will it be stored by the physician? When this is known the formulation chemists can investigate the properties of the molecule in different buffers and at different temperatures; a final formulation can be selected and long-term stability trials begun. Results with monoclonal antibodies can be surprising. Some molecules store equally well at +4°C and -20°C; some are unstable at one or other of these temperatures; some are easily freeze dried, some are not. With classical pharmaceutical compounds it is sometimes possible to perform accelerated stability studies at higher temperatures, since a direct relationship between time and temperature can be determined. With proteins, accelerated stability studies have to be used with caution, as product deterioration can obviously be quite different at higher temperatures.
Preclinical safety Any new drug, no matter how theoretically sound its mode of action, could fail to be efficacious in humans. Before marketing, efficacy has to be proved by a series of clinical trials. Before these trials can begin safety has to be established2°. Some of the in vitro safety checks that have to be made on the cell line and to validate the purification process have already been mentioned. Analytical and manufacturing methods have to be validated and formulation and stability studies adequately recorded. The aim of the scientists is to gather together all the data on the potential product and present them to the regulatory authorities who will then evaluate the information and decide whether the clinical trials can begin. To research scientists accustomed to preparing data for publication in scientific journals, the process of assembling and presenting the data to the required format for regulatory authorities can take an unexpectedly long time. Preclinical safety studies have to be done to Good Laboratory Practice (GLP). This is a way of working to standard procedures and study protocols, with quality assurance checks, which in theory could be the way that all scientists operate. In practice those used to research laboratories find GLP cumbersome and time consuming. Matching the research workers' obvious desire to get into the clinic as soon as possible with the need to make sure that the enabling studies are done in the correct way is often a difficult part of the move from the research to the development culture. GLP has to be performed, but the time taken to work in this way should not be underestimated. An important part of the preclinical data package is the animal studies. This is a time consuming and expensive element of the development programme. It can be
132
Vol. 12 No. 4 1991
the most problematical in design and interpretation of results and with antibodies there are extra problems not found with classical pharmaceuticals 2°. It is usual to present animal data showing efficacy in an in vivo model. However, not all monoclonal antibodies to human antigens also recognize an antigen found in another species in which an appropriate model of the target indication can be established. Again scientific ingenuity is often required to devise a model that can be used to convince those developing the drug and the regulatory authorities that the antibody is efficacious in vivo. With B72.3 the animal model used was a mouse xenograft model in which human antibody is introduced into nude mice bearing human colorectal turnout xenografts. Appropriate localization to the tumour was demonstrated 13. The murine form of B72.3 had already been demonstrated to locate to tumours in human patients in vivo 13 and this added to confidence that efficacy of the chimeric product might be established in humans. However, because the humanized form of B72.3 is different, some demonstration of efficacy in animals had to be achieved with this molecule. A second group of animal experiments concerns the pharmacokinetics of the administered antibody. How long will it stay in the animal? In which tissues does it persist? A very short half-life is unlikely to lead to a practical drug, particularly one that will be administered by injection, and this is a potential problem if antibody fragments are developed. Undue persistence in nontarget tissues can also be undesirable. Methods for measuring the antibody in body fluids and tissues and for looking at its in vivo metabolites have to be established so that its pharmacokinetics can be studied in animals. Pharmacokinetic studies in animals must be treated with some caution. The pattern of metabolism of large humanized molecules might be very different in nonhuman species than it is in humans. With B72.3 it was appropriate to use radiolabelled forms of the antibody for pharmacokinetic and biodistribution studies because this was the form that would be administered to patients. However, it is not necessarily appropriate to use radiolabelled antibody for animal studies when unlabelled antibody will be used in humans. In these cases, assays must be devised for detecting the administered humanized monoclonal in the blood to look at half-life, in the urine to look at clearance, and in the tissues to look for persistence. The final group of animal studies examine toxicity. Here, workers in biopharmaceuticals can often encounter an unusual problem. The aim of the toxicology programme is to look for safety problems with the new drug in more than one species. Doses are administered to test animals to such a level that some sign of toxicity must be seen. Pathological or clinical signs are demanded of a toxicology study. Safety can then be assessed by comparing the window between the dose which will be administered to humans with the dose that was necessary to demonstrate toxicity in animals. Another rationale for designing a toxicology study such that adverse signs are seen is that the most sensitive organs can be identified and clinical trials designed to closely monitor safety in those organs. With biopharmaceuticals, the difficulty can sometimes be in demonstrating any kind of toxicity, even at the Immunology Today
highest dose that it is practical to deliver. While this makes the developers of the drug feel comfortable it can sometimes lead to a clash of cultures between the classical toxicologists and the biopharmaceutical companies. With chimeric B72.3, the toxicology studies that were required were limited, first because there was a data base on the murine product in humans which indicated no toxicity due to specificity, and second because the dosing regime that would be given to the target patient population was restricted and therefore extensive toxicology was unnecessary. However, some toxicology trials had to be performed. To illustrate the point made above, in mice given five times the human dose equivalent of chimeric B72.3, no toxicological signs were seen, either clinically or histopathologically. In rabbits, there was only one clinical sign - blood glucose levels showed a marginally significant increase over control levels after administration of a high dose of B72.3. (With a classical new chemical entity such clean toxicology data would be unusual.) These were all single dose studies. Repeat dose toxicology would not be appropriate because the animals would presumably develop anti-human antibody responses, which hopefully would not occur in humans. Again, this makes it difficult to translate classical toxicology practice to biopharmaceuticals such as humanized monoclonals 2°. Finally, an important safety property of monoclonals is their specificity. Chimeric B72.3 was shown not to crossreact with an extensive panel of normal human tissues in an in vitro study 13.
To the clinic and beyond This review stops at the clinic. The first development phase is over, the scientists have collected enough data to satisfy themselves and the regulatory authorities that the drug will be safe to administer to humans. They have also begun to establish a manufacturing process, based on a stable cell line, which will yield consistent product for the duration of the clinical trials and beyond. This development programme has taken several years from the construction of the chimeric antibody genes and has required considerable scientific rigour and ingenuity at all steps along the way. The final stage, which again will take a number of years, is to establish by clinical trials that the molecule is safe in humans and is truly efficacious, if necessary comparing it with existing therapies. Then, it may be that product managers can take it over and sell it! We would like to thank our colleagues in Celltech for helpful discussions, in particular John Birch, David Broad, Chris Hill, Roger Holdsworth and Don MacLean. Susan Bright and John Adair are at Celltech Limited, 216 Bath Road, Slough SL1 4EN, UK and David Secher is at Gonville and Caius College, Cambridge, UK.
References 1 Morrison, S.L., Johnson, M.J., Herzenberg, L.A. and Oi, V.T. (1984) Proc. Natl Acad. Sci. USA 81, 6851-6855 2 Reichman, L., Clark, M., Waldmann, H. and Winter, G. (1988) Nature 332, 325-327 3 Courtney-Luck, N.S., Epenetos, A.A., Moore, R. et al. (1986) Cancer Res. 46, 6489-6493
133
rot 12 No. 4 1991
4 Goldstein, G., Fucello, A.J., Norman, D.J. et al. (1986) Transplantation 42, 507-570 5 Skerra, A. and Pluckthorn, A. (1988) Science 240, 1038-1041 6 Better, M., Chang, C.P., Robinson, R.R. and Horwitz, A.H. (1988) Science 240, 1041-1043 7 Chaudhary, V.K., Queen, C., Junghans, R.P. et al. (1989) Nature 339, 394-397 8 Sahagan, B.G., Dorai, H., Saltzgaber-Muller, J. et al. (1986) J. Immunol. 137, 1066-1074 9 LoBuglio, A.F., Wheeler, R.H., Trang, J. et al. (1989) Proc. Natl Acad. Sci. USA 86, 4220-4224 10 Begent, R.H.J., Ledermann, J.A., Bagshawe, K.D. et al. (1990) Antibody, Immunoconjugates and Radiopharmaceuticals (Vol. 3): Abstracts from the Fifth International Conference on Monoclonal Antibody Immunoconjugates for Cancer p. 86 11 Colcher, D., Horan-Hand, P., Nuti, M. and Schlom, J.
iiiii!i!i/!!i!i
(1~81) Proc. Natl Acad. Sci. USA 78, 3199-3203 12 Whittle, N., Adair, J., Lloyd, C. et al. (1987) Protein Eng. 1,499-505 13 Colcher, D., Milenic, D., Roselli, M. et al. (1989) Cancer Res. 49, 1738-1745 14 Queen, C., Schneider, W.P., Selick, H.E. et al. (1989) Proc. Natl Acad. Sci. USA 86, 10029-10033 15 Cockett, M.L, Bebbington, C.R. and Yarranton, G.T. (1990) Biotechnology 8,662-667 16 Committee for Proprietary Medical Products (1987) Trends Biotechnol. 5, G1-G4 17 Rhodes, M. and Birch, J. (1988) Biotechnology 6, 518-523 18 Duncan, M.E., Charlesworth, F.A. and Griffin, J.P. (1987) Trends Biotechnol. 5,325-328 19 Sherwood, R. (1988) Trends Biotechnol. 6, 135-136 20 Committee for Proprietary Medical Products (1989) Trends Biotechnol. 7, G13-G16
!!ii !3iliii!i!i!i!iiiiii!ii! U! !ii!0
Is selective IgA deficiency associated with central HLA genes or alleles of the DR-DQ region? M.
French
and
R. Dawkins
(Immunol. Today, 1990, 11, 271-
274) proposed that products of the central MHC genes might be involved in the susceptibility to selective IgA deficiency (IgA-D) and various autoimmune diseases associated with IgA-D. Our findings, summarized below, do not support this hypothesis. We have recently described positive associations with three D R - D Q haplotypes: DR1,DQw5; DR7,DQw2; and DRw17,DQw2, as well as a strong negative association with DRw15,DQw6 in IgA-D individuals 1. When looking for a shared feature between the observed D R - D Q associations, we observed that the DQf3 chains of the three 'susceptibility' haplotypes all had a neutral alanine or valine at position 57 of the DQJ3 chain. The 'protective' allele had the negatively charged aspartic acid at this position. Thus, selective IgA-D was shown to be one of the as yet rare HLA-associated diseases that are related to specific amino acids or epitopes of the major histocompatibility complex (MHC) class II molecules. In studies on patients with common variable immunodeficiency or combined IgA-IgG2 deficiency, we have found similar associations
!!iiiiii!iiiiiiii!i!iiii!ii!i!!i!ililiiii!i i!i !iiiii!!iilil!iiii!iiiiiiiili...
(Ref. 2 and L. Hammarstr6m et al., unpublished). In a study of patients with IgA-D or common variable immunodeficiency, Schaffer and co-workers 3 found that three genetic events/ markers of the HLA class III region were shared by the two immunodeficiency disorders: deletion of the C4A gene, deletion of the 21hydroxylase A pseudogene, and rare C2 restriction fragments. We have investigated the occurrence of C4A deletions in a large group of individuals with selective IgA-D (n = 95). IgA-D was found to be strongly positively associated with C4A deletion (P
