TiPS - May 1993 [Vol. 141
`
`139
`
`to be created
`rodent monoclonal antibodies
`enabled
`technology
`Hybridoma
`against human pathogens
`and cells, but these had limited clinical utility.
`reviewed here by Greg Winter and William Harris, is
`Protein engineering,
`now generating
`antibodies
`for treatment of infectious disease, autoimmune
`disease and cancer by ‘humanizing’ rodent antibodies. Humanized antibodies
`have
`improved pharmacokinetics,
`reduced
`immunogenicity
`and have been
`used to clinical advantage.
`
`gens that are difficult to elicit from
`a human immune response’.
`The engineering of antibodies
`facilitated
`by
`the modular
`
`is
`
`arrangement of protein domains:
`the heavy- and light-chain variable
`(V) domains are responsible
`for
`binding
`to antigen, and the con-
`stant domains
`to effector
`func-
`tions. As both complement and
`cell-mediated killing require fully
`glycosylated antibody,
`the engin-
`eered mAbs are expressed
`in
`mammalian hosts. Each antibody
`domain is encoded by a different
`genetic exon and, to build recom-
`binant antibodies,
`the exons are
`pasted
`together. The exons en-
`coding
`the variable domains
`(V
`genes) can be cloned
`from the
`genomic DNA of a B-cell hybrid-
`oma: more conveniently,
`the V
`genes of hybridomas are isolated
`
`is an adaptor mol-
`The antibody
`ecule containing binding sites for
`antigen at one end and for effector
`molecules at the other, and has
`evolved to bind to a vast range of
`antigens. Binding alone may be
`some
`sufficient
`to neutralize
`toxins and viruses, however, more
`commonly,
`the antibody
`triggers
`the complement
`system and cell-
`mediated killing. Although anti-
`bodies
`are natural
`therapeutic
`agents,
`it has proved difficult
`to
`make human monodonal
`anti-
`bodies
`(mAbs)
`by hybridoma
`technology.
`unfortunately
`Rodent mAbs
`have
`serious
`disadvantages:
`a
`short half-life in serum; only some
`of the different classes can trigger
`human effector functions; and the
`mAbs can also elicit an unwanted
`immune
`response
`in patients
`(human anti-mouse antibodies or
`HAMA). HAMA can result in en-
`hanced clearance of the antibody
`from the serum, blocking of its
`therapeutic effect and hypersen-
`sitivity reactions. These problems
`have prompted the use of protein
`technologies
`to
`engineering
`‘humanize’ rodent mAbs by trans-
`planting
`antigen-binding
`sites
`from rodent to human antibodies.
`In principle, humanizing
`allows
`access
`to a large pool of well-
`characterized
`rodent mAbs
`for
`including
`those with
`therapy,
`specificities
`against human anti-
`
`of
`is at the MRC Laboratory
`Winter
`Molecular Biology and is Deputy Director
`at
`the MRC Centrefor Protein Engineering, Hills
`Road, Cambridge, UK CB2 2QH; W. Harris
`is Professor of Genetics at the Department
`of Molecular
`and Cell Biology, Marischal
`College, University
`Aberdeen, Aberdeen,
`UK A69 IAS.
`
`5’ . . . . . . 3’ 5’ . . . . . . 3’ lg mRNA template Ig mRNA template PCR amplification of V genes ‘, 1. . . VH ,I.,’ . . . . VI_ I,’ . . ,’ ‘. .’ L. _’ Cloning V genes “*I... _,,-+” expression vector for heavy chains expression - vector for light chains
`
`heavy-
`mouse
`from the mRNA of
`chain V genes
`Fig. 7. Cloning
`(human) genes encoding constant domains,
`into vectors comprising
`B-cell hybridoma
`for expression of mouse humen chimaeric antibodies
`
`in mammalian cells. @ 1993. Elsevicr Sciencs Pubbsherr Ltd (UK) 0165 - 6147/‘WS06.00
`
`Pfizer v. Genentech
`IPR2017-01488
`Genentech Exhibit 2055
`
`Greg Winter and William J. Harris
`G.
`of
`of
`(VH) and light- (VL)
`a
`

`

`Fig. 2. The @sheet framework structure of heavy- (VH) and lighf- (VA) chain variable domains with hyperwariable loops 1-6. Reproduced, with permission, from Winter, G. and Milstein. C. (1991) Nafure 349, 239-299. from the mRNA by use of the polymerase chain reaction (PCR). The V genes are readily linked to those exons encoding constant domains for expression of mAbs2 (Fig. 1). Expression vectors have been built with both antibody and viral promoters and enhancers, with V and C genes as different exons (‘genomic’) or linked together (‘cDNA’). Different markers for selection of trans- formed cells are available and in both myeloma and CHO hosts3,4, mAb expression is greatly en- hanced by amplifying the number of integrated copies, resulting in yields of up to 0.7 gl-’ in fer- menters5. Building humanized antibodies The first generation of human- ized antibodies were simple chimaeric mAbs, in which the variable domains of a rodent mAb are transplanted to the constant domains of human antibodies (Fig. 1). This reduces the immuno- genic@ of the rodent mAb (see below) and allows the effector functions to be selected for the therapeutic application. Thus, the human yl isotype appears to be the most effective for complement and cell-mediated killing, while the human yg4 isotype appears more suitable for imaging and blocking’. The second generation of humanized antibodies were the so-called CDR-grafted antibodies, in which the antigen-binding loops of the rodent mAb were built into a human antibody. The architecture of each antibody V-domain con- sists of a B-sheet ‘sandwich’ sur- mounted by antigen-binding loops (complementarity deter- mining regions or CDRs): in dif- ferent antibodies, these CDR loops are hypervariable in sequence (Fig. 2). It is this hyper- variability that allows the anti- body repertoire to bind a poten- tially vast array of antigens. By transplanting (or grafting) the CDRs from rodent mAb to human antibody, the antigen-binding site can also be transferred’; indeed the same human frame- work can be used for mounting different antigen-binding sitesy-9. However, to recreate the antigen- binding site it is also necessary to consider other possible inter- actions between the B-sheet framework and the loops. With the help of molecular modelling it TiPS - May 1993 [Vol. 141 is possible to design framework substitutions that maintain key contacts with the CDR loops. For example, with the rat anti- body CAMPATH- directed against the CDw52 antigen of human lymphocytes, the simple grafting of the CDRs failed to transplant the binding activity to a human antibody. When the three-dimensional folding of the VH-CDRl loop of the rat antibody and its contacts with the rat framework were modelled by computer graphics, the framework amino acid residue Phe27 was predicted to pack against the loop. However, in the human frame- work of the CDR-grafted anti- body, Phe27 was replaced by Ser27; indeed when Ser27 was mutated to Phe in the CDR- grafted antibody, the binding activity was restored’. In other examples, enhancement of anti- gen affinity was achieved step- wise by combining several frame- work substitutionsiO,“. Indeed analysis of antibody structures is leading towards the identification of sets of framework residues that may exert an influence on CDR structure” and also on the packing of the strands of the B- sheet12. The first CDR-grafted anti- bodies were based on the known crystallographic structures of the human myeloma proteins7-9. CDR-grafted antibodies have also been built with consensus human frameworks based on several human heavy chains13. The use of a single or a limited number of human frameworks offers the prospect of a range of therapeutic antibodies that are almost ident- ical, apart from the CDR sequences. Conversely a range of framework structures should be capable of supporting the CDR loops, and ‘hyperchimaeric’ CDR-grafted antibodies have used mouse- human frameworks. For example, to humanize a mouse antibody directed against the human IL-2 receptor (anti-Tat), a human framework sequence was selected by homology. A molecular model was then used to identify those framework residues of the rodent antibody that might interact with the antigen-binding loops, and these were built into the selected human frameworkI (Fig. 3). Chimaeric frameworks have also been proposed in which the
`
`

`

`TiPS - May 1993 IVol. 141
`
`internal residues that pack be- tween the domains and with the antigen-binding loops are derived from the rodent sequence and the solvent-accessible residues are taken from a human sequencei5. Most generally, all of the rodent CDRs are transplanted from mouse to human antibody. How- ever, some CDRs are more im- portant than others for binding of antigen, as evident from the crys- tallographic structures of anti- body-antigen complexes. The interaction of antibody loops with antigen involves both main-chain and sidechain contacts: as the CDR loops of mouse and human antibodies fold in a limited number of ways16, it is possible to main- tain some main-chain contacts while varying some of the side- chains (and sequence) of the CDRsr7. There is some loss in binding affinity on CDR grafting but, in combination with some frame- work alterations, it is usually possible to obtain a reshaped antibody with an affinity within three-fold of the parent mono- clonal antibody. High binding affinities may be critical for neutralization of a cytokine or toxin in the serum; they appear to be less important where multiple interactions can occur with high avidities, as with multimeric (cell surface) antibody binding to repeated epitopes on a viral coat’. However, small improvements in affinity have been seen for some CDR-grafted antibodiesls, and binding affinities can also be im- proved
`Both chimaeric and CDR- grafted antibodies appear to have better pharmacokinetics than rodent mAbs, with extended serum half-life (>75 hours) in humans and cynomolgus monkeys. Likewise the immunogenicity is reduced. Much of the HAMA re- sponse to mouse antibodies in patients is directed against the constant region: chimaeric anti- bodies and CDR-grafted antibodies appear to elicit much less response with the immunogenic epitopes being located in the variable regions. Indeed much of this re- sponse is directed against the anti- gen-binding site (for review see Ref. 4). Fig. 3. Computer model of the main-chain backbone of the humanized anti-Tat antibody. Red, complementarily-determining regions; blue, altered framework residues. Reproduced, by kind permission of L. Korn, from the prospectus of Protein Design Laboratories. No antibody response was detected against the CDR-grafted anti-CDw52 antibody during the treatment of two patients with non-Hodgkin’s lymphoma for up to 43 days with escalating doses of antibody ranging from 1 to 20 mg per day**. Also, no response was reported when the antibody was used in a single course of therapy for rheumatoid arthritis patients22 or in conjunction with an anti- CD4 antibody in treatment of a patient with intractable systemic vasculitis23. However, an anti- body response was detected on further treatment of the rheuma- toid patients22. Antibody re- sponses were not detected against other CDR-grafted antibodies used for radio-imaging in tumour patients24 or treating acute graft- versus-host disease25. eluding tumour cell antigen: S. Some of the targets are summar- ized in Table I. CDR-grafted anti- bodies against lymphocyte markers have already been used to clinical advantage. The anti- CDw52 antibody was used to deplete a large tumour mass in two lymphoma patients, and to achieve clinical remission2’. Like- wise, this antibody resulted in significant clinical benefit for a patient with systemic vasculitisz3, and for rheumatoid arthritis patients2*. The anti-Tat antibody was used for immunosuppression following allogeneic marrow transplantation, and resulted in improvement in several cases of acute graft-versus-host disease*‘. Future
`
`As shown above, humanized antibodies can be engineered from rodent mAbs. Their use has been demonstrated in the clinic, and they have a longer serum half- life and reduced immunogenicity Chimaeric and CDR-grafted antibodies have been constructed against a wide range of viral and bacterial pathogens, and against human cell-surface markers in-
`
`prospects
`
`in vitro,
`by chain shuffling”, or random mutation”‘.
`
`Using humanized antibodies
`
`

`

`142 TABLE I. CDR-grafted antibodies for therapy (see also Ref. 3) Tarcret Clfnfcal potential CDw52 CD3 CD4 IL-2 receptor Tumour necrosis factor f3 Human immunodeficiency virus Rous sarcoma virus Herpes simplex virus Lewis-Y p185HER‘2 Placental alkaline phosphatase Carcfnoembryonic antigen lymphomas. systemic vasculitis. rheumatoid arthritis organ transplantation organ transplants, rheumatoid arthritis, Crohn’s disease leukaemias and lymphomas, orgail transplants, graft-versus-host disease septic shock AIDS respiratory syncytial virus infection neonatal, ocular and genital herpes infection cancer cancer cancer cancer compared to rodent mAbs. They therefore appear to be promising as therapeutics, especially for a single course of treatment. It is not yet clear whether humanized (or even human) antibodies will elicit a blocking immune response over longer or several courses of treat- ment. The immunogenic@ of human- ized antibodies is likely to depend on several factors, including the immune state of the patients, and the dose and regimen of antibody administration. The target is also likely to be important; antibodies directed against cell-bound anti- gens might be expected to be more immunogenic than those binding to soluble antigen. Furthermore, since foreign frame- work regions can elicit an immune responsez6, we might also expect that the differing strategies to select and mutate human frame- works (as described above) could lead to reshaped antibodies with differing immunogenicity. For ex- ample, human antibody frame- works that are mutated (in viva or in vitro) with respect to the germ-line segments could prove immunogenic: even buried resi- dues could form the critical element of a T-cell epitope if presented as a denatured peptide by a class II MHC moleculez7. Indeed it may be desirable to design CDR-grafted antibodies by using framework regions based on human germ- line V-gene segments. The design of antibodies for therapy would certainly be rev- olutionized if in
`entirely, but how it can be bypassed, for example by changing the idiotype of the engineered antibody, and whether the antibodies can be used for long enough to achieve clinical benefit. So far we have focussed almost entirely on the construction and use of glycosylated antibodies ex- pressed in mammalian cells. How- ever, the use of antibody frag- ments may be advantageous in some applications, since they pen- etrate tissues more readily, and are cleared more rapidly from the serum. This may help in neutral- izing and clearing drugs from the serum, or in imaging tumours with radioactive entities coupled to the fragments’. Although antibody fragments, lacking the glycosylated Fc portion, cannot trigger effector functions, they could in principle be equipped to do so, for example by chemically linking Fab frag- ments togethe? as bispecific antibody fragments (with one arm binding a tumour cell antigen and the other binding and triggering effector cells such as cytotoxic T cells or monocytes). Furthermore, antibody frag- ments can be expressed by secretion from bacteria29*30, and can be readily derived from the V genes of hybridomas, or from V gene repertoires. The repertoires are cloned for display on the sur- face of filamentous bacteriophage by fusion of the encoded antibody fragment to a coat protein of the phage, and phage with the desired activities selected by binding to antigen. Indeed this technology mimics the strategy of immune selection, and human antibody fragments with speci- ficities against many different foreign and human self-antigens
`
`assays were available to test the immuno- genicity of different constructs. However the key practical issues are not whether the immune
`
`response can be avoided
`
`vitro
`
`TiPS -May 1993
`
`[Vol. 141 have been isolated from the same ‘single pot’ of phages (see Refs 31, 32 for review). Over the past century we have seen three generations of anti- body therapeutics: polyclonal animal antibodies, rodent mono- clonal antibodies and now humanized antibodies. We antici- pate that the use of ‘repertoire selection’ technologies to make human antibodies and fragments will provide the next generation. However, in the meantime it seems likely that humanized anti- bodies will prove clinically useful for treating several diseases, and the experience should prove valuable for designing and formulating the next generation of antibody therapeutics.
`
`References
`
`Winter, G. and Milstein, C. (1991) Nature 349, 293-299 Orlandi. R.. Giissow. D. H.. lones, P. T. and Winter, G. (1989)
`
`Proc.sNr!l Acad.
`
`Sci. USA 86, 3833-3837
`
`l-142
`Adair, J. R. and Mountain, A. (1992) Biofech. Gen. Eng. Rev. 10,
`130,l~O
`Biotech-
`
`nology
`10, 169-175 Morrison, S. L. (1992)
`Immunol. 10,239-265
`
`Annu. Rev.
`
`Jones, P. T., Dear, P. H., Foote, J., Neuberger, M. S. and Winter, G. (1986) Nature 321,522-525 Riechmann, L., Clark, M., Waldmann, H. and Winter, C. (1988) Nature 332, 323-327 . . Tempest, P. R. et al. (1991) Biofecnnocogy 9. 2hL271 10 Kettleborough, C. A., Saldanha, J., Heath, V. J., Morrison, C. J., and Bendig, M. M. (1991)
`
`Protein Engin. 4,
`
`773-783 11 Foote, J. and Winter, G. (1992) J. Mol. Biol. 224,487499 12 Saul, F. A. and Poliak. R. 1.11993) J. Mol. Biol. 230,15-20 ’ I 13 Carter, P. et a/. (1992) Proc. Natl Acad.
`
`Sci. USA
`89, 4285-4289 14 Queen, C. et al. (1989) Proc. Nafl Acad.
`Sci. USA
`
`Mol.
`
`86, 10029-10033 15 Padlan, E. A. (1991) Mol. Immunol. 28, 489-498 16 Chothia, C. and Lesk, A. (1987) 1.
`Biol. 196, 901-917 17 Glaser, S. M., Vasquez, M., Payne, P. W. and Schneider, W. P. (1992) J. Immunol. 149. 2607-2614 18 Co, M. S., Avdolavic, N. M., Caron, P. C.
`(1992) J. Immunol. 148, 1149-1154 19 Marks, J. D., Griffiths, A. D., Malmqvist, M. ef al. (1992)
`20 Hawkins, R. E., Russell, S. J. and Winter, G. (1992) J. Mol. Biol. 226, 889-896 21 Hale, G. et al. (1988) Lancef ii, 1394-1399 22 Isaacs, J. D. et al. (1992)
`340, 748-752 23 Mathieson, P. W. et al. (1990) Nero Engl. 1. Med. 323, 250-254 24 Hird, V. ef al. (1991) Br. J.
`64, 911-914 25 Anasetti, C. et al. (Abstr.) American
`
`et al.
`
`Biotechnology 10,779-789
`
`Lancet
`
`Cancer
`
`Adair, J. R. (1992) Immunol. Rev.
`Bebbington, C. R. ef al. (1992)
`

`

`TiPS - May 1993 [Vol. 141
`
`Science
`
`Immunol. Reu.
`
`143 Society of Haematology, Anaheim, November 1992 26 Briiggemann, M., Winter, G., Waldmann, H. and Neuberger, M. S. (1989) 1. Exp. Med. 170, 2153-2157 27 Allen, P. M., Masueda, G. R., Haber, E. and Unanue, E. R. (1985) J. Immunol. 28 29 30 135. 368-373 1041-1043 Carter, I’. ef nl. (1992) Biotechnology 10, 163-167 31 Skerra, A. and Pluckthun, A. (1988)
`240, 1038-1041 Better, M., Chang, C. P., Robinson, R. R. and Horwitz, A. G. (1988) Science 240, 32 Marks, J. D., Hoogenboom, H. R., Griffiths, A. D. and Winter, G. (1992) 1. Biol. Chem. 267, 1-4 Hoogenboom, H. R., Marks, J. D., Griffiths, A. D. and Winter, G. (1992)
`130, 41-68 Herman Waldmann and Stephen Cobbold
`
`many body cells. In addition, T
`cells have unique clonally distrib-
`uted receptors
`that respond
`to
`these antigens displayed
`to them
`on specialized
`cells
`in the lym-
`phoid tissues (antigen-presenting
`cells or AK). They then prolifer-
`ate and differentiate
`to effector
`mode (Fig. 1). It is now clear that
`the cells with the greatest ability
`to present antigen and activate T
`cells are the dendritic cells’,*. Two
`features of dendritic
`cells
`that
`endow
`them with
`this property
`are their possession of a particular
`array of cell surface ligands comp-
`lementary to an array of adhesion
`molecules on T cells, and their
`abundant MHC class II expression.
`A further requirement for T-cell ac-
`tivation is one of collaboration or
`help from other T cells responsive
`to the same antigen%’ (Fig. 2).
`
`agents.
`immunosuppressive
`useful
`are potentially
`antibodies
`Monoclonal
`antibody can be used to guide the
`Short courses of CD4KD8 monoclonal
`immune system of experimental
`animals to accept organ grafts and to arrest
`reviewed by Herman Waldmann and
`autoimmunity.
`This reprogramming,
`Stephen Cobbold,
`is accompanied
`by potent T-cell dependent,
`‘infectious’
`regulatory mechanisms. A goal for therapeutic
`immunosuppression
`should be
`to understand and harness these innate immunoregulatoy mechanisms.
`
`to them in the
`ments displayed
`clefts of major histocompatibility
`complex (MHC) class I and II mol-
`ecules expressed on the surfaces of
`
`im-
`The ideal form of therapeutic
`munosuppression would be one
`that could be given
`a short-
`term period to achieve
`long-term
`unresponsiveness
`to the desired
`antigen, without
`impairing
`the
`response
`to
`infectious
`agents.
`Current immunosuppressive
`regi-
`mens are relatively non-antigen-
`specific, require long-term admin-
`istration
`and
`incur a sustained
`risk of infection and undesirable
`side-effects.
`tolerance
`If we are to achieve
`in auto-
`as a therapeutic
`goal
`immunity or in transplantation,
`then it is essential
`that we under-
`stand how
`the body normally
`establishes
`‘self tolerance’. From
`this basis we can determine which
`of these natural mechanisms might
`be exploited
`for pharmaceutical
`control. As thymus-derived
`cells
`(T cells) are required
`for driving
`most forms of immune response
`it
`is appropriate
`that we focus our
`discussion on how to control their
`functions.
`
`T-cell
`or
`foreign
`recognize
`T cells
`‘nonself’ antigen as peptide
`frar
`
`H. Wutdmann
`is Kay K,-nJal/ ProJcssor of
`Therapeutic
`Immunology
`and S. Cobbold
`I’S
`Associate ut the Immunology
`Division, Department
`of Pathology,
`Road, Cambridge, UK CB2
`IQF’.
`Tennis Court
`
`T lymphocyte I Antigen-Presenting Cell Interaction Molecules LFA-l,CD2 cD4, CD8 LFA-2, ICAM- CD28, B7 etc.
`
`Interleultins
`l-12
`
`Tumolu necmais iilctors
`
`antigen-
`interactions between a T ccl: and
`three types of molecular
`Fig. 1. There
`the foreign
`presenting cell (APC). First, the specific T-cell receptor
`(TCR)
`recognizes
`in the
`antigen,
`in the form
`processed peptide
`[the antigen
`(Ag) peptide], bound
`a
`the major histoc,?mpatibility
`complex
`(MHC) mo/ecules of the APC. Secondly,
`Cleft of
`Uecules
`on the T cell bind to their ligands on the APC. These
`series of adhesior,
`include CD4 and CDB, wnich are co-receptors
`for the MHC mo/ecJes,
`and CD28 which
`interacts with the APC
`ligand 87 to provide
`‘costimulation
`Thirdly, the T cell expresses
`receptors
`for various factors that regulategrowth
`and differentiation
`(cytokines), such as
`interleukin 2, some
`which are produced by other activated T cells and form the basis
`of ‘he/p’
`Although
`the T-cell receptor binding provides
`the primary
`trigger the T cell, the outcome, which can be either a proliferative
`response or
`the induction of a non-responsive
`state, depends on further signals from these adhesion
`molecules
`and growth
`factor
`receptors
`in order
`to put
`antigen
`recognition
`‘in
`context:
`
`0 1993.
`
`Elscvicr Science Publishers Ltd (UK) 0165 - 6147/93/506 00
`
`over
`recognition of nonself
`Senior Research
`are
`an
`of e
`‘.
`of
`and collaboration.
`signal to
`the
`