0022-1767/92/1484-1149S02.00/0
`THE JOURNAL OF IMMUNOLOGY
`Copyright 0 1992 by The American Association of Immunologlsts
`
`Vol. 148. 1149-1 154. No. 4. February 15. 1992
`Printed in U.S.A.
`
`CHIMERIC AND HUMANIZED ANTIBODIES WITH SPECIFICITY FOR THE
`CD33 ANTIGEN'
`
`MAN SUNG C0,2* NEVENKA M. AVDALOVIC,' PHILIP C. CARON,+ MARK V. AVDALOVIC,*
`DAVID A. SCHEINBERG,' AND CARY QUEEN"
`
`From *Protein Design Labs, Inc.. Mountain View, CA 94043 and 'Memorial Sloan-Kettering Cancer Center,
`New York, NY 10021
`
`Received for publication September 3, 1991.
`Accepted for publication November 21. 1991.
`The costs of publication of this article were defrayed in part by the
`payment of page charges. This article must therefore be hereby marked
`advertisement in accordance with 18 U.S.C. Section 1734 solely to indi-
`cate this fact.
`' This work was supported in part by The Lucille P. Markey Charitable
`Trust and by Grant NIH CA55349. D.A.S. is a Lucllle P. Markey Scholar.
`The nucleotide sequence(s) reported in this paper has been submitted
`to the GenBank/EMBL Data Bank with accession numbers M83098 and
`M83099.
`'Address correspondence and reprint requests to Man Sung Co. Protein
`Design Labs, Inc.. 2375 Garcia Ave.. Mountain View, CA 94043.
`'Abbreviations used in this paper: AML. acute myelogenous leukemia:
`HAMA. human anti-mouse antibody: PCR, polymerase chain reaction;
`CDR, complementarity-determining region.
`
`MATERIALS AND METHODS
`Cloning of V region cDNAs. The V domain cDNAs for the H and
`L chains of murine M195 were cloned by an anchored PCR method
`1. First, total RNA was prepared
`(13). which is outlined in Figure
`using the hot phenol method. Briefly. 1 X 10' M 195 hybridoma cells
`were resuspended in 1.2 ml of RNA extraction buffer (50 mM sodium
`acetate, pH 5.2 , 1% SDS]. vortexed, and incubated at room temper-
`ature for 2 min. The cell lysates were then incubated with 0.6 ml of
`phenol, pH 5.2, a t 65°C for 15 min followed by a 15-min incubation
`a microcentrifuge: the aqueous
`on ice. The extract was spun in
`phase was recovered and ethanol precipitated twice. The RNA pellet
`1149
`
`L and H chain cDNAs of M195, a murine mAb that
`munogenicity of murine antibodies, rDNA technology can
`binds to the CD33 Ag on normal and leukemic mye-
`(5). Such
`be applied to construct chimeric antibodies
`loid cells, were cloned. The cDNAs were used in the
`antibodies combine the V region of a mouse antibody
`construction of mouse/human IgGl and IgG3 chi-
`with a human antibody C region, thus retaining the bind-
`meric antibodies. In addition, humanized antibodies
`ing specificity of the murine antibody while presenting
`were constructed which combined the complemen-
`less foreign amino acid sequence to the human immune
`tarity-determining regions of the M195 antibody
`system. In some but not all cases, chimeric antibodies in
`with human framework and constant regions. The
`fact have improved complement-dependent cytotoxicity
`human framework was chosen to maximize homol-
`and antibody-dependent cellular
`cytotoxicity function
`ogy with the M195 V domain sequence. Moreover, a
`relative to mouse antibodies (6) or are less immunogenic
`computer model of M195 was used to identify sev-
`in human patients (7). However, because the entire V
`eral framework amino acids that are likely to inter-
`domain of a chimeric antibody, about one third of the
`act with the complementarity-determining regions,
`molecule, is of mouse origin, chimeric antibodies may
`and these residues were also retained in the human-
`still provoke a substantial HAMA response when used to
`ized antibodies. Unexpectedly, the humanized IgGl
`treat humans.
`and IgG3 M195 antibodies, which have reshaped V
`To reduce further the immunogenicity of murine anti-
`regions, have higher apparent binding affinity for
`the CD33 Ag than the chimeric or mouse antibodies.
`bodies, Winter and
`colleagues (8-10)
`constructed re-
`shaped or "humanized" antibodies by combining only the
`smallest required part of a mouse antibody, the CDR,
`The mAb M195 is a murine IgG2a antibody reactive
`with human V region frameworks and C regions. A dis-
`with the CD33 Ag (1-3). M195 binds to early myeloid
`advantage of this approach is that the CDR may adopt a
`cells, some monocytes, and cells of most myeloid leuke-
`new conformation after being grafted onto the human
`
`mias but not to the earliest hematopoietic stem cells. The
`framework so the humanized antibody often has sub-
`M 195 Ag is also absent from any other hematopoietic or
`stantially reduced affinity for the Ag (101. To overcome
`nonhematopoietic tissue. These properties make the an-
`this problem, we have previously chosen a human frame-
`tibody a n ideal candidate for therapy of AML3 and chronic
`work a s homologous as possible to
`the original mouse
`myelogenous leukemia. The efficient cellular binding and
`framework and used computer modeling to identify sev-
`internalization of M 1 9 5 have allowed use of the radiola-
`eral key residues in the mouse framework which contact
`beled antibody in trials for AML therapy. These trials
`the CDR (1 1-12). These residues must also be transferred
`have demonstrated specific targeting of I3lI-M195 to leu-
`
`to the human framework to maintain the conformational
`kemic cells in
`the bone marrow (4). The murine M195
`integrity of the CDR.
`antibody, however, does not kill leukemic cells by com-
`In this paper we describe the cloning of the murine
`plement-dependent cytotoxicity with human complement
`M195 H and L chain V region cDNAs by a new method.
`or by antibody-dependent cellular cytotoxicity with hu-
`Chimeric antibodies of the human IgG1 and IgG3 isotypes
`man effector cells. The HAMA response may also pre-
`were constructed to improve the effector functions of the
`clude long term use of the murine antibody in patients.
`mouse antibody. Applying our general procedures, we
`To increase the effector function and reduce the im-
`then designed and synthesized humanized M 1 9 5 anti-
`bodies that retain the binding site of the murine antibody
`while having the potential to reduce immunogenicity fur-
`ther.
`
`Downloaded from
`
`http://www.jimmunol.org/
`
` by guest on January 24, 2018
`
`Pfizer v. Genentech
`IPR2017-01488
`Genentech Exhibit 2051
`
`

`

`1150
`
`HUMANIZED ANTIBODIES BINDING TO CD33
`
`Downloaded from
`
`http://www.jimmunol.org/
`
` by guest on January 24, 2018
`
`ends of the M195 VL cDNA clone. The 5’-primer contained a n XbaI
`site followed by the ATG codon and the next 15 nucleotides of VL.
`The 3“primer
`included the last 21 nucleotides of the VL coding
`sequence (noncoding strand) followed by
`the 17 nucleotides that
`follow the J.1 segment in mouse genomic DNA and then an XbaI
`
`site. The fragment generated by PCR with these primers was digested
`with XbaI and cloned into the XbaI site of the pVk vector. The
`correct orientation and sequence of the cloned VL segment in pVk
`were verified by sequencing. The chimeric H chain gene was con-
`structed similarly, using PCR with appropriate primers on the M 195
`VH cDNA and cloning into the XbaI sites of pVgl and pVg3.
`Constructfon of humanized genes. Nucleotide sequences were
`selected which encode the protein sequences of the humanized M195
`H and L chains. including signal peptides, generally utilizing codons
`found in
`the mouse sequence. Several degenerate codons were
`changed to create restriction sites or to remove undesirable ones.
`For each V region gene, two pairs of overlapping oligonucleotides on
`alternating strands were synthesized (Applied Biosystems 3808 DNA
`the entire coding sequences as
`synthesizer), which encompassed
`well as a splice donor signal, and contained suitable restriction sites
`at their ends. The oligonucleotides were 11 0 to 140 bases long with
`15-base overlaps. dsDNA fragments were synthesized with Klenow
`polymerase from
`the 5’-pair and separately
`from the 3’-pair of
`oligonucleotides, digested with restriction enzymes, ligated into the
`pUC18 vector, and sequenced. A 5”fragment and a 3”fragment with
`correct sequences were then excised from pUC18 and ligated to-
`gether into the XbaI sites of the expression vectors pVk. pVgl. or
`pvg3.
`Transfectfon. Transfection was by electroporation using a Gene
`Pulser apparatus (Bio-Rad) at 360 V and 25-microfaraday capaci-
`tance according to the manufacturer’s instructions. Before transfec-
`tion, the L- and H chain-containing plasmids were linearized using
`BamHI. extracted with phenol-chloroform. and ethanol precipitated.
`All transfections were done using 20 pg of each plasmid DNA and
`about 10’ Sp2/0 cells (ATCC CRL 1581) in PBS. The cells from each
`transfection were plated into one 96-well tissue culture plate. After
`48 h selective medium (DMEM + 10% FBS + HT media supplement
`(Sigma) + 1 p g h l mycophenolic acid) was applied. After the wells
`had become confluent with surviving colonies of cells, medium from
`each well was assayed for the presence and quantity of secreted
`antibodies by ELISA. A high yielding clone from each transfection
`was grown up to produce antibody for purification.
`Purfffcatfon of antlbodfes. The chimeric IgGl and IgG3 and hu-
`manized lgGl antibodies were purified from culture supernatant by
`affinity chromatography on protein A-Sepharose CL-4B (Pharmacia).
`The bound antibodies were eluted with 0.1 M glycine HC1 (pH 3.0)
`and neutralized with 1 M Tris-HC1 (pH 8.0). The buffer was ex-
`changed into PBS by passing over a PDlO column (Pharmacia) or by
`dialysis. The humanized IgG3 antibody was purified similarly but
`using protein G-agarose (Pierce Chemical Co.. Rockford. IL) instead
`of protein A-Sepharose. The isotypes of the purified antibodies were
`verified using a n Ouchterlony immunodiffusion kit (The
`Binding
`Site Ltd., Birmingham, UK). The final antibody concentration was
`determined assuming that 1 mgml has an A ~ w of 1.35.
`Affttrttty measurements. The purified antibodies were labeled
`with NaIz5I using chloramine-T, to 2 to 10 pCi/pg of protein. Scat-
`chard analysis was made by binding dilutions of labeled antibody to
`lo5 HL60 cells for 90 min at 0°C. The cells were washed in RPMI
`medium and counted. To avoid nonspecific and FcR binding, the
`assays were done in the presence of human serum. The least squares
`method was used to fit a line to the plot of bound/free vs bound
`antibody, and the apparent K, was read from the slope of the line.
`
`RESULTS
`Cloning of V region cDNAs. The M195 H and L chain
`V domain cDNAs were cloned using an anchored PCR
`method (Fig. l), with 3’-primers that hybridized to the C
`regions and 5”primers that hybridized to the G-tails.
`Immunoglobulin chains have been cloned previously by
`PCR using mixed 5”primers (18, 19). but that method
`may not yield the true sequence of the 5’-part of the
`chain and may not suffice to clone chains with unusual
`5’-sequences. Two independent M 1 9 5 L chain clones
`obtained by our method were sequenced and found to be
`identical: similarly, two H chain clones had the same
`sequence. The nucleotide sequences of the L and H chain
`V regions have been deposited in GenBank: the translated
`protein sequences of the mature chains are shown in
`
`V
`
`V
`
`cDNA 3‘
`
`3’ GGGG
`w
`EcoRl
`
`Reverse Transcription
`
`-
`J G Tailing
`
`C
`
`C
`
`I
`
`I
`
`TTTT 5’
`
`TTTT 5’
`
`Hindlll
`
`I PCR
`
`t
`
`C
`
`
`
`“
`
`V
`
`1
`3’
`5’
`EcoRl
`Hindlll
`Figure 1. Scheme for cloning of light and H chain V region cDNAs.
`Horfzontal arrrows represent oligonucleotide primers.
`
`was resuspended in H 2 0 and quantitated by an OD260 reading. cDNA
`was synthesized from the RNA by incubating 5 pg of total RNA with
`40 ng of dT12-1B (Pharmacia LKB Biotechnology Inc.. Piscataway.
`NJ). 200 U of Moloney murine leukemia virus reverse transcriptase
`(Bethesda Research Laboratories, Bethesda, MD). 40 U of RNasin
`(Promega Biotec. Madison, WI), 50 mM Tris-HC1. pH 8.3, 75 mM KCl.
`10 mM dithiothreitol, 3 mM MgC12. and a 0.5 mM concentration of
`each dNTP in a 2 0 4 reaction volume for 60 min a t 3 7 T . The cDNA
`was purified by phenol extraction and ethanol precipitation. A tail
`of dGs was added to the 3’-terminus of the cDNA by incubating it
`with 15 U of terminal deoxynucleotidyl transferase (Bethesda Re-
`search Laboratories), 0.1 M potassium cacodylate. pH 7.2. 2 mM
`C0Cl2. 0.2 mM dithiothreitol, and 1 mM dGTP in a 2 0 4 reaction
`volume for 30 min at 37°C. Under the conditions described, G-tails
`generally contained about 20 bases. One half of the G-tailed product
`was then amplified to clone the VL gene and the other half amplified
`to clone the VH gene, using Taq polymerase. The VL gene was
`amplified with the primers TATATCTAGAATTCCCCCCCCCCCCCC
`CCC and TATAGAGCTCAAGCTTGGATGGTGGGAAGATGGATAC
`AGTTGGTGC, which, respectively, anneal to the G-tail and the K L
`chain C region. The VH gene was amplified similarly, but using the
`downstream primer TATAGAGCTCAAGCTTCCAGTGGATAGAC(C
`AT)GATGGGG(GC)TGT(TC)GTTTTGGC, which anneals to the C
`indicate base
`region of y chains. The sequences in parentheses
`degeneracies, which were introduced so the primer would be able to
`recognize all y chains isotypes. EcoRl and HfndIIl sites were included
`
`in the upstream and downstream primers for convenient subcloning.
`An alternate set of restriction sites (XbaI and SacI) was also included
`in the primers for the rare event that an EcoRI or Hind111 site is
`present in the V region genes. The PCR reactions were performed in
`a programmable heating block using 30 rounds of temperature cy-
`cling (92°C for 1 min. 50°C for 2 min, and 7 2 T for 3 min). The
`reactions included the G-tailed product, 1 pg of each primer. and 2.5
`U of Taq polymerase (Perkin-Elmer Cetus Instruments. Nonvalk, CT)
`in a final volume of 100 pl. with the reaction buffer recommended
`by the manufacturer. The PCR product bands were excised from a
`low melting agarose gel. digested with restriction enzymes.
`and
`cloned into the pUC18 vector for sequence determination.
`Constructfon of expression vectors. pVk [Fig. 2A) was con-
`structed by replacing the EcoRI-XbaI fragment in the p V ~ l plasmid
`(1 1) with a 600-bp FnuDII fragment containing a human CMV
`enhancer and promoter (14), to which appropriate linkers were
`attached. The same fragment was inserted similarly into the pVyl
`plasmid. A fragment in the modified plasmid containing the hyg
`gene was then replaced with a fragment containing a mutant gene
`for dihydrofolate reductase (15) to construct pVgl (Fig. 28). The
`plasmid pVg3 was constructed similarly to pVgl. with a 3400-bp
`HfndIII-PuuII fragment containing the genomic CYs gene (16) replac-
`ing the C,, gene, using linkers so the XbaI and BamHI sites were
`preserved. All reactions were carried out under standard conditions
`(17).
`Constructfon ofchfrnerfcgenes. Togenerate the chimeric L chain
`gene. primers were synthesized which annealed to the 5‘- and 3‘-
`
`

`

`Downloaded from
`
`http://www.jimmunol.org/
`
` by guest on January 24, 2018
`
`
`
`
`
`HUMANIZED ANTIBODIES BINDING TO CD33
`
`1151
`
`Ftgure 2. Diagram of plasmid vectors. Cod-
`ing regions are shown as boxes, and important
`restriction sites are indicated. Not drawn to
`scale.
`
`EcoRl
`
`A
`
`Amp
`
`I
`
`I
`
`was determined by ELISA. The best producing clones
`Figure 3 (upper lines). The VL domain belongs to the
`secreted 2.5 pg of chimeric IgGl and 6 pg of chimeric
`mouse K chain subgroup 111 and uses the J,1 segment. The
`IgG3/106 cells/24 h. Chimeric IgGl and IgG3 antibodies
`VH domain belongs to the mouse H chain subgroup I1 and
`secreted into the medium were purified on protein A-
`uses the JH4 segment (20).
`Expression vectors. Plasmid vectors were constructed
`Sepharose. Human IgG3 antibodies generally do not bind
`protein A (25). so the ability of the chimeric IgG3 antibody
`for the expression of chimeric and humanized L and H
`to bind to protein A may be mediated through the V region
`chain genes. The plasmid pVk (Fig. 2A) contains the
`(26). The purified IgGl and IgG3 chimeric antibodies were
`human genomic C, segment (21) including about 300 bp
`analyzed by SDS-PAGE (Fig. 4). There is some heteroge-
`of the preceding intron and the poly(A) signal. The intron
`is preceded by a unique XbaI site into which a L chain V
`neity in the murine and chimeric H chains, which may
`be a result of variable glycosylation.
`region can be cloned. Transcription of the complete L
`Design of humanized M195 V domain. To retain the
`chain gene will be initiated by the strong human CMV
`binding affinity and specificity in the humanized anti-
`major immediate early promoter and enhancer (14). The
`bodies, the general procedures of Queen et al. (1 1) were
`pVk plasmid also contains a selectable marker gene for
`followed. First, a human antibody V region with maximal
`xanthine-guanine phosphoribosyltransferase (gpt) (22).
`sequence homology to the mouse M195 V region was
`Plasmids pVgl (Fig. 2B) and pVg3 for expression of y l
`selected to provide the framework sequences for the hu-
`and 73 H chain genes are similar to pVk but, respectively,
`manized antibodies in order to minimize the chance that
`contain the Cyl and Cy3 C region genes (16, 23) and
`the CDR conformation would be distorted when grafted
`another selectable marker. The H chain C region genes
`onto the human framework. The L and H chain V regions
`are genomic clones that include the CHI, hinge, cH2, and
`were taken from the same human antibody to reduce the
`CH3 exons with the intervening introns and about 200
`possibility of incompatibility in assembly of the two
`bp of the intron preceding CH1. The mutant dihydrofolate
`chains. Based on a sequence homology search against
`reductase selectable marker can be used for gene ampli-
`the National Biomedical Research Foundation Protein
`fication (24).
`Synthesis of chimeric antibodies. PCR with appropri-
`Identification Resource, the Eu antibody was selected to
`provide the framework sequences for humanization of
`ate primers was used to precisely copy the VL region,
`M195. The complete Eu L and H chain V regions are,
`including the signal sequence and J segment, from an
`respectively, 54 and 5 1% homologous to the murine M 195
`M195 L chain cDNA clone. The 5’-PCR primer was de-
`L and H chain V regions.
`signed to insert an XbaI site before the ATG codon. The
`Next, the computer programs ABMOD and ENCAD (27)
`3’-primer was designed to insert the splice donor signal
`that normally follows mouse J,1 in genomic DNA after
`were used to construct a molecular model of the M 195 V
`the J segment, followed by an XbaI site. The PCR- gen-
`domain. Inspection of the refined model of murine M 195
`revealed several amino acid residues in the framework
`erated fragment was then cloned into the XbaI site in the
`pVk vector (Fig. 2A). Doing so created a complete chi-
`which have significant contacts with the CDR residues.
`Specifically, residues Phe-40, Ile-52, and Asp-74 in the L
`meric K L chain gene with a miniintron between the
`chain interact with the CDR. Residues Tyr-27, Thr-30,
`mouse V-J and human C, segments. Similarly, PCR with
`Ile-48, Lys-67, Ala-68, Arg-98, and Trp-106 in the H
`suitable primers was used to insert the M 1 9 5 VH region,
`chain also show significant interactions with the CDR.
`including the signal sequence and J segment and fol-
`lowed by a splice donor signal, into the XbaI sites of pVgl
`These murine residues were retained in
`the V region
`framework of the humanized M 195 antibodies.
`(Fig. 2B) and pVg3. The resulting plasmids contain com-
`Different human L and H chain V regions exhibit sub-
`plete chimeric y l and 73 H chain genes, respectively,
`stantial amino acid homology within the framework re-
`with a miniintron between the mouse V-J segment and
`gions (20). However, a given V region usually contains
`the human CHI exon.
`amino acids atypical of other human V regions at several
`The chimeric L chain-containing plasmid was trans-
`framework positions. The Eu antibody contains such
`fected into Sp2/0 mouse myeloma cells together with
`unusual residues at positions L10, L67, and L110 in the
`either the chimeric y l- or chimeric 73-containing plas-
`L chain and positions H93, H95, H107, H108, and H110
`mid, and cells were selected for expression of the gpt
`in the H chain. At these positions, we chose to use a
`gene. Production of antibody by surviving clones of cells
`
`

`

`1152
`
`HUMANIZED ANTIBODI
`:ES BINDING TO CD33
`
`A
`
`1
`
`I
`
`D I V L T Q S P A S L A V S L G Q R A T
`I
`I
`
`
`I
`I
`I
`
` I
`I
`
`
`I
`I
`I
`I
`D I Q M T Q S P S S L S A S V G D R V T
`
`
`
`21 I S C R A S E S V D N Y G I S F M N W F
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`
`I T C R A S E S V D N Y G I S F M N W E
`
`41 Q Q K P G Q P P K L L I Y A A S N Q G S
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`
`I
`I
`I
`I
`I
`
`Q Q K P G K A P K L L I Y A A S N O G S
`
`61 G V P A R F S G S G S G T D F S L N I H
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`
`I
`I
`l
`l
`
`I
`G V P S R F S G S G S G T Q F T L T I S
`
`81 P M E E D D T A M Y F C Q Q S K E V P W
`I
`I
`I
`I
`I
`I
`I
`I
`I
`
`I 1
`I
`I
`S L Q P D D F A T Y Y C Q O S K E V P W
`
`101 T F G G G T K L E I K
`I
`l
`l
`
`I
`l
`l
`
`I
`l
`l
`
`- T F G Q G T K V E I K
`
`B
`
`1
`
`E V Q L Q Q S G P E L V K P G A S V K I
`I
`l
`l
`
`I
`l
`l
`
`I
`l
`l
`
`I
`I
`l
`l
`
`Q V Q L V Q S G A E V K K P G S S V K V
`
`Q, c
`
`c
`
`c
`
`(3 I
`V
`
`M
`(3 I
`V
`
`c
`
`(3 3 I
`
`M
`c3
`3 I
`
`92
`67
`
`45
`
`22
`
`Ffgure4. SDS-PAGE analysls of M195 antlbodies. 1 pg of each antl-
`body was run on the 10% gel. whlch was stained wlth Coomassle blue.
`LeJt lane. m.w. standards of the lndlcated slzes. CHGI, chlmerlc antlbody
`of IgGl isotype; CHC3. chimeric antlbody of IgG3 isotype: HUGI. human-
`ked antibody of IgGl isotype; HUGS. humanlzed antibody of lgG3 Isotype.
`
`Downloaded from
`
`http://www.jimmunol.org/
`
` by guest on January 24, 2018
`
`TABLE I
`Afffnftfes of MI95 antfbodfes
`K. (XlO' "'1
`
`Antlbody Apparent
`Murlne
`2.21 f 0.73"
`Chimerlc I g G l
`2.28
`5.90 * 1.4"
`0.69 It 0.13"
`Chimerlc IgG3
`Humanlzed IgGl
`Hurnanlzed IcG3
`3.33
`" The average of three independent experiments is shown f the SD.
`
`21 S C K A S G Y T F T D Y N M H W V K Q S
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`
`S C K A S G Y T F T D Y N M H W V R Q A
`were surrounded by XbaI sites. The DNA segments were
`inserted into the XbaI sites of the appropriate expression
`41 H G K S L E W I G Y I Y P Y N G G T G Y
`vectors (Fig. 2) to produce plasmids containing complete
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`
`humanized K , y 1, and 73 genes. The humanized L chain-
`P G Q G L E W L G Y I Y P Y N G G T G Y
`containing plasmid was transfected into Sp2/0 cells to-
`gether with either the humanized y l - or humanized y3-
`containing plasmid, and cells were selected for expres-
`sion of the gpt gene. The best antibody-producing clones
`secreted 8 pg of humanized IgGl and 3 pg of humanized
`IgG3/106 cells/24 h. IgG1 antibody secreted into the me-
`dium was purified on protein A-Sepharose. Unlike the
`chimeric IgG3 antibody, the humanized IgG3 antibody
`did not bind protein A. so it was purified on protein G-
`agarose. The humanized IgGl and IgG3 antibodies were
`shown to be pure by SDS-PAGE (Fig. 4). The H chain
`isotypes of the humanized and chimeric antibodies were
`verified by Ouchterlony immunodiffusion.
`Aginity measurements. Flow cytometry was used to
`show that the chimeric and humanized M195 antibodies
`bind to HL60 and U937 cells, which express the CD33
`Ag. but not to several cell lines that do not express CD33
`(data not shown). Binding of the radiolabeled chimeric
`and humanized antibodies to HL60 cells was specific and
`for the CD33 Ag of
`saturable. The apparent affinities
`murine M195 and the various genetic constructs were
`determined by Scatchard analysis (Table I). Murine M 195
`was found to have an apparent K, of 2.2 X lo9 M"
`(Fig.
`5A). consistent with previous results (1). Chimeric IgGl
`M195 has the same apparent affinity as murine M195
`whereas the chimeric IgG3 has a slightly lower apparent
`affinity. Surprisingly, the humanized IgGl M195 anti-
`body has an apparent K, of 5.9 X log M" (Fig. 5B). about
`
`61 N Q K F K S K A T L T V D N S S S T A Y
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`
`I
`l
`l
`
`I
`N O K F K S g & T I T A D E S T N T A Y
`
`81 M D V R S L T S E D S A V Y Y C A R G R
`I
`I
`
`I
`l
`l
`
`I
`I
`I
`I
`I
`I
`I
`I
`I
`
`I
`M E L S S L R S E D T A Y Y X C A R G R
`
`101 P A M D Y W G Q G T S V T V S S
`I
`I
`I
`I
`I
`I
`I
`I
`I
`I
`
`I
`I
`I
`I
`I
`
`P A M D Y E S Q G T L V T V S S
`Ffgure3. Sequences of the humanized M195 llght ( A ) and heavy (B)
`chaln V domalns (lower lfnes). aligned wlth the respectlve murlne M195
`V domalns (upper Ifnes). Vertfcal marks indicate identity of the amino
`aclds. The CDR are underlined. Residues in the Eu framework which
`were replaced by murlne residues or consensus human resldues are
`double underlfned.
`
`consensus human residue rather then the Eu residue in
`the humanized antibody. Several of the unusual residues
`in Eu occur in the H chain J segment, which is signifi-
`cantly mutated from any human genomic JH segment.
`The final amino acid sequences of the humanized M195
`L and H chain V regions are shown in Figure 3. aligned
`with the murine sequences.
`Synthesis of humanized antibodies. DNA segments
`encoding the humanized M195 L and H chain V regions
`were constructed by total gene synthesis from overlap-
`ping oligonucleotides. These genes included signal se-
`quences, J segments, and splice donor sequences and
`
`

`

`1
`
`0.06 1
`
`0.05
`
`0.041
`
`1
`
`0.000
`
`A
`
`/
`
`
`
`. \ .
`
`0.010
`0.005
`0.015
`Bound (nM)
`
`0.020
`
`1153
`HUMANIZED ANTIBODIES BINDING TO CD33
`for the Ag. In contrast, construction of a humanized
`antibody by replacing the mouse V region framework
`with a human framework may alter key contacts between
`the framework and the CDR, which are required to main-
`tain the CDR conformation (30). The altered CDR confor-
`mation may in turn reduce or abolish the binding affinity
`of the antibody for the Ag.
`To retain high affinity when humanizing a mouse an-
`tibody, we have previously introduced two concepts (1 1).
`First, a human framework is selected which maximizes
`homology with the sequence of the mouse antibody V
`region. Such a framework is less likely to introduce dis-
`tortions into the CDR. For the humanized M 195 antibody,
`we chose the framework of the human Eu antibody,
`which we have also used to humanize the anti-Tac and
`anti-gD antibodies (1 1, 12). Eu is used frequently because
`it is the only human antibody having
`an H chain in
`subgroup I for which the sequence of the L chain is also
`available (20). and the H chains of many murine mAb
`are most homologous to human subgroup I. However, we
`have also successfully used the framework of the human
`Pom antibody and of other human antibodies when they
`provide greater homology (12) and (M. S. Co, unpublished
`data). The utility of choosing a homologous human frame-
`work has recently been specifically confirmed by other
`investigators (29).
`In addition to selecting an appropriate human frame-
`work, we use a computer model of the murine antibody
`to identify a small number of amino acids in the mouse
`framework which make key contacts with the
`CDR.
`These residues are retained in the humanized antibody.
`In humanized M195, three L chain and seven H chain
`murine residues were kept for this reason. It may be
`argued that the mouse framework amino acids will make
`the humanized antibody more immunogenic when used
`in human patients. However, when humanized anti-Tac,
`which incorporates a number of murine framework res-
`idues, was tested in cynomolgus monkeys, any monkey
`antibody response was directed to the anti-Tac CDR and
`not to the framework (31). For the fully humanized anti-
`bodies that have been examined (CAMPATH-1 (10). anti-
`respiratory syncytial virus (281, anti-CD4 (291, and anti-
`Tac (W. Schneider and c. Q., unpublished data)) using a
`homologous human framework or retaining some murine
`framework residues or both have been shown to be crit-
`ical for retaining high binding affinity. For the M195
`antibody, these methods actually
`led to a humanized
`antibody with higher apparent binding affinity, as deter-
`mined by Scatchard analysis, than the mouse or chimeric
`antibodies. Elsewhere, we have also directly verified the
`increased apparent binding affinity of humanized M 195
`relative to murine M195 by competition binding of one
`antibody against the other4 (36). The reason for the in-
`crease in affinity of M195 upon humanization is under
`investigation but may relate to a change in glycosylation
`patterns.
`Initial animal and clinical studies with chimeric and
`humanized antibodies have been encouraging. Human-
`ized CAMPATH- 1 has induced remissions in two patients
`with non-Hodgkin lymphoma (32) and one patient with
`Caron. P. C.. M. S. Co. M. K. Bull. N. M. Avdalovic, C. Queen, and D.
`Scheinberg. 1991. Humanized M195 (anti-CD33) monoclonal antibodies:
`potential for therapy of myelogenous leukemia. Subrnlttedfor publlca-
`tion.
`
`Downloaded from
`
`http://www.jimmunol.org/
`
` by guest on January 24, 2018
`
`0
`0 . 0 0 ~ " " ~ ~ " ' ' ~ "
`0.000
`0.005
`0.010
`0.01 5
`Bound (nM)
`
`0.020
`
`Flgure 5. Representative Scatchard analyses of binding of '251-labeled
`murine ( A ) and humanized IgGl (B) M195 antibodies to HL60 cells.
`
`%fold higher than murine M195. The affinity determi-
`nations of the murine M 195 and humanized IgGl anti-
`bodies were performed three times to verify this finding,
`each time with very similar results [Table I]. The human-
`ized IgG3 antibody has an apparent
`affinity slightly
`higher than the murine antibody but not as high as the
`humanized IgGl form. Repeating the experiments sug-
`gested that the chimeric and humanized IgG3 antibodies
`were not completely stable in solution over time.
`
`DISCUSSION
`To date, a multitude of chimeric antibodies and at least
`seven fully humanized antibodies have been described.
`The humanized antibodies, respectively, bind to the pan-
`lymphocyte CAMPATH-1 Ag (101, the p55 chain of the
`IL-2 receptor (1 1). the gB and gD glycoproteins of herpes-
`virus (12). the fusion protein
`of respiratory syncytial
`virus (28). the CD4 Ag (29) and, as described here, the
`CD33 Ag. Because the V and C regions of antibodies are
`independent protein domains (20), it is possible to con-
`struct a chimeric antibody
`by replacing the mouse C
`region with a human C region without altering the con-
`formation of the V region, including the CDR. Hence,
`chimeric antibodies generally retain full binding affinity
`
`

`

`Downloaded from
`
`http://www.jimmunol.org/
`
` by guest on January 24, 2018
`
`1154
`
`
`
`
`
`HUMANIZED ANTIBODIES
`
`BINDING TO CD33
`
`Acknowledgments. We would like to thank H. Selick
`for participating in the construction
`of the expression
`vectors, P. Payne for assistance in the antibody modeling,
`and M. K. Bull for technical assistance.
`
`12. Co. M. S., M. Deschamps. R. J. Whitley, and C. Queen. 1991.
`systemic vasculitis (33), without inducing an antibody
`Humanized antibodies for antiviral therapy. Proc. Natl. Acad. Sct.
`response. The chimeric 17- 1 A antibody has displayed a
`USA 88:2869.
`longer half-life and reduced immunogenicity
`in colon
`13. Loh, E. Y., J. F. Elliott, S. Cwirla, L. L. Lanier. and M. M. Davis.
`1989. Polymerase chain reaction with single-sided specificity: analy-
`cancer patients relative to the original mouse antibody
`sis of T cell receptor d chain. Science 243:217.
`(7). Humanized anti-Tac has shown increased half-life,
`14. Boshart, M.. F. Weber, G. Jahn. K. Dorsch-Hasler. B. Fleckenstein,
`reduced and delayed immunogenicity, and improved ef-
`and W. S