1 Engineering Antibodies for Therapy
`Celltech Research, Ce/ltech Ltd, Slough, UK
`
`A . MOUNTAIN AND J . R . ADAIR
`
`Introduction
`
`Antibodies have long been viewed as potential agents for targeted drug
`delivery and other therapeutic interventions, largely with a view to exploiting
`the combination of high specificity and affinity of the antibody-antigen
`interaction. Since the development of rodent monoclonal antibody (MAb)
`technology (Kohler and Milstein, 1975) it has been possible in principle to
`produce rodent MAbs to virtually any antigen, and a large number of rodent
`MAbs relevant to human therapy have been generated. MAbs have already
`been used clinically for the diagnosis and therapy of several human disorders,
`notably cancer and infectious diseases, and for the modulation of immune
`responses. The target antigens have been tumour-associated antigens (T AAs,
`Boyer et al., 1988; Herlyn, Menrad and Koprowski, 1990), specific cell type
`markers, viruses, bacteria and specific human proteins of physiological
`
`Abbreviations: ADCC, antibody-dependent cellular cytotoxicity; bp, base pairs; BSA, bovine
`serum albumin; c[antibody name], mouse variable region-human constant region chimeric[ anti­
`body name]; ADCMC, antibody-dependent complement-mediated cytotoxicity; cDNA, comple­
`mentary DNA; CDR, complementarity determining regions; CEA, carcinoembryonic antigen;
`CTL, cytotoxic T lymphocyte; d, days; DHFR, dihydrofolatc reductase; DNS, dansyl; ELISA,
`enzyme-linked immunosorbent assay; E:T, effector-to-target ratio; HTC, fluorcsccin isothiocy­
`anate; GM-CSF, granulocyte-macrophagc-colony-stimulating factor; gpt, xanthine/guanine phos­
`phoribosyl transferase gene; h, hours; HAMA, human anti-mouse antibody; HbsAg, hepatitis B
`surface antigen; hCMV, human cytomegalovirus; hEGFR, human epidermal growth factor
`receptor; HIV, human immunodeficiency virus; hph, hygromycin B phosphotransfcrasc gene;
`HRP, horse radish peroxidase; hyg', hygromycin resistance; IC,., quantity required for 11%
`inhibition of activity; ID,., quantity of virus required for n% infection; IFNy, interferon y; i.n.,
`intra-nasal; i.p., intra-peritoneal; i.v., intra-venous; kbp, kilobase pairs; KLH, kcyhold limpet
`haemocyanin; LT, lymphotoxin; LTR, long terminal repeat unit; MAb, monoclonal antibody;
`MLR, mixed lymphocyte reaction; mRNA, messenger RNA; MSX, methionine sulphoximine;
`MTX, methotrexate;
`rzeo, neomycin phosphotransferase; NIP, 5-iodo-4-hydroxy-3-
`nitrophenacetyl; NP, 4-hydroxy-3-nitrophenacetyl; NP-cap, NP-caproic acid; OD,., optical
`density at rz nm;% i.d.g-1, percentage of injected dose per gram of tissue; o-PDM, N,N'-1.2-
`phcnylcnedimaleimide; PBMC, peripheral blood mononuclear cells; PC, phosphorylcholine;
`PEG, polyethylene glycol; PGK, phosphoglycerate kinase; pfu, plaque-forming units; PLAP.
`placental alkaline phosphatase; PMN, polymorphonuclcar lymphocyte: p/o, promoter/operator;
`rbs, ribosome binding site; RES, reticulo-endothelial system; rIL-2, recombinant interleukin 2;
`s.c., subcutaneous; SDM, site-directed mutagenesis; SRBC; sheep red blood cells; TAA,
`tumour-associated antigen; TNB, thionitrobcnzoate; TNP, trinitrophenyl; V11, heavy chain
`variable domain; V L> light chain variable domain.
`8iotec!mology a11d Genetic Engi11ceri11g Reviews- Vol. 10. December 1992
`0264--8725192110/l-142 $20.00 + $0.00 © lnlcrccpl Ltd. P.O. Box 716, Andover. Hampshire SPlO l YG. UK
`1
`
`1 of 142
`
`BI Exhibit 1067
`
`

`

`A. MOUNTAIN AND J.R. ADAIR
`2
`importance (particularly cytokines or their receptors). At the end of 1991
`there were 132 biotechnology-based medicines in formal clinical development
`(Pharmaceutical Manufacturers Association Report, 1992) or awaiting final
`approval from the US Food and Drugs Administration (FDA). Sixty-six of
`
`the 132 were for cancer therapy and 58 of the 132 were MAbs. Three rodent
`
`MAbs have so far been approved and launched as products: OKT3 is a naked
`MAb and has been approved by the FDA for treatment of acute kidney
`transplant rejection; OncoScint and MyoScint are Mab-isotope conjugates
`which have been approved outside the US as imaging agents for colorectal
`cancer and myocardial infarction, respectively. The human MAb 'Centoxin'
`has also been approved in Europe for treatment of septic shock. The total
`market for MAbs is presently around $330 million per annum and is estimated
`to grow approximately ten-fold by 1996. Therapeutic MAbs will account for
`most of this market.
`Although the specificity of MAbs undoubtedly gives them immense poten­
`tial in medicine, rodent MAbs are certainly not ideal therapeutic agents. The
`five most important issues and technical challenges in the development of
`MAb-based therapies are: (1) identifying MAbs of suitable affinity and
`specificity; (2) overcoming human immune responses against rodent MAbs
`and against any cell-killing agents attached to them; (3) identifying and
`
`harnessing appropriate cell-killing agents; ( 4) achieving appropriate pharma­
`
`cokinetics and biodistribution; (5) achieving economic manufacture, which is
`of particular relevance for highly engineered MAbs and for MAb-cytotoxic
`agent conjugates (as opposed to naked MAbs).
`As the above statistics indicate, a large proportion of the MAb-based
`agents presently in clinical development are for treatment of cancer and in
`this review the development of anti-cancer MAbs and MAb conjugates will
`largely be used to illustrate the approaches being taken to address the five key
`issues. The review begins with a brief description of the structure of
`antibodies and antibody genes, followed by a summary of the arguments and
`evidence relating to the importance of affinity and specificity for MAb-based
`therapies. We then briefly summarize the available clinical results with naked
`rodent MAbs. Next we describe the approaches being taken to overcome the
`immunogenicity in patients of rodent MAbs, which is certainly the most
`serious and general problem for MAb-based therapies. This section concen­
`trates largely on antibody humanization, which is the most promising solution
`to the problem. The processes developed for efficient cloning of antibody
`genes and for production of engineered whole antibodies are then described.
`This is followed by a summary of the approaches being taken to improve the
`pharmacokinetics and biodistribution of MAbs, focusing particularly on the
`development of engineered antibody fragments, and then by a summary of
`production systems being used for such fragments. The first half of the review
`is then completed by a summary of the various cell-killing strategies being
`developed for MAb-based therapies. The second half of the review is largely
`devoted to a detailed summary of the construction, expression, pre-clinical
`studies and data on efficacy and immunogenicity for engineered MAbs and
`MAb conjugates that have been used in clinical studies by the time of writing.
`
`2 of 142
`
`BI Exhibit 1067
`
`

`

`3
`Engineering antibodies for therapy
`The review finishes with some conclusions and a summary of recent advances
`in antibody technology which may lead eventually to the successors of
`humanized rodent MAbs for therapy.
`
`The structure of antibodies and antibody genes
`
`In order to understand the later sections of this review (especially those on
`antibody gene cloning and humanization) it is necessary to have some
`knowledge of the structure and organization of antibodies and their genes.
`We give a brief description here.
`Higher mammals have five classes of immunoglobulin, termed IgG, lgM,
`IgA, IgE and IgD. The structures and functions of these five classes have
`
`MAbs of therapeutic potential are of the IgG class, and have the basic
`structure shown in Figure IA. IgG antibodies have a tetrameric structure
`consisting of two identical 55 kDa glycosylated proteins (termed 'heavy
`chains') and two identical 25 kDa proteins, which are normally not glycosy­
`lated (termed 'light chains'), covalently Jinked by disulphide bridges. The
`
`been very well described by Roitt, Brostoff and Male (1987). Almost all
`proteins are organized into discrete folding domains of around 110 amino
`
`acids which are encoded in the genome on separate exons (Figures 1 B and
`1 C). Each light chain associates with and is covalently linked via a disulphide
`bridge to a cysteine in the N-terminal region of one heavy chain, and the
`C-terminal half of the heavy chains associate with each other to form a Y- or
`T-form structure. The heavy chains are also covalently linked to each other
`via disulphide bridges in the hinge domain.
`Sequence information is now available for hundreds of antibodies of many
`different species and reveals that the N-terminal domains of each chain are
`much more variable in sequence than the other domains. The N-terminal
`domains are therefore termed 'variable domains' and the others 'constant
`domains'. Three non-contiguous regions within these variable domains are
`particularly variable and are usually referred to as 'hypervariable loops' or
`'complementarity determining regions' (CDRs). This sequence variation is
`postulated to provide the variability (within these otherwise highly conserved
`proteins) which enables antibodies to recognize and bind to a very wide range
`
`of antigens (Wu and Kabat, 1970; Kabat et al., 1987). The proposal has been
`(Amit et al., 1986; Boulot et al. , 1987; Colman et al., 1987; Sheriff et al. , 1987;
`Davies et al. , 1989; Padlan et al., 1989; Tulip et al. , 1989, reviewed in Alzari et
`al., 1987; Bentley et al. , 1990; Bhat et al. , 1990; Davies, Padlan and Sheriff,
`1990 and Poljak, 1991). The variable region residues that are not part of the
`
`confirmed by structural studies, which show that the hypervariable sequences
`are (in most cases) associated on the surface of the antibody as a set of loops.
`The loops form a large surface patch and are in contact with antigen in cases
`for which structural information on the antibody-antigen complex is available
`
`CDR or loops together constitute the 'framework' of the variable region. It
`has been shown that the exons for the variable domains are assembled from a
`number of repeated gene families - V and J for the light chain and V, D and
`J for the heavy chain - by a series of recombination events during the
`
`3 of 142
`
`BI Exhibit 1067
`
`

`

`+:. ?> �
`77 "i ·�
`''"' nm "' mo "' i -
`0 c � >
`�';>,�·
`2::
`llllli
`II• /� 111
`\
`--. r
`. . · I"'
`=c11=-cH�1-=-"'�', o�o�:--�',
`:
`!:':H_ J_ Ill. :I&!
`
`
`z
`> z
`0 '-<
`�
`
`>-0
`
`
`:i::i
`
`A
`
`B
`
`c
`
`100-300 V GENES !N 7 CLUSTERS ENCOM?ASS\NG
`1-3Mb OF CHROMOSOME
`
`lllllll 111111 111111 IUll llllll 1111111
`
`!!1111
`
`5 JKMINIGENES AND 1 CONSTANT
`REGION GENE IN 4 Kb REGION
`
`//
`
`_____ llill_ I
`
`BI
`L.
`
`I
`
`
`
`DODOO u-.....1 ----.1
`
`HEAVY CHAIN
`
`100-1000 VH GENES IN 10
`INTERSPERSED SUBGROUPS
`
`AT LEAST 12 DH MINIGENES LINKED TO 4 JH MINIGENES
`AND 8 CONST ANT REGION GENES IN A 420 Kb REGION
`OH JH C
`
`L
`
`v
`
`Figure 1. Protein and gene organization of murine IgGK.
`
`1[u --=1 �',---:
`
`H
`
`CH2
`
`CH3
`
`4 of 142
`
`BI Exhibit 1067
`
`

`

`5
`
`Engineering antibodies for therapy
`development of the antibody-producing B-cell lineage. The variable region
`exon along with the signal sequence exon and the promoter/enhancer
`(involved in transcription) is then juxtaposed with the constant region gene
`family by further recombination events for subsequent expression (Figures
`JA, JB; reviewed by Alt, Blackwell and Yancopolous, 1987). The organiza­
`tion of the lg loci in mice and humans has been reviewed recently by Lai,
`Wilson and Hood (1989).
`The constant regions tend to be conserved in sequence among antibodies of
`a given species, and also to a lesser extent between species. Light chains have
`a single constant domain for which there are two gene loci, CK and Cl. .. IgGs
`have three constant domains on the heavy chains, CHl, CH2 and CH3.
`Between the CHl and CH2 domains (for IgGs) is a short proline-rich peptide
`sequence termed the 'hinge' which contains the cysteines that bridge the two
`heavy chains. IgGs also have a site in the CH2 domain for N-linked
`glycosylation, which is required for structural integrity of the antibody and for
`some of its effector functions. Sequence motifs within the CH2 and CH3
`domains are responsible for the effector functions, such as complement
`activation and binding to other cells of the immune system. In humans and
`rodents there are four different types of IgGs, termed 'isotypes', which vary
`in their spectrum of effector functions as a result of amino acid sequence
`variation in the constant regions (Burton, 1990). In humans there are a
`number of immunologically distinct variants of IgGl, 2 and 3, termed
`allotypes (Gorman and Clark, 1990). These allotypes are racially distributed,
`for example the Glm(3) marker predominates in Caucasian IgGl whereas
`Glm(l,17) predominates in Asian and Japanese individuals.
`Until the advent of recombinant DNA technology antibody fragments
`(Figure 2) were generated by proteolytic digestion. Pepsin cleaves IgGs on
`the C-terminal side of the hinge, liberating an antigen binding fragment
`referred to as the F(ab')i. Papain cleaves on the N-terminal side of the hinge
`and liberates two F( ab) fragments and a single Fe fragment. The F( ab)
`fragments have a single antigen binding site (monovalent), while the F(ab'h
`has two (bivalent). The term F(ab') means monovalent but with the hinge
`sequence also present. The heavy chain of the F(ab) or F(ab') is usually
`referred to as the Fd or Fd'. The variable domains of the heavy and light
`chains (V H and V d together comprise a fragment called the Fv. This is the
`smallest fragment which retains the full antigen binding activity of the
`monovalent antibody. Although the Fv can be obtained for some antibodies
`by proteolytic digestion of the IgG the process is very inefficient. Fvs
`dissociate into VH and V L under physiological conditions, and so are not
`useful for therapy. The single chain Fv (scFv) represents the most successful
`strategy for stabilizing the Fv. It has V H and V L linked by a short peptide
`linker (between the C-terminus of one domain and the N-terminus of the
`other) and expressed as a single polypeptide chain. It is possible to make scFv
`variants for most MAbs which retain most or all of the monovalent antigen
`binding activity of the MAb. In some cases V H alone displays significant
`antigen binding activity, an observation which has led to use of the term
`'single domain antibodies' (DAbs; Ward et al. , 1989). Molecular biology
`
`5 of 142
`
`BI Exhibit 1067
`
`

`

`6
`
`A. MOUNTAIN AND J.R. ADAIR
`procedures now allow the efficient production of the F(ab), F(ab'), F(ab'h
`and Fv, as well as novel engineered variants of these fragments and of course
`whole IgG.
`
`MAb affinity and specificity
`
`Laboratories developing MAbs for therapy usually choose the highest affinity
`MAb available because it is widely assumed that high affinity confers a
`therapeutic advantage. There is, however, very little actual experimental data
`relevant to the importance of affinity. This is largely because all comparative
`studies in patients or animal models have concerned MAbs which recognize
`different epitopes as well as having different affinities. For example compar­
`ative studies have been performed in an animal model with rodent MAbs
`which neutralize human TNF-cx as part of a programme to develop a MAb for
`treatment of septic shock, in which TNF-cx is an important mediator. Two
`mouse MAbs were compared for ability to prevent pyrexia induced in rabbits
`by human TNF-cx (R. Foulkes, personal communication). The 100-fold
`difference of the MAbs in ability to bind TNF-cx in vitro was shown to give a
`seven-fold difference in the doses required for complete neutralization. A
`similar correlation between affinity and effective dose has also been observed
`for MAbs neutralizing IL-5 (M. Bodmer, personal communication). These
`demonstrations of the importance of affinity are not quite conclusive,
`however, since the antibodies were directed to different epitopes.
`Most other relevant data concern MAbs recognizing T AAs, for which the
`affinity issue is complicated by several factors, including the tendency of such
`antigens to be shed from the tumour into the circulation, and tumour
`penetration. Circulating antigen may in some cases interfere with MAb
`localization to the antigen on the tumour, and MAbs which recognize
`different epitopes of the antigen on the tumour cells may also be differentially
`affected by the presence of circulating antigen. It has also been suggested that
`higher affinity antibodies may show poorer tumour penetration through an
`increased tendency to bind tightly to the antigen on the tumour cells close to
`the blood vessel through which the MAb gains access to the tumour. Many
`studies have been performed on the tumour localization in colorectal cancer
`patients of MAbs recognizing the TAA's carcinoembryonic antigen (CEA)
`and polymorphic epithelial mucin (PEM). For both antigens the affinities of
`the MAbs concerned have been measured and reported to be different.
`Although marked differences between the MAbs in tumour localization are
`also reported the data do not allow the effect of affinity differences on
`localization to be distinguished from differences in circulating half-life
`(conferred by differences in isotype and immune response), tumour site and
`size, tumour vascularity and permeability. Perhaps the most direct evidence
`(Schlom et al. , 1992) on the importance of MAb affinity for tumour therapy
`concerns two MAbs recognizing the T AA T AG72, which is expressed on
`several human tumour types. MAb B72.3, which binds this antigen, has been
`administered to over 1000 patients and shown to localize to about 75% of
`gastrointestinal, ovarian, prostate and breast tumours. B72.3 has gained
`
`6 of 142
`
`BI Exhibit 1067
`
`

`

`Engineering antibodies for therapy
`
`7
`
`HEAVY CHAIN
`
`Fab
`
`Hin go
`
`Chernlcally
`
`cross-linked F(ab')
`
`2
`
`Single chain Fv
`
`Fv
`
`
`
`Cross-linked Fv
`
`Bivalent
`single chain
`Fv-hlnge
`
`Bivalent
`chemically
`cross-linked
`slngle chain
`Fv-hinge
`
`Figure 2. Engineered antibody fragments.
`
`7 of 142
`
`BI Exhibit 1067
`
`

`

`A. MOUNTAIN AND J.R. ADAIR
`8
`regulatory approval as part of a colorectal cancer imaging agent called
`'OncoScint'. Recently a MAb CC49 has been characterized which binds to an
`overlapping epitope on T AG72 and which has approximately seven-fold
`greater affinity for this antigen than B72.3 (dissociation constants are
`16·2 x 10-9 and 2·5 x 10-9 M, respectively). The two MAbs have the same
`range of reactivities to normal adult tissues. The higher affinity MAb was
`clearly shown to have a greater anti-tumour effect (two- to three-fold at the
`same dose) than B72.3 when both MAbs were conjugated to the therapeutic
`radioisotope 1311 and used in tumour regression experiments in nude mice.
`Other studies in animal models have concluded that higher affinity MAbs give
`greater tumour uptake at low doses which do not saturate the antigen, but not
`at higher doses (Sung et al., 1992).
`Other relevant data concern comparisons of MAbs and MAbs conjugated
`to cytotoxic agents for ability to bind to, and kill, cancer cells in tissue culture.
`Cell killing by MAbs carrying low molecular weight cytotoxic drugs or protein
`toxins requires internalization of the conjugates and intracellular release of
`the active cytotoxic agent. A humanized variant of the MAb CTMOl , which
`binds to PEM, has been identified which has an affinity two- to three-fold
`greater than its parent murine MAb but retains the same specificity (J .R.
`Adair et al., unpublished). Calicheamicin conjugates of the humanized and
`murine forms have been compared for binding to tumour cell lines expressing
`PEM, for internalization and cell killing in vitro and for tumour regression in
`animal models. The higher affinity humanized form performs better for all
`three parameters (L. Hinman, personal communication).
`In summary, there are not yet sufficient data to demonstrate conclusively a
`general correlation between affinity and efficacy for therapeutic MAbs or
`their conjugates in animal models or in patients. Indeed, ethical consider­
`ations make it very difficult to carry out such comparisons in the clinic. Such
`demonstrations will require thorough head-to-head efficacy comparisons in
`animal models of MAbs with different affinities but which recognize the same
`epitope and have equivalent pharmacokinetics and biodistribution. Recently
`murine B72.3 and a humanized variant with a 100-fold lower affinity have
`been compared for ability to localize to tumours in mice (D. King, personal
`communication). The murine antibody showed somewhat greater localiza­
`tion. Although definitive studies on the importance of affinity remain to be
`done, it is very likely that for most therapeutic applications MAbs with
`minimum dissociation constants of 10-9 to 10-10 M will be essential to achieve
`efficacy and to permit economically realistic doses. It is therefore extremely
`important to retain at least most of the MAbs affinity for its antigen through
`the antibody engineering and conjugation procedures required to render it
`suitable for therapeutic use.
`One of the important advantages of MAbs and MAb conjugates over
`conventional low molecular weight drugs is their specificity for the target
`molecule. This is very likely to be reflected in a lower failure rate of
`therapeutic MAbs in development at the stage of toxicology. Specificity is of
`particular importance for anti-cancer MAbs because unfortunately no anti­
`gens have yet been identified which are expressed exclusively on tumour cells.
`
`8 of 142
`
`BI Exhibit 1067
`
`

`

`9
`Engineering antibodies for therapy
`This is why the term 'tumour-specific antigen' has largely been replaced by
`the more accurate 'tumour-associated antigen'. Expression of the antigen on
`normal tissues is a major potential problem for tumour therapy with MAbs
`and MAb conjugates, since it can lead to dose-limiting toxicities. It is
`therefore important when selecting such MAbs for development to include a
`step in the screening cascade which evaluates binding of the MAb to normal
`tissues. lmmuno-histochemical studies are usually performed which indicate
`the ability of the MAb to bind to tumours and to a range of normal tissues.
`MAbs can then be chosen which bind antigens that are highly expressed on
`tumours and also on a large proportion of cells within any one tumour. The
`most suitable of these MAbs will be those for which the antigen shows little
`expression on normal tissues - especially those likely to be involved in
`dose-limiting toxicities - or shows expression on a much smaller proportion
`of the cells. In some favourable cases the antigen may be expressed on
`important normal tissues but may be in a cellular location that is much less
`accessible to a therapeutic MAb in the circulation than is the antigen on the
`surface of tumour cells. The most frequent normal tissue expressing the T AA
`is of course the tissue of origin of the tumour. In many cases the T AA is much
`more highly expressed on tumour cells than on these normal cells, and this
`may be reflected in the levels of MAb uptake by tumour and normal tissues at
`relatively low doses. There are many T AAs for which it is suggested that the
`tumour form of the antigen is structurally different from that found on normal
`cells. Considerable effort is presently going into identifying MAbs recogniz­
`ing specifically the aberrantly glycosylated forms of glycosphingolipid and
`glycoprotein T AAs suggested to be present on tumour cells (Hakomori,
`1991a, b).
`A related potential problem of specificity is caused by the tendency of
`many T AAs to be shed from the tumour into the circulation. In some cases
`this circulating antigen may interfere with localization of the MAb to the
`tumour, especially at low doses. MAb-antigen complex formation in the
`circulation may cause toxicity when the MAb carries a cytotoxic agent
`through deposition of the agent in the organs of clearance of the complex. In
`one targeting study with colorectal cancer patients a particular anti-CEA
`MAb was shown to localize efficiently to 42 out of 43 tumours examined
`(Boxer et al. , 1992). In the remaining case the patient had a high level of
`circulating CEA. On the other hand the presence of circulating PEM was
`shown to increase levels of immune complexes in the circulation and to
`enhance tumour localization in colorectal cancer patients administered the
`anti-PEM antibody ICR2 (Davidson et al. , 1991). It is thus far from clear yet
`whether circulating antigen will be a general problem for cancer therapy with
`MAbs or MAb conjugates. In cases where circulating antigen is shown to
`interfere with MAb targeting, the specificity of MAbs is such that it may well
`be possible to identify and use MAbs capable of binding preferentially to the
`tumour-bound form of the antigen. The MAb CTMOl , for example, appears
`to bind preferentially to tumour-bound PEM rather than to PEM shed into
`the circulation (T. Baker, personal communication).
`
`9 of 142
`
`BI Exhibit 1067
`
`

`

`A. MOUNTAIN AND J.R. ADAIR
`10
`Clinical data with naked rodent MAbs
`
`The clinical use of naked rodent MAbs has so far largely focused on the
`treatment of cancer, and on suppression of immune responses involved in
`auto-immune disease, graft versus host disease (GVHD) and transplant
`rejection. Naked rodent MAbs have in general proven very ineffective in
`cancer therapy, with only 23 partial remissions and three complete remissions
`reported among the initial 185 patients included in 25 clinical trials (Catane
`and Longo, 1988). This is partly because most of these MAbs were not
`directed against cell surface structures with functions required for tumour cell
`proliferation, partly because HAMA responses prevented repeated adminis­
`tration, and partly because most rodent antibodies are very inefficient at
`recruiting human immune effector mechanisms. (Murine IgG2as and rat
`IgG2bs are rather more efficient in this respect than other isotypes.) Some
`partial responses have been observed in B-cell lymphoma patients treated
`with MAbs directed to B-cell lg idiotypes (Meeker et al. , 1985). The MAb
`most widely used clinically is OKT3, which binds to the CD3 antigen of the
`T-cell receptor (TCR) complex that is expressed on virtually all circulating
`T-cells. OKTI has been approved by the FDA for the treatment of acute
`renal allograft rejection on the basis of its superiority in randomized clinical
`trials (Ortho Multi-Center Transplant Study Group, 1985) over conventional,
`broad spectrum immunosuppressive agents (93% reversal of acute rejection
`episodes for OKTI compared to 75% for conventional agents). Despite being
`itself an immunosuppressive agent, murine OKT3 elicits HAMA responses in
`patients (see later), and also leads to toxicity problems arising from cytokine
`release which accompanies T-cell activation in response to binding of the
`MAb. OKTI therapy also leads to much broader immunosuppression than is
`desirable, with increased incidence of viral infections and B-cell neoplasms.
`Anti-Tac is another rodent MAb which is intended for use in clinical studies
`of renal allograft rejection and which appears much more promising. This
`MAb blocks the binding of IL-2 to the IL-2 receptor, which is expressed on
`T-cells participating in allograft rejection, in certain auto-immune disorders
`and in one type of acute T-cell leukaemia (ATL). Of 20 ATL patients treated
`with anti-Tac seven showed remission lasting from one to at least 17 months
`(Waldmann, 1989, 1991a, b). Murine anti-Tac is presently in clinical evalua­
`tion for acute allograft rejection, and has been successfully humanized (see
`below).
`
`Approaches to overcoming rodent MAb immunogenicity
`
`There is a great deal of evidence demonstrating that the administration of
`antibodies from rodents (and other species) to humans results in an immune
`response in the great majority of patients, which limits the use of such
`antibodies to one or perhaps two doses (Lind et al. , 1991, and references
`therein). For mouse antibodies the response has been termed the 'HAMA'
`(human anti-mouse antibody) response. Administration of further doses
`leads to accelerated clearance and in many cases to complete abrogation of
`
`10 of 142
`
`BI Exhibit 1067
`
`

`

`1 1
`Engineering antibodies for therapy
`efficacy. I t can also lead to allergic reactions and in severe cases to
`anaphylactic shock. Clearance of the complexes which form between the
`administered MAb and the HAMA antibodies can also occur through routes
`which deposit cytotoxic agents carried by the former in undesirable locations,
`giving toxicity in the organs of clearance or in the reticulo-endothelial system
`(RES). In some cases the HAMA titre increases with the kinetics expected of
`a secondary response, consistent with the presence of a pre-existing antibody
`with anti-mouse specificity (Schroff et al. , 1985; Shawler et al. , 1985;
`Courtenay-Luck et al. , 1987; Khazaeli et al. , 1991). Antibody detected during
`the early phase of the response tends to be directed to the Fe portion of the
`antibody, but later on reactivities outside this region - and especially to the
`variable region - can be detected. In some cases components of the response
`are directed to the antigen binding site, termed 'anti-idiotype' responses. It
`has been clearly demonstrated that HAMA interferes with tumour localiza­
`tion by anti-TAA antibodies (see for example Ledermann et al. , 1988;
`Goldenberg, Sharkey and Ford, 1987) and with the immunosuppressive effect
`of OKT3 in the course of acute allograft rejection (Chatenoud et al. , 1986).
`Most antibody-based therapies are very unlikely to achieve success with a
`single dose. The only exceptions to this are likely to be in acute indications
`such as septic shock, for which it is possible that a single dose of an anti-TNF
`or anti-ILl antibody can neutralize sufficient of these cytokines to prevent the
`serious (and usually fatal) organ damage which otherwise occurs. It is
`abundantly clear that it will not be possible in general to eradicate solid
`tumours with a single dose of an antibody or an antibody conjugate. Many
`different approaches are therefore being taken to allow repeated administra­
`tion of therapeutic antibodies. One obvious approach is to isolate and use
`human rather than foreign antibodies. Although there are some human
`antibodies presently in clinical evaluation - notably the anti-LPS antibody
`'Centoxin' for treatment of septic shock - the technology does not yet exist
`routinely to isolate human antibodies of suitable affinity and specificity.
`Recent technology developments which may eventually lead to more routine
`isolation of therapeutically useful human MAbs are summarized at the end of
`this review. In the short term the use of at least the binding site from a
`non-human antibody remains the only generally applicable method of exploit­
`ing the specificity and affinity of the antibody-antigen interaction for therapy.
`Several approaches have been suggested and attempted for overcoming the
`HAMA response to allow repeated administration of therapeutic mouse
`MAbs. One is to use conventional immunosuppressants at sub-toxic doses to
`immune response.
`reduce the patient's ability to mount an effective
`Co-treatment with corticosteroids and azathioprine, for example, was shown
`to delay and diminish the HAMA response to murine OKT3 (Chatenoud et
`al. , 1986). Similarly co-treatment with cyclosporin A was shown to reduce
`and delay the HAMA response of colorectal cancer patients treated with the
`radiolabelled anti-CEA antibody A5B7 (Ledermann et al. , 1988). The effect
`was sufficient to allow four doses of the conjugate to be administered.
`Tumour uptake of the conjugate increased with each dose in the cyclosporin
`co-treated patients but not in those who were not given the immunosuppres-
`
`11 of 142
`
`BI Exhibit 1067
`
`

`

`12
`A. MOUNTAIN AND J .R. ADAIR
`sant. Although the use of conventional immunosuppressants in these cases
`allowed more than a single administration the number of effective doses was
`still quite small. Such general immunosuppression is clearly undesirable in
`any case since it leads to the risk of infection and since the involvement of the
`patients immune system may well often be required in order to achieve cures.
`It is also possible that one consequence of general immunosuppression will be
`accelerated cancer prog