`
`http://science.sciencemag.org/
`
`
`
`on January 19, 2018
`
`Monoclonal Antibodies in
`Diagnosis and Therapy
`
`THOMAS A. WALDMANN
`
`Monoclonal antibodies have been applied clinically to the
`diagnosis and therapy of an array of human disorders,
`including cancer and infectious diseases, and have been
`used for the modulation of immune responses. Effective
`therapy using unmodified monoclonal antibodies has,
`however, been elusive. Recently, monoclonal antibody-
`mediated therapy has been revolutionized by advances
`such as the definition of cell-surface structures on abnor-
`mal cells as targets for effective monoclonal antibody
`action, genetic engineering to create less immunogemc
`and more effective monoclonal antibodies, and the arming
`of such antibodies with toxins or radionucides to en-
`hance their effector function.
`
`T HE DEVELOPMENT OF MONOCLONAL ANTIBODY TECHNOL-
`ogy by Kohler and Milstein (1) provided an enormous
`opportunity for examination of a range of previously elusive
`issues. For example, monoclonal antibodies are being used in
`radioimmunoassays, enzyme-linked immunosorbent assays, immu-
`nocytopathology, and flow cytometry for in vitro diagnosis, and in
`vivo for diagnosis and immunotherapy of human disease. However,
`in the area of immunotherapy, monoclonal antibodies are just
`beginning to fulfill the promise inherent in their great specificity for
`recognizing and selectively binding to antigens on cells.
`Monoclonal antibodies have largely been applied clinically to the
`diagnosis and therapy of cancer and the modulation of the immune
`response to produce immunosuppression for treatment of autoim-
`mune and graft versus host diseases (GVHD) and for prevention of
`allograft rejection. Human monoclonal antibodies have also been
`applied clinically against cytomegalovirus, Varicella zoster virus, and
`the various specific serotypes of Pseudomonas aeruginosa, Escherichia
`coli, and Klebsiella pneumoniae. For example, in a multicenter
`clinical trial involving 200 patients with Gram-negative bacteremia,
`mortality was reduced in patients who received an immunoglobulin
`M (IgM) monoclonal antibody (HA-LA) that binds specifically to
`the lipid A domain of endotoxin (2). In other studies, monoclonal
`antibodies that are specific for leukocyte adhesion molecules, such as
`lymphocyte functional antigen-1 (LFA-1) or intercellular adhesion
`molecule-i (ICAM-1), the adhesion partner of LFA-1, have been
`used to inhibit accumulation of neutrophils and thereby reduce
`tissue damage in animal models of bacterial meningitis, hemorrhagic
`shock, and myocardial reprofusion injury (3).
`The use of monoclonal antibodies has also been proposed for
`therapy of myocardial infarction (4), for reversal of drug toxicity
`(digitalis intoxication), and for fertility control. Despite this wide-
`
`The author is in the Metabolism Branch, National Cancer Institute, National Institutes
`of Health, Bethesda, MD 20892.
`
`21 JUNE 1991
`
`ranging interest, the "magic bullet" of antibody therapy that has
`been the dream of immunotherapists since the time of Paul Ehrlich
`has proved to be elusive (5). Only one monoclonal antibody, OKT3,
`has been licensed for clinical use. Furthermore, the initial use of
`unmodified murine monoclonal antibodies in human patients with
`cancer was disappointing, with only 23 partial and 3 complete
`remissions reported among the initial 185 patients included in 25
`clinical trials (6). A number of factors explain the low therapeutic
`efficacy observed. Unmodified murine monoclonal antibodies are
`immunogenic and elicit a human immune response to the murine
`antibodies. Moreover, most mouse monoclonal antibodies are not
`cytocidal against neoplastic cells in humans because the antibodies
`do not participate in human complement or cell-mediated cytotox-
`icity. In most cases, the antibodies were not directed against a vital
`cell-surface structure such as a receptor for a growth factor required
`for tumor cell proliferation. In an attempt to circumvent these
`problems, researchers have developed human and humanized anti-
`bodies, prepared by genetic engineering, that are less immunogenic
`than murine antibodies. Cytotoxic action has been augmented by
`arming the antibodies with toxins or radionuclides. Finally, cell
`surface antigenic targets have been defined for effective monoclonal
`antibody action.
`In this article I summarize information on in vivo use of mono-
`clonal antibodies for diagnosis and therapy, with special reference to
`the most seminal discoveries and recent advances in (i) definition of
`cell-surface structures on abnormal cells as targets for effective
`monoclonal antibody action, (ii) development of genetic engineer-
`ing approaches for creating more effective agents, and (iii) develop-
`ment of techniques for arming monoclonal antibodies with radio-
`effector function.
`nucides or toxins and thereby increasing
`Comprehensive recent reviews provide more detail (7-9).
`
`Cell Surface Antigenic Targets
`The ideal immunosuppressive monoclonal antibody would be one
`that abrogates responses to a defined antigen and preserves respons-
`es to all others. However, the monoclonal antibody that is most
`widely used clinically (OKT3) is directed against the CD3 (duster of
`differentiation) antigen of the T cell receptor complex that is
`expressed on virtually all circulating T cells. The Food and Drug
`Administration licensed the use of OKT3 for the treatment of acute
`renal allograft rejection on the basis of randomized clinical trials
`showing the superiority of this treatment (93% reversal of acute
`rejection episodes) to more conventional, broad-spectrum immuno-
`suppressive agents (75% reversal) (10). There are, however, toxici-
`ties associated with the use of OKT3. Antibodies to CD3 initiate in
`vivo T cell activation that is accompanied by release of tumor
`necrosis factor, interferon -y, and, in some cases, interleukin-2
`(IL-2), which leads to an acute clinical syndrome that involves high
`fever, vomiting, diarrhea, and occasionally respiratory distress.
`1657
`
`ARTICLES
`
`Pfizer v. Genentech
`IPR2017-01489
`Genentech Exhibit 2054
`
`
`
`Downloaded from
`
`http://science.sciencemag.org/
`
`
`
`on January 19, 2018
`
`Furthermore, therapy with antibody to CD3 leads to broad immu-
`nosuppression that is associated with an increased incidence of
`infections and B cell neoplasms.
`Monoclonal antibodies have been used to target the polymorphic
`a and 1 subunits of the human T cell antigen receptor (TCR). In
`pilot studies, one monoclonal antibody (BMA 031) specific for the
`constant region of the TCR was used successfully in treatment of
`acute organ allograft rejection and GVHD (11). Monocdonal anti-
`bodies that specifically identify the hypervariable antigen-binding
`region of the TCR represent more antigen-specific agents. Janson
`and co-workers (12) prepared a murine monoclonal antibody reac-
`tive with an idiotypic determinant of the hypervariable region ofthe
`TCR expressed by a patient's malignant leukemic cells. There was an
`80% reduction in the number of leukemic cells after infusion of this
`monoclonal antibody. However, associated toxic side effects includ-
`ed fever, chills, nausea, vomiting, diarrhea, and shortness of breath.
`Monoclonal antibodies directed toward idiotypic determinants of
`the TCR may also be useful in the treatment of autoimmune disease
`because it has been observed that there is an extreme restriction in the
`TCR variable region usage in the T cells responsible for experimental
`allergic encephalomyelitis (EAE), a murine model of multiple sdero-
`sis. The majority of T cells specific for myelin basic protein that can
`transfer the disease express a specific V138 phenotype (13). Further-
`more, in vivo treatment with a monodonal antibody to V138 provided
`protection from disease transfer by encephalitogenic donal T cells.
`Evidence for such restricted usage ofthe TCR variable region is being
`sought in spontaneous human neurological diseases, induding multi-
`ple sderosis and tropical spastic paraparesis.
`Most anti-idiotypic antibodies utilized in treatment of human
`lymphoid malignancies are directed toward B cell immunoglobulin
`idiotypes. Idiotypic monodonal antibodies to immunoglobulin
`were used (14) for treatment ofpatients with B cell lymphomas. The
`results were encouraging; one patient manifested an uninterrupted
`remission for more than 4 years after therapy. A partial response was
`observed in approximately half of the remaining patients. However,
`there were certain difficulties. During maturation, B cells use somatic
`hypermutation to increase their diversity. Thus, there is a tremendous
`idiotypic heterogeneity within the B cell population, which permits
`some neoplastic B cells to escape the attack of an antibody directed
`toward an individual idiotype. The TCR does not undergo somatic
`hypermutation and is less susceptible to this problem.
`Monoclonal antibodies that recognize surface molecules that
`facilitate cell-cell interactions are also effective immunosuppressive
`agents. For example, antibodies specific for LFA-1 or its adhesion
`partner, ICAM-1, inhibit a series of T cell functions and thereby
`inhibit allograft rejection and GVHD (15). The CD4 and CD8
`proteins, which are important in T cell activation, recognition of
`major histocompatibility complex gene products, and cooperative
`cellular interactions, are also useful targets. In animal models,
`antibodies to CD4 or CD8 suppressed allograft rejection, GVHD,
`and autoimmune reactions (16). Certain antibodies to CD4 are of
`special interest because they induce immunological tolerance in mice
`to simultaneously administered human, rat, and rabbit immuno-
`globulins. This form of tolerance was induced in T helper cells but
`not in B cells (17), and thus CD4 cells may be made tolerant when
`simultaneously confronted with certain antigens and monoclonal
`antibodies to CD4. Effective CD4 monoclonal antibody tolerance
`therapy is not confined to special protein antigens but is exploitable
`in transplantation. For example, short-term administration ofa mono-
`donal antibody to CD4 induced cellular depletion and led to indefi-
`nite survival of adult pancreatic islet cell transplants in mice without
`subsequent immunotherapy (18). Furthermore, administration of
`antibodies to CD4 ameliorated Type II collagen-induced arthritis in
`rats, murine EAE, and autoimmune myasthenia gravis (16).
`
`1658
`
`The monodonal antibodies to primate CD4 examined to date are
`not effective in tolerance induction in subhuman primate models.
`Despite this limitation, the administration of monoclonal antibodies
`to CD4 has led to improvement in preliminary clinical trials
`involving patients with severe rheumatoid or psoriatic arthritis (19).
`The inducible a chain of the interleukin-2 receptor (IL-2Ra)
`expressed on the surface of activated or abnormal T cells is also a
`target for effective monoclonal antibody immunotherapy (20).
`Interleukin-2 receptor (IL-2R) is a useful target because resting cells
`do not express the IL-2Ra, whereas the T cells participating in
`allograft rejection and abnormal T cells in certain autoimmune
`disorders express this receptor. To exploit this difference in expres-
`sion, we used anti-Tac, a monoclonal antibody that blocks the
`binding of IL-2 to IL-2Ra, in the treatment of patients with human
`T cell leukemia/lymphoma virus-I (HTLV-I)-associated, IL-2Ra-
`expressing adult T cell leukemia (ATL) (20). The 20 patients treated
`in this study did not suffer any untoward reactions. Of the 20 treated
`patients, seven had remissions, three of these partial, one mixed, and
`three complete, lasting from 1 to more than 17 months after
`anti-Tac therapy. Unmodified anti-Tac has also been used for
`prevention of early allograft rejection episodes in patients receiving
`renal allografts. The IL-2/IL-2R system offers a variety of other
`possibilities for relatively specific immune intervention strategies
`that will be considered below.
`
`Genetically Engineered Antibodies
`Although murmne antibodies are of value in therapy of human
`diseases, their effectiveness is limited because rodent monoclonal
`antibodies have a short survival time in humans and induce an
`immune response that neutralizes their therapeutic effect. Further-
`more, the responses induced by murine antibodies are limited
`because they only weakly recruit human effector elements and are
`relatively ineffective as cytocidal agents. To circumvent these diffi-
`culties, genetically engineered antibody variants were produced that
`combine the rodent variable or hypervariable regions with the
`human constant or constant and variable framework regions (21-
`25). The ability to genetically engineer antibodies represents a
`quantum leap in immune intervention that is comparable to the
`immunological revolution initiated by the introduction ofmonoclo-
`nal antibodies.
`After the demonstration that lymphoid cells can express cloned
`transfected immunoglobulin genes, mouse-human chimeric mono-
`clonal antibodies to tumors were generated with specificities direct-
`ed toward antigens expressed by colorectal, mammary, pancreatic,
`and B cell and T cell malignancies (21, 22). Jones et al. (23) proposed
`that because V domains represent a framework of P sheets topped
`with antigen-binding loops and because 1-framework structures of
`most crystallized antibodies are nearly invariant, the specificity ofthe
`antibody combining site might be independent of the framework
`region. Thus, to further reduce the immunogenicity of rodent
`elements, Winter and colleagues generated humanized antibodies
`that retained only the antigen-binding complementarity-determin-
`ing regions (CDRs) from the parent rodent monoclonal antibody in
`association with human framework regions (23, 24). Unfortunately,
`in some cases humanized antibodies produced by this approach have
`reduced binding affinity for antigen when compared to the original
`rodent antibody. Queen and co-workers (25) addressed this prob-
`lem in two ways. First, the human framework was chosen to be as
`homologous as possible to the original mouse antibody in order to
`reduce deformation of the transplanted mouse CDRs. Second,
`computer modeling was used to identify several amino acids in the
`mouse antibody framework that, although outside the CDRs, were
`SCIENCE, VOL. 252
`
`
`
`Downloaded from
`
`http://science.sciencemag.org/
`
`
`
`on January 19, 2018
`
`likely to interact with the CDRs or antigen. These specific amino
`acids were retained in the humanized antibody.
`One of the goals in the generation of humanized antibodies is the
`reduction oftheir immunogenscity. Although humanized antibodies
`are less immunogenic than their murine counterparts (26-28), the
`idiotypic element may be immunogenic. Furthermore, the presence
`of allotypes on the human immunoglobulin G (IgG) framework
`may provide foreign carrier determinants, thereby enhancing im-
`mune responses to the idiotypic element (29). In the case of
`anti-Tac, the humanized version was dramatically less immunogenic
`than the parent murine monoclonal antibody when administered to
`cynomolgus monkeys (26). Similarly, in the first clinical trial of a
`chimeric antibody, a murine human chimeric IgGI antibody specific
`for a gastrointestinal tumor antigen elicited an antibody response
`against the chimeric antibody in only one of ten patients studied
`(27). Furthermore, no anti-monoclonal antibody response was
`detected in two patients with non-Hodgkin's lymphoma studied
`who received CAMPATH-1, an antibody recognizing an antigen
`expressed by human lymphocytes (28). The pharmacokinetics of an
`antibody molecule may also be altered genetically. The survival time
`of a monoclonal antibody can be altered to increase its period of
`effective action or, alternatively, to accelerate its clearance. For
`example, the pharmacokinetics of radiolabeled humanized anti-Tac
`differed substantially from that of murine anti-Tac when adminis-
`tered to normal cynomolgus monkeys, with a prolongation of the
`survival half-time of humanized anti-Tac to 103 hours, as compared
`to 38 hours for murine anti-Tac (26). Prolonged survival of
`humanized monoclonal antibodies in humans was observed by
`LoBuglio and co-workers (27). The catabolic rate of an immuno-
`globulin is determined by the CH2 domain of the Fc region of the
`immunoglobulin. Thus, the longer survival of humanized antibodies
`observed probably reflects replacement of the murine IgG CH2
`domain with the human IgG CH2 domain (30, 31).
`Antibodies or their fragments can also be genetically engineered
`to have more rapid clearance. This might be desirable when a
`monoclonal antibody is conjugated to a radionuclide for use in
`radioimmunoscanning. For example, antigen-binding fragment
`(Fab), F(ab')2, or single chain Fv fragments of monoclonal anti-
`bodies have survival half-lives of less than 5 hours. Rapid turnover
`can also be accomplished by the deletion of the CH2 domain as
`demonstrated for an antibody reactive with the disaloganglioside
`GD2 expressed on human tumors of neuroectodermal origin (32).
`Effector functions can be improved by introduction of human
`constant regions that impart biological activity to a murine antibody
`that lacks effector function but has the desired binding specificity. The
`human IgG subclasses differ in their antitumor activity and in their
`capacity to induce complement or antibody-dependent cell-mediated
`cytotoxicity. The IgGl subclass appears to be superior to the other
`subclasses in most functions (33). The human IgGl versions of the
`CAMPATH-1 antibody and the L6 monodonal antibody to a carci-
`noma-associated antigen were more effective in antibody-dependent
`cell-mediated cytotoxicity (ADCC) than the parent rodent antibodies
`(8, 24). Furthermore, the humanized version of anti-Tac induces
`ADCC with human mononudear cells, a function absent from the
`original mouse monoclonal antibody (34).
`
`Human Monoclonal Antibodies
`One solution to the problems of immunogenicity and poor
`recruitment of effector functions characteristic of rodent monoclon-
`als is to produce human monoclonal antibodies. Human antibodies
`of appropriate specificity and of high affinity have been difficult to
`isolate. The use of mouse myeloma cells as fusion partners for
`
`21 JUNE 1991
`
`human cells often leads to preferential loss of human chromosomes
`and instability of the hybrids. For ethical reasons one cannot
`immunize humans with certain tumor antigens. An alternative, the
`immortalization of human cells by Epstein-Barr virus, often gener-
`ates lines that produce only low amounts of IgM-type antibodies.
`Human antibodies may be produced more easily with the use of
`SCID-hu mice (35), that is, immunodeficient mice reconstituted by
`human peripheral blood or human fetal thymus, bone marrow, and
`lymph nodes. When peripheral blood is used, antibodies may be
`produced if the donor has already been primed with antigen. Such
`antibodies have not yet been used in therapeutic trials.
`An alternative approach to the production of human monoclonal
`antibodies has been reported that bypasses hybridoma technology
`(7, 36, 37). The immunoglobulin V-region genes from B cells were
`cloned with the use of the polymerase chain reaction technique. The
`antibody derivatives were then expressed in E. coli and screened for
`ability to bind antigen. Initially, heavy chain V regions were
`expressed alone or with an irrelevant light chain V region. However,
`a large combinatorial library of the immunoglobulin repertoire of
`the mouse in phage lambda has now been generated (36). Heavy
`and light chain libraries were prepared in phage lambda and used to
`generate a large array of random heavy plus light chain pairs
`expressed in bacteria in the form of Fab molecules. The screening for
`binding of antigens to hapten was rapid and permitted the analysis
`of many monospecific Fab-producing clones. In similar studies,
`Mullinax and co-workers (37) identified human antibody fragment
`clones specific for tetanus toxoid in a bacteriophage lambda immu-
`noexpression combinatorial library prepared using messenger RNA
`derived from human peripheral blood lymphocytes. The bacterio-
`phage clones are directly amenable to genetic manipulation for
`preparing complete immunoglobulins of the desired isotype. Fur-
`ther studies will be required to determine whether this approach will
`allow the efficient identification of human monoclonal antibodies of
`sufficiently high affinity for clinical studies.
`
`Biflmctional Antibodies
`Antibodies with two distinct binding activities have been gener-
`ated to deliver radionuclides, toxins, cytotoxic drugs, or host
`cytotoxic cells to specific cellular targets (38). Their predominant use
`has been to direct cytotoxic cells to target and lyse cells that they
`normally would not lyse. Such bifunctional antibodies have been
`prepared by chemical cross-linking, disulfide exchange, or the
`production of hybrid-hybridomas (quadromas). Bispecific antibod-
`ies have also been produced by introduction of two sets of immu-
`noglobulin heavy and light chains into myeloma cells or by con-
`struction of single-peptide bispecific antibodies with the use of
`peptide linkers between the variable domains of two distinct mono-
`clonal antibodies. To be effective, the bispecific antibody must
`retarget the cytotoxic cell from its natural ligand to the ligand
`identified by the monoclonal antibody. Furthermore, the antibody
`must activate the cytotoxic cells into functional effectors without the
`normal major histocompatibility complex- and antigen-specific re-
`strictions. The most effective bispecific monoclonal antibodies have
`the CD3 antigen on cytotoxic T cells or the CD16 Fc y R III
`receptor on natural killer cells as their nonantigen specificity.
`Bispecific antibodies with specificity against both tumor targets and
`CD3 or CD16 effector cells are effective in mediating the killing of
`tumor cells in vitro and in vivo. For example, murine anti-Tac does
`not participate in ADCC with human mononuclear cells. In con-
`trast, murine anti-Tac-anti-CD3 and anti-Tac-anti-CD16 bifunc-
`tional agents used in conjunction with peripheral blood mononu-
`clear cells killed targets that express IL-2 receptors. A universal
`1659
`ARTICLES
`
`
`
`Downloaded from
`
`http://science.sciencemag.org/
`
`
`
`on January 19, 2018
`
`bispecific antibody for retargeting effector cells to tumor cells can be
`generated with the use of a bispecific hybrid antibody with dual
`specificities for CD3 and for a rat immunoglobulin light chain
`allotype (39). This bispecific antibody mediates retargeting of
`effector cells to a range of tumor cells, each coated with rat
`monoclonal antibodies bound to surface antigens.
`Human T cells that had been retargeted by bifunctional antibodies
`were used to treat established human ovarian carcinoma in a nude
`mouse model (40). Peripheral blood lymphocytes from patients
`were incubated overnight with IL-2 and treated with heteroconju-
`gates containing antibody to CD3 cross-linked to an appropriate
`antibody to tumor and were injected interperitoneally into tumor-
`bearing mice. Tumor growth was inhibited. Bispecific antibodies
`have also been used in nontumor systems. For example, thrombol-
`ysis was enhanced by targeting oftissue plasminogen activator (tPA)
`by bispecific antibodies to tPA and fibrin (4). Thus, although
`problems associated with their manufacture remain to be resolved,
`bispecific antibodies show great promise as therapeutic agents.
`
`Monoclonal Antibody-Cytotoxic Agent
`Conjugates
`The limited efficacy of many unmodified monoclonal antibodies
`led to an alternative approach, the use of these agents as carriers of
`cytotoxic substances. An array of toxins of bacterial and plant origin
`have been coupled to monoclonal antibodies for production of
`immunotoxins (8, 41). The strategy is to select from nature a toxic
`protein and then to modify the toxin so that it will no longer
`indiscriminately bind and kill normal cells but will instead kill only
`the cells expressing the antigen identified by the monoclonal anti-
`body. The majority of toxins targeted to cell surfaces by immuno-
`conjugates act in the cytoplasm, where they inhibit protein synthe-
`sis. After binding to cell surface antigens, immunotoxins are taken
`up by endocytosis and delivered to endosomes. Fragments of some
`toxins (for example, diphtheria toxin) are then translocated across
`the membrane of this organelle. Other immunotoxins (for example,
`ricin) are routed further to the trans-Golgi network, where a
`minority undergo translocation to the cytoplasm. Unfortunately,
`most are routed to lysosomes, where they are degraded. In the
`cytoplasm, the toxins used clinically act either to adenosine diphos-
`phate (ADP)-ribosylate elongation factor 2 (for example, Pseudo-
`monas exotoxin) or to inactivate the 60S ribosomal subunit so that
`it has a decreased capacity to bind elongation factor 2 (for example,
`ricin). Less than ten toxin molecules in the cytoplasm are sufficient
`to kill the cell; however, more must bind to the cell surface to
`compensate for the inefficiencies in internalization and translocation.
`Although immunotoxins are simple in concept, the first-genera-
`tion immunotoxins were relatively ineffective. Several requirements
`must be fulfilled for an immunotoxin to be effective (41). In
`particular, (i) the immunoconjugate should be specific and should
`not react with normal tissues. Binding to tissues that do not express
`antigen can be reduced by removal of the nonspecific natural
`cell-binding subunits or domains of the toxin. Furthermore, be-
`cause plant glycoprotein toxins contain mannose oligosaccharides
`that bind to cells of the reticuloendothelial system and, in some
`cases, also contain fucose residues that are recognized by the
`receptors on hepatocytes, deglycosylation of plant toxins may be
`required to avoid rapid clearance and potential cytotoxic effects on
`these cells. (ii) The linkage of the toxin to the antibody should not
`impair the capacity of the antibody to bind antigen. (iii) The
`immunotoxin must be internalized into endosomic vesicles. Thus,
`toxins directed by monoclonal antibodies to surface receptors that
`are normally internalized may be more active than those directed
`
`1660
`
`toward noninternalizing cell surface molecules. (iv) The active
`component of the toxin must translocate into the cytoplasm. These
`various goals can be in conflict; thus, the removal of the B chain of
`ricin reduces nonspecific binding but also reduces the capacity of the
`residual A-chain monoclonal antibody to translocate across the
`endosomic vesicle membrane. (v) For in vivo therapy, the linkage
`must be sufficiently stable to remain intact while the immunotoxin
`passes through the tissues of the patient to its cellular site of action.
`The first generation of heterobifunctional cross-linkers used to bind
`the toxin to the monoclonal antibody generated disulfide bonds that
`were unstable in vivo. This problem was solved in part by the
`synthesis of more stable cross-linkers, which used phenyl or methyl
`groups, or both, adjacent to the disulfide bond to restrict access to
`the bond.
`The development by Pastan and co-workers (41) of IL-2R-
`directed Pseudomonas exotoxin (PE) conjugates for the treatment of
`IL-2R-expressing ATL demonstrates recent progress in the devel-
`opment of effective immunotoxins. PE, chemically coupled to
`anti-Tac, showed specificity in vitro (42). However, only a few
`milligrams of this agent could be given to patients without produc-
`ing liver damage because the toxin had not been sufficiently changed
`so that it would no longer bind to normal liver cells. Functional
`analysis of deletion mutants of the 66-ld PE (43) showed that
`Domain III was responsible for ADP-ribosylation of elongation
`factor 2; Domain II helped in translocation of the toxin to the
`cytosol, whereas Domain I was responsible for unwanted ubiquitous
`cell binding. A PE molecule from which the 26-d) Domain I had
`been deleted (PE40) had full ADP-ribosylating activity but extreme-
`ly low cell-killing activity when used alone. PE40 conjugated to
`anti-Tac inhibited protein synthesis in T cell lines expressing Tac but
`not in lines not expressing the IL-2R. However, immunotoxins
`made by chemical attachment of this truncated PE to anti-Tac
`yielded a product that was heterogeneous. Active single-chain Fv
`fragments of antibodies have been produced in E. coli by linking of
`the light and heavy chain variable domains with a peptide linker
`(44). Chaudhary and co-workers (45) used this genetic engineering
`approach to produce a single-chain antibody toxin fusion protein
`[anti-Tac (Fv)-PE40] in which the variable regions of anti-Tac were
`joined in peptide linkage to PE40 to generate an immunotoxin that
`was cytotoxic to human cell lines bearing IL-2R and to freshly
`obtained ATL cells but not to receptor-negative cells.
`The antitumor activity of immunotoxins has been evaluated in
`animal models since their introduction (46). Treatment usually
`delayed the appearance of tumors and prolonged the lifespan of the
`animals. In a few cases, there was complete regression of the tumor.
`Several in vivo clinical trials in humans have involved the paren-
`teral administration of immunotoxins. In certain cases, for example,
`with monoclonal antibody-toxin conjugates directed toward breast
`or ovarian cancers, severe neurological toxicity was observed because
`of an unanticipated crossreactivity of the monoclonal antibody with
`an antigen in the central nervous system (47). The most common
`toxicity observed was capillary leakage that resulted in hypoalbu-
`minemia, edema, fatigue, and myalgia.
`The results of in vivo clinical trials in patients with cancer with
`first-generation immunotoxins did not fulfill the hopes engendered
`by in vitro and animal model studies. There were only two complete
`remissions and eight partial remissions among the 127 patients
`treated in ten clinical trials with toxin-conjugated monoclonal
`antibodies directed toward ovarian, breast, colorectal, and lymphoid
`neoplastic cells. More encouraging results were obtained when
`benign diseases were treated with immunotoxins or when modified
`immunotoxins were used in therapy of patients with cancer. For
`example, 9 of 22 patients manifested a mixed or partial response,
`and one had a complete remission after treatment of a refractory B
`SCIENCE, VOL. 252
`
`
`
`Downloaded from
`
`http://science.sciencemag.org/
`
`
`
`on January 19, 2018
`
`cell leukemia-lymphoma with a monoclonal antibody to CD19
`conjugated to a modified ricin toxin that had galactose binding sites
`blocked sterically (48). Furthermore, 22 of 32 evaluable patients
`with GVHD had a favorable early response in at least one organ
`after therapy with monoclonal antibody to CD5 conjugated to ricin
`A (49).
`An alternative approach for the delivery of cytotoxic agents to
`cancer cells involves the use of monoclonal antibodies as carriers for
`enzymes to tumor cell surfaces (50). The enzymes are chosen for
`their ability to convert drug precursors injected parenterally into
`active antineoplastic drugs. The active cytotoxic agents formed can
`then penetrate nearby tumor cells and cause the death of these cells.
`A number of prodrugs (drugs in an inactive form that can be
`transformed at the tumor into active anticancer drugs by antibody-
`enzyme conjugates) have been developed. The antibody-enzymes
`conjugates were shown to localize to tumors through the activity of
`the monoclonal antibodies that bind to tumor-associated antigens.
`In vivo studies showed that prodrug administration after monoclo-
`nal antibody-enzyme infusion can result in antitumor activities
`significantly greater than the activities of the prodrugs, drugs, or
`monoclonal antibodies given alone.
`
`Radiolabeled Monoclonal Antibodies
`Toxin conjugates do not pass easily from the endosome to the
`cytosol. Furthermore, the toxins are immunogenic and thus provide
`only a short therapeutic window before the development of anti-
`bodies directed toward the toxin. Radiolabeled monoclonal anti-
`bodies have been developed as alternative immunoconjugates for
`delivery of a cytotoxic effector to target cells and for radioimaging
`(8, 51). Radioimmunodetection with the use of radiolabeled mono-
`clonal antibodies, most often with monoclonal antibodies to carci-
`noembryonic antigen, is widely used to complement other ap-
`proaches for tumor detection. Although intact IgG antibodies are
`retained better by tumors and thus appear to be better for therapy,
`F(ab')2 and Fab fragments are preferred for imaging because both
`targeting and blood clearance are more rapid, which reduces the
`background. Tumors as small as 0.5 cm, which are sometimes
`missed by other radiological methods, can be imaged with antibod-
`ies or antibody fragments labeled with suitable radionuclides.
`One advantage in the use of radiolabeled monoclonal antibody
`con