REVIEW
`Antibody-Enzyme Conjugates for
`Cancer Therapy
`
`Roger G. Melton, Roger F. Sherwood*
`
`Downloaded from
`
`http://jnci.oxfordjournals.org/
`
` by guest on November 24, 2014
`
`The use of antibody-enzyme conjugates directed at tumor-
`associated antigens to achieve site-specific activation of
`prodrugs to potent cytotoxic species, termed "antibody-
`directed enzyme prodrug therapy" (ADEPT), has attracted
`considerable interest since the concept was first described in
`1987. Prodrug forms of both clinically used anticancer
`agents and novel cytotoxic compounds have been developed
`to take advantage of potential prodrug-generating technol-
`ogy employing a variety of enzymes with widely differing
`substrate specificities. A particular advantage of the ADEPT
`approach is that it may allow the use of extremely potent
`agents such as nitrogen mustards and palytoxin, which are
`too toxic to be readily used in conventional chemotherapy.
`Preliminary studies using an antibody-enzyme conjugate
`constructed with a bacterial enzyme and a murine mono-
`clonal antibody not only have established the value of the
`ADEPT technique, but also have highlighted the potential
`problem of immunogenicity of proteins of nonhuman origin.
`This problem has been tackled in the first instance by the
`use of immunosuppressive agents, but long-term solutions
`are being investigated in the development of second-genera-
`tion ADEPT systems, including the development of human
`antibody-human enzyme fusion proteins and catalytic an-
`tibodies. Such improvements, coupled with further refine-
`ment of the prodrug-drug element of the system and the
`wide variety of antibody-enzyme-drug combinations avail-
`able, should mean that ADEPT-based approaches will form
`an important element of the search for the anticancer drugs
`of the future. [J Natl Cancer Inst 1996;88:153-65]
`
`The advent of monoclonal antibody technology in the 1970s
`was thought at the time to herald the fruition of the search for
`the "magic bullet" first proposed by Ehrlich in the early years of
`this century (/). In practice, however, this goal has remained
`elusive, inasmuch as many obstacles have presented themselves
`and have proved to be very difficult to overcome. To date, there
`have been numerous attempts to improve the cytotoxicity of the
`antibody "missiles" by attaching a variety of "warheads" to
`them—e.g., cytotoxic drugs such as doxorubicin or metho-
`trexate, toxins such as ricin A-chain or Pseudomonas exotoxin,
`and radioisotopes (2). Recently, enzymes capable of activating
`
`prodrugs to active drugs have been the focus of considerable in-
`terest and form the subject of this review.
`For any form of antibody-mediated targeting to be successful,
`it is axiomatic that there must be selective expression of the tar-
`get antigen by the tumor cells. This provides the first barrier to
`overcome, for the only well-characterized tumor-specific an-
`tigens described to date are the idiotypic determinants on the
`surface immunoglobulins of B-cell lymphomas (3). Tissue-
`specific antigens are also known, e.g., prostate-specific antigen
`(PSA), but have yet to be exploited to any great extent in tar-
`geted immunotherapy regimens. Many antigens, e.g., oncofetal
`antigens such as carcinoembryonic antigen (CEA) or placental
`alkaline phosphatase, and other potential targets such as growth
`factor receptors are present in elevated levels in tumor tissue,
`but most, if not all, appear to be found to a greater or lesser ex-
`tent in other tissues. These limitations have not prevented ex-
`ploitation of the better characterized tumor-associated antigens
`(e.g., CEA) as targets for therapy, even though it is also well
`documented that tumor cells exhibit considerable antigenic
`heterogeneity (4-6). Tumor cells may express varying levels of
`antigen, and some cells may not express the target antigen at all,
`such that a single monoclonal antibody conjugate targeted at a
`tumor mass will not bind to all the cells present, although other
`antibodies may be used to target alternative antigens on such
`cells. This is a disadvantage for immunoconjugates constructed
`using cytotoxic moieties that need to bind to the target cells and
`be internalized to exert their effect; however, the problem may
`be overcome to some extent by the use of antibody "cocktails"
`in which a mixture of antibodies aimed at two or more discrete
`antigens is used in noncompetitive combination (7).
`The ability of conjugates to gain access to the target cells is
`also essential, yet there is ample evidence that antibodies
`penetrate tumors poorly (8,9). Thus, the best access is at the
`tumor periphery, and cells located near the center of the tumor
`mass may be inaccessible to immunoconjugates, even though
`
`* Affiliations of authors: R. G. Melton, Centre for Applied Microbiology and
`Research, Porton Down, Salisbury, U.K.; R. F. Sherwood, Duramed Europe
`Ltd., Magdalen Centre, Oxford Science Park, Oxford, U.K.
`Correspondence to: Roger G. Melton, Ph.D., Centre for Applied Microbiol-
`ogy and Research, Porton Down, Salisbury SP4 OJG, U.K.
`See "Notes" section following "References."
`
`Journal of the National Cancer Institute, Vol. 88, No. 3/4, February 21, 1996
`
`REVIEW
`
`153
`
`IMMUNOGEN 2172, pg. 1
`Phigenic v. Immunogen
`IPR2014-00676
`
`

`

`less levels of the latter are particularly high (11). Another source of
`antigen may arise from the lysis of tumor cells following the active
`drug release from the prodrug by the action of the enzyme in the
`antibody-enzyme conjugate bound to cell surfaces. However, it is
`unlikely that there is any antibody-enzyme conjugate in the cir-
`culation at that time; thus, such antigen release may not be
`problematic except in situations where repeat cycle(s) of therapy
`are administered before this antigen is cleared from the body.
`A further advantage of ADEPT is that a single enzyme
`molecule has the potential to cleave many prodrug molecules—
`up to 800 mol/mol of enzyme per second in the case of the ben-
`zoic acid mustard substrates of carboxypeptidase G2 (13)—
`providing an amplification effect giving high levels of drug lo-
`calized at the tumor; this may be an important advantage in the
`clinic, in view of the typically low localization of immunocon-
`jugates in humans (14). There is also some evidence that high
`levels of drug generated at the surface of tumor cells are more
`effective than equivalent concentrations of free drug (75). The
`prodrug is an integral component of ADEPT systems and re-
`quires careful design in its own right. An ideal prodrug would
`be one with a large differential in cytotoxicity between drug and
`prodrug, which is a good substrate for the enzyme under
`physiologic conditions and for which there is no mammalian
`homologue capable of performing the same reaction. The use of
`human enzymes requires qualification of the latter statement, as
`discussed below. Equal cytotoxicity of the released active drug
`toward proliferating and quiescent cells is also desirable if
`residual deposits of viable but nonproliferating cells with the
`potential for outgrowth are to be eradicated. Once formed, it
`would be desirable for the drug to have a very short half-life,
`limiting the possibility of escape of active drug back into the cir-
`culation and access to healthy tissue. Development of drug
`resistance may limit the effects of conventional active drugs
`produced from prodrugs; thus, the nitrogen mustard group of
`compounds has tended to be the most commonly used group of
`active drugs for ADEPT because these compounds are not cell
`cycle specific, can be hydrolyzed to nontoxic forms, can kill
`both well-oxygenated and hypoxic cells, and cells develop only
`low levels of resistance to them (16). The potential release of
`the active drug into the circulation is conceivable after tumor
`cell lysis, but most drugs proposed for ADEPT systems are
`potent alkylating or intercalating agents and are likely to remain
`bound irreversibly to intracellular targets.
`In selecting an enzyme, one would look for activity under
`physiologic conditions, low immunogenicity, and little or no
`equivalent endogenous enzyme in humans. The enzyme needs
`to have a high specific activity, because the amounts that can be
`delivered by antibody vectors are limited; however, the defini-
`tion of optimal enzyme kinetics is clear cut and, as we shall see,
`under certain circumstances, a low kcat (i.e., the turnover rate of
`enzyme, which is expressed as number of moles of substrate
`utilized per mole of enzyme per second) may be preferable to a
`higher kcat (17). If a human enzyme is selected in an attempt to
`develop a nonimmunogenic conjugate, it will probably need to
`be an intracellular enzyme so that endogenous enzyme does not
`give rise to activation of the prodrug in the blood. When
`coupled to an antibody, the enzyme must localize efficiently at
`the tumor, and the development of immune responses to non-
`
`Downloaded from
`
`http://jnci.oxfordjournals.org/
`
` by guest on November 24, 2014
`
`they may express antigen. Cells in such inaccessible anoxic
`zones may be quiescent but possess the ability to regrow if ex-
`posed to a supply of oxygen and nutrients as a result of more
`vulnerable cells in oxic zones being killed. For this reason,
`radioimmunoconjugates have been extensively studied, because
`the emissions from the isotopes used clinically are capable of
`penetrating several cell layers, killing cells distal from where the
`antibody is localized. Although this approach is attractive, in
`practice the relatively low tumor uptake of the radioimmuno-
`conjugate may require large patient exposures, giving rise to
`dose-limiting myelosuppression and necessitating the use of
`autologous bone marrow grafts {10).
`
`Antibody-Directed Enzyme Prodrug Therapy as
`an Alternative Strategy
`
`An alternative strategy that has recently been the subject of
`considerable interest is the use of antibodies as vectors for en-
`zymes capable of activating a nontoxic drug precursor, termed a
`"prodrug," to a potently cytotoxic moiety. This approach, which
`has been termed "antibody-directed enzyme prodrug therapy"
`(ADEPT) (11), or "antibody-directed catalysis" (72), offers the
`potential to overcome most of the problems associated with the
`use of drug or toxin immunoconjugates described above. The
`system is shown by the diagram in Fig. 1. An antibody-enzyme
`conjugate is injected and allowed to localize at the tumor while
`clearing from the rest of the tissues. After a suitable interval, the
`prodrug is administered; while remaining innocuous to the nor-
`mal tissues, it is converted to a cytotoxic form by enzyme local-
`ized within the tumor tissue. Being of low molecular weight, the
`active drug can diffuse to adjacent tissues, providing a bystander
`effect, although this may be of questionable benefit if the
`released drug is relatively stable and can diffuse back into the
`circulation. It is implicit in the ADEPT approach that the an-
`tigen to which the antibody-enzyme conjugate is targeted must
`remain extracellular; thus, secreted antigen that accumulates in
`the tumor interstitial spaces may assist in achieving high levels
`of conjugate at the tumor site. Although shed antigen present in
`the circulation will act as a competitor with the tumor site for
`binding of target antigen, there is evidence that tumor localiza-
`tion of conjugate can occur in the face of circulating antigen un-
`
`Anti body-enzyme
`conjugate
`
`Fig. 1. Schematic representation of antibody-directed enzyme prodrug therapy.
`
`154 REVTEW
`
`Journal of the National Cancer Institute, Vol. 88, No. 3/4, February 21. 1996
`
`IMMUNOGEN 2172, pg. 2
`Phigenic v. Immunogen
`IPR2014-00676
`
`

`

`human enzymes is undesirable, since localization of the con-
`jugate is likely to be drastically reduced upon readministration
`after the onset of an immune response (18) and there will be a
`risk of hypersensitive reactions. However, immunosuppressive
`drug regimens designed to extend the time window before the
`onset of the human immune response observed in patients (19)
`have been developed to counteract this problem (20). The gains
`obtained using such approaches in animal models have been
`modest, but they may be sufficient to allow two or three repeat
`courses of an ADEPT cycle.
`An important factor in early stages of the development of an
`ADEPT system is the production of an antibody-enzyme con-
`jugate without loss of antigen-binding and enzymatic activities.
`The techniques and heterobifunctional agents typically used are
`analogous to those developed for use in the construction of im-
`munotoxins. An important requirement for conjugates prepared
`for use in ADEPT is that the linkage should be stable, since a
`period of up to or greater than 72 hours may elapse between the
`administration of conjugate and the administration of prodrug,
`in order to allow unbound conjugate to clear from the circula-
`tion. For this reason, conjugation chemistry based on thioether
`linkages has been virtually universally adopted.
`A number of experimental ADEPT systems have been de-
`veloped, with a range of different enzymes, and it is thus con-
`venient to categorize them by class of enzyme (Table 1).
`
`Enzymes used for ADEPT systems to date include carboxypep-
`tidases G2 and A, alkaline phosphatase, glycosidases, penicillin
`amidase, P-lactamase, and cytosine deaminase. All are at a very
`early stage of development, with only the carboxypeptidase G2
`system having been tested clinically to date. Typically, these
`systems have been tested in nude mouse models with human
`xenograft targets. These represent probably the best models cur-
`rently available, but they have clear limitations. For example,
`the subcutaneous tumor implant is nonmetastatic and, therefore,
`does not represent the typical clinical situation, where the treat-
`ment of advanced metastatic disease is likely to be the early tar-
`get, with the eventual prospect of moving to minimal residual
`disease following surgery to debulk the tumor. Moreover, the
`levels of tumor localization achievable in mice far exceed those
`that have been found clinically with radiolabeled antibodies. The
`fact that .nude mice are immunodeficient also prevents considera-
`tion of the problem of immunogenicity, an issue of considerable
`importance, as has been revealed by early clinical studies (21,22).
`Once a workable system has been established,.there is likely to be a
`lengthy development stage while all aspects of the system are re-
`evaluated and optimized. Typically, the necessary development
`work will include the synthesis of improved prodrugs, substitution
`of human enzymes for bacterial enzymes where possible, the adop-
`tion of chimeric or human antibodies and fragments thereof, the
`generation of nonimmunogenic clearing systems, and the develop-
`
`Table 1. Some potential antibody-directed enzyme prodrug therapy (ADEPT) systems described in the literature
`
`Downloaded from
`
`http://jnci.oxfordjournals.org/
`
` by guest on November 24, 2014
`
`Reference No.
`
`(13,27)
`
`(13.19,26.27)
`
`(I9.21.28J0)
`
`(153437)
`
`(42,44)
`
`(42)
`
`(45)
`(48)
`
`(46)
`
`(47)
`
`(5839)
`
`(61)
`(4930)
`
`(54)
`
`(79)
`
`(12.67.68)
`
`(65.66)
`
`(69-72)
`
`Enzyme
`
`Antibody
`
`Antigen
`
`Prodrug
`
`Active drug
`
`Carboxypeptidase G:
`
`WI4
`
`Human chorionic
`gonadotropin
`
`p-N-bis (2-Chloroethyl)benzoyl
`glutamic acid
`4-[(2-Mesyloxyethyl)(2-chloroethyl)
`amino] benzoyl glutamic acid
`
`Carboxypeptidase G2
`
`A5B7
`
`Carcinoembryonic
`antigen
`
`4-[(2-Mesyloxyethyl)(2-chloroethyl)
`amino] benzoyl glutamic acid
`
`p-N-bis (2-Chloroethyl)
`amino baaenzoic acid
`4-[(2-Mesyloxyethyl)-
`(2-chloroethyl)
`amino| benzoic acid
`4-[(2-Mesyloxyethyl)-
`(2-chloroethyl)
`amino benzoic acid
`Methotrexate
`
`Etoposide
`
`Mitomycin C
`
`Doxorubicin
`Mitomycin C
`
`In vitro cell line
`(in vivo tumor model)*
`
`CC3 choriocarcinoma
`(CC3)
`
`LS174T colon carcinoma
`(LSI74T)
`
`UCLA-P3 lung
`adenocarcinoma
`H3347 colon carcinoma
`(H3347)
`H298I lung adenocarcinoma
`(H2981)
`H298I lung adenocarcinoma
`L540 Hodgkin's lymphoma
`
`Methotrexate-a-alanine
`
`Etoposide phosphate
`
`Mitomycin C phosphate
`
`Doxorubicin phosphate
`Mitomycin C phosphate
`
`Carboxypeptidase A
`
`KS1/4
`
`Alkaline phosphatase
`
`L6
`
`L6
`
`L6
`HRS-3/API
`
`L6
`
`BW431
`
`Penicillin amidase
`
`L6
`
`p-Glucuronidase
`
`RHI
`
`323/A3
`
`BW431
`
`P-Lactamase
`
`CEM2314
`
`Cytosine deaminase
`
`L6
`
`L6
`
`Lung adenocarcinoma-
`associated antigen
`Tumor-associated
`carbohydrate
`
`CD30 HodgkirTs
`lymphoma antigen
`
`Carcinoembryonic
`antigen
`Lung adenocarcinoma-
`associated antigen
`
`Uncharacterized
`hepatocarcinoma
`antigen
`Tumor-associated
`glycoprotein
`Carcinoembryonic
`antigen
`Carcinoembryonic
`antigen
`
`*A11 cell lines tested are of human origin.
`
`/VJV-bis(2-Chloroethyl)aminophenyl
`phosphate
`Etoposide phosphate
`
`WJV-bis-(2-Chloroethyl)-
`4-hydroxyaniline
`Etoposide
`
`H2981 lung adenocarcinoma
`(H298I)
`S W1398 colon carcinoma
`
`/V-Phenylaceiamido doxorubicin
`/V-Phenylacetamido melphalan
`A'-(4'-Hydroxyphenylacetyl)palytoxin
`p-bis-2-Chloroelhylaminophenyl
`p-D-Glucopyranoside uronic acid
`(K>«-bulyl salt)
`Epimbicin-glucuronide
`
`Doxorubicin-glucuronide
`
`Cephalosporin-vinca alkaloid
`
`Cephalosporin mustard
`
`5-Fluorocytosine
`
`Doxorubicin
`Melphalan
`Palytoxin
`WJV-bis-(2-Chloroethyl)-
`4-hydroxyaniline
`
`H2981 lung adenocarcinoma
`
`H2981 lung adenocarcinoma
`HepG2 hepatoma
`
`Epirubicin
`
`Doxorubicin
`
`Desacetylvinblastine
`hydrazide
`Phenylene diamine
`mustard
`Fluorouracil
`
`MCF-7 breast cancer
`OVCAR-3 ovarian cancer
`(L0V0, Mz-Sto-l)
`colon carcinoma
`LS 174 colon carcinoma
`(LS174T)
`H298I lung adenocarcinoma
`
`H2981 lung adenocarcinoma
`(H2981.HT29 colon
`carcinoma)
`
`Journal of the National Cancer Institute, Vol. 88, No. 3/4, February 21, 1996
`
`REVIEW 155
`
`IMMUNOGEN 2172, pg. 3
`Phigenic v. Immunogen
`IPR2014-00676
`
`

`

`Downloaded from
`
`http://jnci.oxfordjournals.org/
`
` by guest on November 24, 2014
`
`ment of fusion proteins or bifunctional catalytic antibodies with
`suitable binding specificity for tumor-associated antigens. With
`such a range of parameters to be optimized, the development of
`ADEPT systems is clearly complex and likely to be time-con-
`suming and expensive.
`
`Enzymes and Prodrugs Used in Currently
`Available ADEPT Systems
`
`Carboxypeptidase G2
`
`The first ADEPT system was proposed by Bagshawe (25) in
`1987; it used carboxypeptidase G2 (CPG2), which was originally
`isolated as a methotrexate-degrading enzyme from Pseudo-
`monas species and subsequently cloned in Escherichia coli (24),
`to cleave a deactivating glutamate moiety from a benzoic acid
`mustard (25). This remains probably the best characterized sys-
`tem and is the only system to date for which pilot clinical
`studies have commenced (19,21,22). The initial prodrug to be
`synthesized was 4-[A'^V-bis(2-chloroethyl)amino]benzoyl-L-glu-
`tamic acid from which CPG2 cleaves the glutamic acid to yield
`4-[/vVV-bis(2-chloroethyl)amino]benzoic acid (Fig. 2, A). Sub-
`sequently, modified versions were developed with one or both
`of the 2-chloroethyl arms replaced by more reactive 2-mesyl-
`oxyethyl moieties (Fig. 2, B) (25). The mono-(2-mesyloxyethyl)
`(MMC1) compound was more cytotoxic and a slightly better
`substrate for the enzyme than the bis(2-chloroethyl) compound,
`but the bis(2-mesyloxyethyl) form was too reactive and rapidly
`hydrolyzed in the absence of enzyme (25,26). The enzyme con-
`jugate-prodrug combination was initially tested in vitro using
`the human chorionic gonadotropin
`(hCG)-expressing JAR
`choriocarcinoma cell line; it was found to give greater than 100-
`fold differential toxicity between drug or prodrug plus enzyme
`(IC50 [i.e., the concentration required to kill 50% of treated cells
`
`compared with untreated controls] = 20 UJW) and prodrug (IC50
`= >800 \iM) (27). In antitumor studies using the hCG-expressing
`human CC3 choriocarcinoma xenograft model in nude mice, the
`combination of W14 (anti-hCG) F(ab')2-CPG2 conjugate followed
`72 hours later by the MMC1 prodrug resulted in complete eradica-
`tion of established CC3 xenografts when prodrug was administered
`as a three-times-divided dose of 10 mg/kg (total dose, 30 mg/kg) at
`16-hour intervals from 72 hours (1327). In contrast, weekly in-
`travenous injections of daily divided doses of methotrexate (5
`mg/kg), hydroxyurea (50 mg/kg), dactinomycin (7.5 |ag/kg), or
`cytarabine (20 mg/kg) failed to retard tumor growth, whereas the
`MMC1 prodrug and the W14 F(ab')2-CPG2 conjugate produced a
`pronounced growth delay of more than 50 days (27).
`When the same two-phase system was applied to the CEA-
`expressing human LS174T colon carcinoma model, using a con-
`jugate constructed with the F(ab')2 fragment of A5B7 anti-CEA
`antibody, it proved to be less effective in that minimal growth
`delays were seen (28). This was probably because of the much
`slower clearance of conjugate from the blood in this model com-
`pared with that in the choriocarcinoma. In the latter case, high
`levels of circulating antigen probably form immune complexes,
`leading to accelerated clearance; however, much lower levels of
`antigen are present in the circulation of CEA-expressing tumors
`(29), and the clearing effect of free antigen is thus not apparent.
`As a result, a second galactosylated clearing antibody, SB43,
`which enzymatically inactivated CPG2, was used to remove un-
`bound conjugate from the circulation, making the system a
`three-phase approach (30). The galactosylated antibody has a
`relatively short half-life in the circulation because it is rapidly
`cleared by hepatic galactose receptors; consequently, it can react
`with the CPG2 conjugate in the circulation but is not ex-
`travasated sufficiently to reach the conjugate in the extravas-
`cular compartment. This situation should avoid the problem of
`
`COOH
`|
`O-NH-CH
`
`Carboxypeptidase G2
`»•
`
`OOH
`
`CICH2CH/
`
`\=
`
`4-[bis(2-chloroethyl)amino]benzoyl-
`L-gljtamic add
`
`CH2
`I
`COOH
`
`4-[bis(2-chloroethyl)amino]benzoicacid
`
`B CH3SO2O-CH2CH2
`
`CICH2CH/
`
`\=
`
`4-[(2-chloroethyl)(2-mesyloxyethyl)amino]benzoyl-L-glutamic acid (MMCI)
`
`4-[bis(2-mesyIoxyethyt)aminolbenzoyl-L-glutamicacid
`
`COOH
`O-NH-CH
`H2
`CH2
`COOH
`
`IC
`
`CH3SO2O-CH2CH2
`
`CH3SO2O-CH2CH2'
`
`Fig. 2. Benzoic acid mustard prodrugs that are substrates for carboxypeptidase G, (CPG2). A) Mechanism of cleavage of benzoic acid mustards by CPG,. B) Alterna-
`tive mustard functional groups.
`
`156 REVIEW
`
`Journal of the National Cancer Institute, Vol. 88, No. 3/4, February 21, 1996
`
`IMMUNOGEN 2172, pg. 4
`Phigenic v. Immunogen
`IPR2014-00676
`
`

`

`Downloaded from
`
`http://jnci.oxfordjournals.org/
`
` by guest on November 24, 2014
`
`clearing antibody reaching and inactivating the tumor-bound an-
`tibody-enzyme conjugate (30). When the three-phase system
`was tested against the same model, prolonged growth delays of
`up to 10-15 days, compared with untreated controls, were seen
`with the MMC1 compound and, to a lesser extent, with the
`bischloro compound (28). Using a superficially identical model,
`Blakey et al. (31) were subsequently able to achieve growth
`delay of 10-15 days without recourse to clearing antibody. The
`principal difference between the experiments lay in the method
`of tumor implantation: Solid tumor fragments were used in the
`earlier work, whereas injection of a cell suspension was done in
`the later experiments.
`In an alternative approach to the use of clearing antibody,
`Melton et al. (32) demonstrated enhanced tumor localization of
`antibody-enzyme conjugate when administered with tumor
`necrosis factor-a (TNF-a) in the murine E3 thymoma model.
`Coadministration of 1.5 (ig TNF-a with conjugate led to a
`twofold increase in tumor localization and only transient in-
`creases in normal tissue localization. It is possible that this sys-
`tem could be used in combination with the galactosylated
`clearing antibody because the two approaches are complemen-
`tary: TNF-a increases tumor localization of conjugate without
`significantly affecting blood levels, and galactosylated antibody
`markedly decreases blood
`levels without affecting
`tumor
`localization. Galactosylated antibody increases the ratios of
`tumor to normal tissue without affecting overall tumor uptake,
`whereas TNF-a provides a mechanism for achieving higher ini-
`tial tumor uptake. Similar effects can be induced by the ad-
`ministration of drugs that modify blood flow, which essentially
`appear to act by inducing localized production of TNF-a (33).
`The preclinical studies using the colon carcinoma xenograft
`model and the MMC1 compound were deemed sufficiently en-
`couraging to justify a limited clinical trial in patients with advanced
`metastatic, poorly differentiated colon or rectal cancer (1921,22).
`Initial dose escalation studies using the monomesyl benzoic acid
`mustard prodrug showed that doses from 200 up to 2500 mg/m2,
`given as six or 12 divided doses over a 3-day period, were well
`tolerated. A total of four male patients aged 55-70 years, all of
`whom had undergone extensive conventional chemotherapy (prin-
`cipally with fluorouracil), subsequently went on to receive 20 000
`U/m2 A5B7 (anti-CEA) F(ab')2-CPG2 conjugate, a small propor-
`tion of which was I25I labeled to facilitate confirmation of tumor
`localization by gamma camera imaging. SB43 clearing antibody
`was administered 36-48 hours later as a 3- to 5-hour infusion, typi-
`cally followed by MMC1 (1200-3000 mg/m2) once measured
`serum enzyme levels were less than 0.02 U/mL. The serum levels
`of CEA and a second tumor antigen, 19-9, fell in all patients by 10-
`15 days after therapy; in two cases, there were objective responses
`in the reduction in size of liver metastases. A patient who had gross
`hepatomegaly and ascites became free of jaundice and ascites, al-
`though liver size was not reduced. Other subjective responses in-
`cluded relief of pain, and three patients gained weight and reported
`improved health. Adverse effects included the development of an-
`tibodies to both components of the conjugate, and the two patients
`receiving the highest doses of prodrug developed severe myelo-
`suppression, although this was reversible once prodrug administra-
`tion was ceased. The antibody responses developed within 10-12
`days, limiting the treatment effectively to a single cycle; however,
`
`this time window may be extended by the use of the im-
`munosuppressive drug cyclosporine (20). The development of
`myelosuppression appears to be due to the relatively long half-life
`of the active drug, which can diffuse out of the tumor and back into
`the circulation, indicating that more reactive drugs with shorter
`physiologic half-lives are desirable. The results of the pilot clinical
`study are encouraging, if one bears in mind the advanced state of
`the disease of the patients in the study, and provide ample justifica-
`tion for further development of the system.
`
`Carboxypeptidase A
`
`Huennekens and co-workers (15,34-37) described a system in
`which carboxypeptidase A (CPA) is used to cleave the alpha-
`carboxyl alanine residue from methotrexate-a-alanine. When
`human lung adenocarcinoma cells were exposed to conjugate
`consisting of CPA coupled to KS1/4 antibody, followed by
`washing and addition of prodrug, the IC50 of methotrexate-a-
`alanine decreased approximately sixfold from 8.9 x 10"6 M to
`1.5 x 10"6 M after 48 hours of incubation. With prolonged in-
`cubation, the toxicity of methotrexate-a-alanine approached
`that of methotrexate (i.e., about 2 x 10"9 M) (15). It is thus clear
`that the enzyme turns over the substrate rather slowly, although
`the authors suggested that methotrexate generated in intimate
`contact with the cells in this way seems to be more cytotoxic
`than the equivalent concentration of free methotrexate. A study
`of the theoretical aspects of ADEPT by Yuan et al. (17) sug-
`gested that a low turnover rate may not be disadvantageous be-
`cause the rate-limiting step is diffusion of drug from the
`capillaries to the extravascular tumor cells; however, this obser-
`vation must be set against the need to generate sufficient drug to
`exert the desired cytotoxic effect. The most important factor in
`attaining tumor-specific activation may be the achievement of
`high tumor-to-normal tissue localization ratios and rapid reac-
`tion of the active drug to prevent its loss back into the circula-
`tion. If these issues can be resolved, then the use of an enzyme
`with a high turnover and affinity for substrate would seem to be
`ideal because the level of prodrug reaching the tumor is unlikely
`to match that present in blood. It seems likely, however, that the
`relatively poor kinetics of this enzyme with the first-generation
`prodrug are unlikely to lead to a workable system in vivo, and a
`number of groups (38J9) recently described the synthesis of
`potentially better substrates based on methotrexate. No animal
`studies have been reported for the CPA/methotrexate-a-alanine
`system, and the likelihood of activation of prodrug by endoge-
`nous CPA remains as a possible disadvantage of this approach.
`In view of the development of resistance to methotrexate and
`the high levels of drug needed to achieve maximal effect (40), it
`is questionable whether this is a suitable drug for ADEPT ap-
`plications, particularly when the kinetics of the enzyme are rela-
`tively poor.
`
`Aminopeptidase
`
`Smal et al. (41) recently described the synthesis of a series of
`2-aminoacyl methotrexate prodrugs designed to be substrates for
`pyroglutamate aminopeptidase or D-aminopeptidase. The cyto-
`toxicity of the prodrug plus enzyme in vitro did not approach
`that of methotrexate alone and required high levels of enzyme to
`achieve any selective cell kill, suggesting that this is not a prac-
`
`Journal of the National Cancer Institute, Vol. 88, No. 3/4, February 21, 1996
`
`REVIEW 157
`
`IMMUNOGEN 2172, pg. 5
`Phigenic v. Immunogen
`IPR2014-00676
`
`

`

`Downloaded from
`
`http://jnci.oxfordjournals.org/
`
` by guest on November 24, 2014
`
`tical working system in its present form. To date, the production
`and testing of antibody-enzyme conjugates have not been
`described, and it is difficult to envisage the delivery of amounts
`of enzyme activity equivalent to those used in the experiments
`described, using an antibody vector.
`
`Alkaline Phosphatase
`
`A number of groups have presented evidence that alkaline
`phosphatase can be used to activate a range of prodrugs (Fig. 3),
`taking advantage of alkaline phosphatase's relative lack of sub-
`strate specificity. Human alkaline phosphatase, in conjunction
`with a humanized monoclonal antibody, may result in a con-
`jugate with reduced immunogenicity by comparison with bac-
`terial enzyme conjugates. However, endogenous alkaline
`phosphatase is widely distributed in the body and is likely to ac-
`tivate prodrug in normal tissues, suggesting that this system is
`primarily of academic interest. To test the concept of using
`alkaline phosphatase, Senter and co-workers used calf intestinal
`alkaline phosphatase to cleave phosphate groups from etoposide
`phosphate (42-44), mitomycin C phosphate (42), doxorubicin
`phosphate (45), and phenol mustard phosphate (46) and have
`reported antigen-specific cytotoxicity in vitro and growth delays in
`vivo with human colon carcinoma xenograft models; conjugates
`constructed with an irrelevant antibody showed greatly reduced
`growth delays, indicating that nonspecific uptake of conjugate is
`relatively low compared with the uptake of conjugate with
`specificity for the appropriate antigen. The use of this enzyme does
`not address the question of possible immunogenicity, but it would
`be expected to have kinetic properties similar to those of the human
`homologue. Other researchers (47,48) described similar systems
`
`using other antibodies directed toward human colon cancer an-
`tigens. Perhaps the most interesting of these describes the use of
`a bifunctional antibody capable of binding both alkaline phos-
`phatase and the CD30 antigen; the purpose was to enable the
`production of conjugates by immunologic interaction rather than
`by chemical coupling (48). Such an approach requires an enzyme
`that is not catalytically inactivated by the antibody binding to it; in
`this instance, the authors were fortunate that the binding of alkaline
`phosphatase by bispecific intact antibody or F(ab')2 fragment did
`not adversely affect the enzyme's ability to activate the phosphory-
`lated mitomycin C prodrug.
`
`Glycosidases
`
`A number of groups have described ADEPT systems based
`on glycosidases (Fig. 4). Roffler and co-workers (49£0) used (3-
`glucuronidase from Escherichia coli to remove a glucuronide
`from hydroxyaniline mustard. This system takes advantage of
`the fact that hydroxyaniline mustard is rapidly detoxified by
`hepatic enzymes by conversion to the glucuronide form (51,52).
`In certain animal tumors, hydroxyaniline mustard glucuronide
`was activated back to hydroxyaniline mustard by elevated levels
`of endogenous (3-glucuronidase to give site-specific cytotoxicity
`(57), although attempts to use this system in humans were not
`successful (53).
`Escherichia coli (3-glucuronidase, coupled to monoclonal an-
`tibody 12.8 that bind