`
`Bw
`
`J. Mol. Biol. (2002) 320, 415–428
`
`Comprehensive Functional Maps of the Antigen-
`binding Site of an Anti-ErbB2 Antibody Obtained with
`Shotgun Scanning Mutagenesis
`
`Felix F. Vajdos1, Camellia W. Adams1, Timothy N. Breece2
`Leonard G. Presta3, Abraham M. de Vos1 and Sachdev S. Sidhu1*
`
`1Department of Protein
`Engineering, Genentech Inc.
`1 DNA Way
`South San Francisco
`CA 94080, USA
`
`2Department of Process
`Sciences, Genentech Inc.
`1 DNA Way
`South San Francisco
`CA 94080, USA
`
`3Department of Immunology
`Genentech Inc., 1 DNA Way
`South San Francisco, CA 94080
`USA
`
`Shotgun scanning combinatorial mutagenesis was used to study the
`antigen-binding site of Fab2C4, a humanized monoclonal antibody frag-
`ment that binds to the extracellular domain of the human oncogene
`product ErbB2. Essentially all the residues in the Fab2C4 complementarity
`determining regions (CDRs) were alanine-scanned using phage-displayed
`libraries that preferentially allowed side-chains to vary as the wild-type or
`alanine. A separate homolog-scan was performed using libraries that
`allowed side-chains to vary only as the wild-type or a similar amino acid
`residue. Following binding selections to isolate functional clones, DNA
`sequencing was used to determine the wild-type/mutant ratios at each
`varied position, and these ratios were used to assess the contributions of
`each side-chain to antigen binding. The alanine-scan revealed that most
`of the side-chains that contribute to antigen binding are located in the
`heavy chain, and the Fab2C4 three-dimensional structure revealed that
`these residues fall into two groups. The first group consists of solvent-
`exposed residues which likely make energetically favorable contacts with
`the antigen and thus comprise the functional-binding epitope. The second
`group consists of buried residues with side-chains that pack against other
`CDR residues and apparently act as scaffolding to maintain the func-
`tional epitope in a binding-competent conformation. The homolog-scan
`involved subtle mutations, and as a result, only a subset of the side-chains
`that were intolerant to alanine substitutions were also intolerant to homo-
`logous substitutions. In particular, the 610 A˚ 2 functional epitope surface
`revealed by alanine-scanning shrunk to only 369 A˚ 2 when mapped with
`homologous substitutions, suggesting that this smaller subset of side-
`chains may be involved in more precise contacts with the antigen. The
`results validate shotgun scanning as a rapid and accurate method for
`determining the functional contributions of
`individual
`side-chains
`involved in protein – protein interactions.
`q 2002 Elsevier Science Ltd. All rights reserved
`
`*Corresponding author
`
`Keywords: phage display; protein engineering; combinatorial mutagenesis;
`antibody; shotgun scanning
`
`Introduction
`
`Monoclonal antibodies have proven invaluable as reagents in biological chemistry, and more recently, as
`therapeutic agents.1 The field of antibody engineering is concerned with technologies that can be used to
`
`Present address: L. G. Presta, DNAX Research Institute of Molecular and Cellular Biology, Inc., 901 California
`Avenue, Palo Alto, CA 94304, USA.
`Abbreviations used: BSA, bovine serum albumin; CC, correlation coefficient; CDR, complementarity determining
`region; CDR-Hn, (where n ¼ 1, 2, or 3), heavy chain CDR 1, 2, or 3; CDR-Ln, (where n ¼ 1, 2, or 3), light chain CDR 1, 2,
`or 3; cP3, C-terminal domain of the M13 bacteriophage gene-3 minor coat protein; ECD, extracellular domain; ELISA,
`enzyme-linked immunosorbant assay; Fab, antigen-binding fragment; Fv, variable fragment; PBS, phosphate-buffered
`saline; rmsd, root mean square deviation; wt, wild-type.
`E-mail address of the corresponding author: sidhu@gene.com
`
`0022-2836/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved
`
`Lassen - Exhibit 1050, p. 1
`
`
`
`416
`
`Shotgun Scanning of Anti-ErbB2 Fab2C4
`
`Table 1. Shotgun scanning codons
`
`Alanine-scana
`
`Homolog-
`scanb
`
`Wild-typec
`
`Codond
`
`m1
`
`m2
`
`m3
`
`Codond
`
`m4
`
`A
`C
`D
`E
`F
`Gp
`H
`Ip
`K
`L
`M
`Np
`Pp
`Qp
`R
`Sp
`T
`V
`W
`Y
`
`GST
`KST
`GMT
`GMA
`KYT
`GST
`SMT
`RYT
`RMA
`SYT
`RYG
`RMC
`SCA
`SMA
`SST
`KCC
`RCT
`GYT
`KSG
`KMT
`
`G
`A
`A
`A
`A
`A
`A
`A
`A
`A
`A
`A
`A
`A
`A
`A
`A
`A
`A
`A
`
`G
`
`S
`
`D
`T
`E
`P
`T
`D
`
`E
`G
`
`G
`D
`
`S
`
`V
`
`P
`V
`T
`V
`V
`T
`
`P
`P
`
`S
`S
`
`KCT
`TSC
`GAM
`GAM
`TWC
`GST
`MAC
`RTT
`ARG
`MTC
`MTG
`RAC
`SCA
`SAA
`ARG
`KCC
`ASC
`RTT
`TKG
`TWC
`
`S
`S
`E
`D
`Y
`A
`N
`V
`R
`I
`L
`D
`A
`E
`K
`A
`S
`I
`L
`F
`
`For each scan, degenerate shotgun codons were designed to
`encode the wild-type amino acid and one or more substitutions.
`Asterisks ( p ) indicate wild-type amino acid residues for which
`both the alanine and homolog-scan codons encode a common
`substitution.
`a The shotgun alanine-scan codon for each amino acid ideally
`encodes only the wild-type or alanine (m1), but the nature of the
`genetic code necessitates the occurrence of two other amino acid
`residues (m2 and m3) for some substitutions. In the case of wild-
`type alanine, the shotgun codon was designed to encode alanine
`and glycine.
`b For the homolog-scan, binomial shotgun codons were
`designed to encode the wild-type and a similar amino acid (m4).
`c Amino acid residues are represented by the single letter
`amino acid code.
`d Equimolar DNA degeneracies in shotgun codons are repre-
`sented by the IUB code (K ¼ G/T, M ¼ A/C, R ¼ A/G, S ¼ G/
`C, W ¼ A/T, Y ¼ C/T).
`
`dissect and rationalize the requirements for anti-
`body structure and function.2 This knowledge can
`then be used to improve or alter particular anti-
`body – antigen interactions, or even to engineer
`completely novel-binding specificities.
`The specificity and affinity of an antibody for its
`cognate antigen is determined by the sequence
`and structure of
`the variable fragment
`(Fv): a
`heterodimer consisting of the N-terminal domains
`of the heavy and light chains. Even within the Fv,
`antigen binding is primarily mediated by the
`complementarity determining regions (CDRs), six
`hypervariable loops (three each in the heavy and
`light chains) which together present a large con-
`tiguous surface for potential antigen binding.
`Aside from the CDRs, the Fv also contains more
`highly conserved framework segments which
`connect the CDRs and are mainly involved in
`conformations,3,4
`supporting
`the CDR loop
`although in some cases, framework residues also
`contact antigen.5,6 As an important step to under-
`standing how a particular antibody functions, it
`would be very useful to assess the contributions
`
`of each CDR side-chain to antigen binding, and in
`so doing, to produce a functional map of the anti-
`gen-binding site.
`Site-directed mutagenesis is a powerful tool for
`mapping binding energetics at protein – protein
`interfaces.7,8
`In this process,
`individual DNA
`codons are systematically altered and the corre-
`sponding mutant proteins are expressed, purified,
`and assayed for activity relative to the wild-type.
`The effects of individual side-chain substitutions
`can then be assessed in terms of DDGmut – wt, the
`difference in binding free energy between the
`mutant and wild-type protein. By analyzing panels
`of point mutants, a detailed map of the binding
`energetics can be obtained, but the process can be
`very laborious because individual mutant proteins
`must be made and analyzed separately. In par-
`ticular, a comprehensive analysis of an antigen-
`binding site would ideally encompass all CDR
`residues, and this would require the analysis of
`dozens or even hundreds of point mutants.9,10
`Recently, a general and rapid combinatorial
`mutagenesis strategy has been developed for
`function.11
`exploring
`protein
`structure
`and
`“Shotgun scanning” mutagenesis uses phage-
`displayed libraries of protein mutants constructed
`using degenerate codons with restricted diversity.
`For example, codons may be chosen to preferen-
`tially allow the wild-type (wt) or alanine in the
`case of a shotgun alanine-scan. The library pool is
`then subjected to binding selections to enrich for
`clones that retain affinity for a binding partner,
`and following selection, DNA sequencing is used
`to determine the ratio of wild-type/mutant (wt/
`mut) at each varied position. This ratio can be
`used to assess binding contributions of each side-
`chain with good correlation to those obtained with
`traditional site-directed mutagenesis. The method
`is very rapid because many side-chains are simul-
`taneously scanned with a single library, and the
`analysis is based on DNA sequencing which
`circumvents the need for protein purification and
`biophysical analysis.
`We used the shotgun scanning approach to
`study the antigen-binding site of a humanized
`monoclonal antibody (humAb2C4) that binds to
`the extracellular domain of the human receptor
`tyrosine kinase ErbB2 (ErbB2-ECD, Kd ¼ 8:5 nM),
`and in so doing, inhibits tumor growth (C.W.A.,
`unpublished results). The antigen-binding portion
`of humAb2C4 was displayed on M13 bacterio-
`phage in an Fab format (Fab2C4), i.e. a heterodimer
`consisting of the light chain and the variable and
`first constant domains of the heavy chain. We
`conducted two different shotgun scans, with each
`scan covering essentially the complete sequences
`of all six CDRs. With a shotgun alanine-scan, we
`assessed the effects of removing all side-chain
`atoms past the b-carbon, fairly drastic mutations
`that can be used to infer the roles of individual
`side-chains in protein structure and function.7
`We also conducted a more subtle scan, termed a
`shotgun homolog-scan, in which we substituted
`
`Lassen - Exhibit 1050, p. 2
`
`
`
`Shotgun Scanning of Anti-ErbB2 Fab2C4
`
`417
`
`Figure 1. Sequences of the Fab2C4 CDRs. The sequence of each CDR is shown along with the position of each resi-
`due in the numbering scheme† of Kabat et al.37 Residues shown to be important for ErbB2-ECD binding in either the
`shotgun alanine or homolog-scan are shown in bold or underlined, respectively (Fwt/mut . 10, see Tables 3 and 4).
`Asterisks ( p ) indicate residues that were not analyzed in the shotgun scans.
`
`Table 2. Fab2C4 shotgun scanning libraries
`
`Mutated regions
`
`Library CDRs
`
`Residues
`
`Shotgun
`codons
`
`Mutagenic
`oligonucleotides
`
`HAa
`
`HAb
`
`LAa
`
`LAb
`
`HHa
`
`HHb
`
`LH
`
`H1,
`H2,
`H3
`H1,
`H2,
`H3
`L1,
`L2,
`L3
`L1,
`L2,
`L3
`H1,
`H3
`H2
`
`L1,
`L2,
`L3
`
`T28, T30, D31, Y32, T33, D50, V51, N52, N53, S54, I58,
`N60, Q61, N95, L96, P98, S99
`
`Alanine
`
`H1-A1, H2-A1,
`H3-A1
`
`D35, P52a, G55, G56, S57, Y59, R62, F63, K64, G65, G97,
`F99a, Y99b, F100, D101, Y102
`
`Alanine
`
`H1-A2, H2-A2,
`H3-A2
`
`Q27, D28, S30, I31, G32, S50, S52, Y53, Y55, Y91, Y92, I93,
`Y94, Y96
`
`Alanine
`
`L1-A1, L2-A1,
`L3-A1
`
`K24, A25, S26, V29, V33, A34, A51, R54, T56, Q89, Q90,
`P95, T97
`
`Alanine
`
`L1-A2, L2-A2,
`L3-A2
`
`T28, T30, D31, Y32, T33, M34, D35, N95, L96, G97, P98,
`S99, F99a, Y99b, F100, D101, Y102
`D50, V51, N52, P52a, N53, S54, G55, G56, S57, I58, Y59,
`N60, Q61, R62, F63, K64, G65
`K24, A25, S26, Q27, D28, V29, S30, I31, G32, V33, A34,
`S50, A51, S52, Y53, R54, Y55, T56, Q89, Q90, Y91, Y92, I93,
`Y94, P95, Y96, T97
`
`Homolog
`
`H1-H, H3-H
`
`Homolog
`
`H2-H
`
`Homolog
`
`L1-H, L2-H,
`L3-H
`
`Diversity
`
`Theoretical
`3.3 £ 107
`
`Actual
`1.5 £ 1010
`
`1.7 £ 107
`
`2.4 £ 1010
`
`8.3 £ 107
`
`1.4 £ 1010
`
`1.6 £ 104
`
`2.5 £ 1010
`
`1.3 £ 105
`1.3 £ 105
`1.3 £ 108
`
`2.4 £ 1010
`2.2 £ 1010
`2.4 £ 1010
`
`Libraries were designed to replace the codons for the indicated residues with either alanine-scan or homolog-scan shotgun codons
`(Table 1). Libraries were constructed using the indicated mutagenic oligonucleotides (see Materials and Methods), and in each case,
`the theoretical diversity (the number of amino acid combinations encoded by the mutagenic oligonucleotides) was exceeded at least
`100-fold by the actual diversity of the constructed library.
`
`each wild-type residue with a similar amino acid,
`to gain insight into which positions require precise
`side-chain geometries and chemistry. When the
`mutagenesis results were mapped onto the three-
`dimensional crystal structure of Fab2C4, each scan
`provided a comprehensive view of how the CDR
`side-chains contribute to the formation of a func-
`
`tional antigen-binding site. The two views are
`distinct yet complementary: together, they provide
`a clearer understanding of antibody structure and
`function than would be possible with either scan
`alone.
`
`Results
`
`† Antibody residues are designated by a letter in lower
`case italics denoting the heavy or light chain (h or l,
`respectively), followed by the amino acid in the one-
`letter code, followed by the position in the chain. For
`example, h D101 denotes an aspartic acid residue at
`position 101 in the heavy chain.
`
`Shotgun alanine-scan of Fab2C4
`
`For the shotgun alanine-scan, we replaced wt
`codons with degenerate
`codons
`that
`ideally
`encoded the wt amino acid or alanine (m1 in
`Table 1), although the nature of the genetic code
`
`Lassen - Exhibit 1050, p. 3
`
`
`
`418
`
`Shotgun Scanning of Anti-ErbB2 Fab2C4
`
`Table 3. Fab2C4 light chain shotgun scan
`
`Wt/mut ratios
`
`Antigen selection
`
`Display selection
`
`Residuea Wt/m1 Wt/m2 Wt/m3 Wt/m4 Wt/m1 Wt/m2 Wt/m3 Wt/m4
`
`K24
`A25
`S26p
`Q27p
`D28
`V29
`S30p
`I31p
`G32p
`V33
`A34
`
`S50p
`A51
`S52p
`Y53
`R54
`Y55
`T56
`
`Q89p
`Q90p
`Y91
`Y92
`I93p
`Y94
`P95p
`Y96
`T97
`
`0.89
`3.7
`3.5
`0.67
`1.1
`6.1
`1.8
`0.91
`3.3
`16
`16
`
`1.0
`1.7
`1.3
`1.9
`3.2
`32
`0.49
`
`8.8
`2.4
`. 166
`1.2
`1.7
`6.7
`13
`0.99
`0.56
`
`4.2
`
`1.5
`
`0.96
`
`2.5
`
`2.8
`
`0.57
`
`97
`4.1
`80
`
`10
`1.1
`. 166
`3.7
`1.6
`30
`
`. 66
`
`4.4
`1.9
`53
`
`70
`. 36
`166
`1.1
`0.81
`5.5
`
`2.1
`
`0.88
`2.8
`2.8
`0.51
`1.8
`3.5
`1.1
`0.64
`4.8
`3.1
`5.5
`
`0.78
`1.6
`1.2
`1.4
`3.0
`4.8
`0.88
`
`3.6
`0.67
`0.94
`0.88
`0.69
`1.3
`9.7
`0.36
`0.28
`
`0.42
`2.0
`2.9
`0.88
`0.99
`2.5
`1.5
`1.7
`2.9
`3.3
`3.6
`
`1.3
`0.90
`1.5
`1.6
`1.7
`1.4
`0.89
`
`0.77
`0.88
`1.8
`1.3
`1.7
`1.9
`1.1
`2.1
`0.89
`
`0.79
`
`0.52
`
`1.2
`
`0.94
`
`2.7
`
`0.56
`
`3.5
`3.7
`2.3
`
`2.4
`1.9
`3.5
`2.1
`1.5
`3.0
`
`18
`
`1.2
`1.0
`0.89
`
`3.4
`2.3
`0.97
`0.84
`0.64
`1.7
`
`2.2
`
`1.0
`1.6
`1.5
`0.73
`1.9
`2.0
`0.87
`0.55
`3.9
`2.8
`2.5
`
`0.87
`0.85
`1.7
`1.3
`2.4
`0.95
`0.76
`
`1.9
`0.71
`1.2
`0.6
`0.53
`0.63
`1.74
`0.91
`0.35
`
`Fwt/mut
`
`m2
`
`5.3
`
`m3
`
`1.8
`
`1.3p
`
`2.7
`
`3.7
`
`1.0p
`
`28
`1.1
`35
`
`4.2p
`0.58p
`.47
`1.8
`1.1
`10
`
`3.7
`1.9
`60
`
`21
`. 16
`138
`0.76
`1.3p
`3.2
`
`. 3.7
`
`0.95
`
`m1
`
`2.1
`1.8
`1.2p
`0.76
`1.1
`2.4
`1.1p
`0.53
`1.1p
`4.8
`4.6
`
`0.77p
`1.9
`0.85p
`1.2
`1.8
`23
`0.55
`
`11
`2.7
`. 92
`0.96
`1.0
`3.6
`12 p
`0.48
`0.62
`
`m4
`
`0.86
`1.8
`1.9p
`0.70p
`1.0
`1.8
`1.3p
`1.2p
`1.2p
`1.1
`2.2
`
`0.89p
`1.8
`0.70p
`1.1
`1.3
`5.1
`1.2
`
`1.8p
`0.94p
`0.76p
`1.5
`1.3p
`2.0
`5.6p
`0.40
`0.80
`
`For each of the listed light chain residues, the effect of each mutation (Table 1) was assessed using data from either the alanine-scan
`libraries (m1, m2, and m3) or the homolog-scan libraries (m4) described in Table 2. The wt/mut ratios were determined from the
`sequences of binding clones isolated after selection for binding to either the ErbB2-ECD (antigen selection) or an anti-tag antibody
`(display selection). The function ratio (Fwt/mut) for each mutation was derived by dividing the antigen selection wt/mut ratio by the
`display selection wt/mut ratio. Fwt/mut provides a quantitative estimate of the effect of each mutation on the binding affinity of
`Fab2C4 for ErbB2-ECD. Deleterious effects are indicated by Fwt/mut values greater than 1.0, and mutations that have large deleterious
`effects (Fwt/mut . 10) are shown in bold text. In cases where a particular mutation was not observed amongst the antigen selection
`sequences, only a lower limit could be defined for the wt/mut ratio and the Fwt/mut (indicated by a greater than sign). Asterisks ( p )
`indicate residues for which the alanine and homolog-scan codons encoded a common substitution.
`a Residues are denoted by the single letter amino acid code and are numbered according to the scheme of Kabat et al.37
`
`necessitated two other amino acid substitutions for
`some residues (m2 and m3 in Table 1). In positions
`where alanine was the wt, we used a degenerate
`codon that encoded alanine or glycine. The six
`CDRs of Fab2C4 encompass a total of 64 residues
`(Figure 1). We constructed two libraries (HAa and
`HAb) that together covered 33 of the 37 heavy
`chain CDR residues and two libraries (LAa and
`LAb) that together covered all 27 light chain CDR
`residues (Figure 1 and Table 2). Each library con-
`tained . 1010 unique members, and thus in each
`case,
`the theoretical diversity for combinatorial
`mutagenesis
`at
`the
`scanned positions was
`exceeded by at least 100-fold (Table 2).
`Phage pools from each library were subjected to
`two different selections. The first selection (display
`selection) isolated variants capable of binding to a
`monoclonal antibody specific for the epitope tag
`fused to the N terminus of the Fab2C4 light chain.
`The second selection (antigen selection) isolated
`variants capable of binding to ErbB2-ECD. Close
`to 100 binding clones were sequenced from each
`
`selection; the sequences were aligned, and at each
`mutated position,
`the occurrences of wt or
`each designed substitution were tabulated (see
`Materials and Methods for details). For each selec-
`tion, these data were used to calculate the wt/mut
`ratio for each mutation at each position (Tables 3
`and 4).
`Because the wt/mut ratio is the statistical prefer-
`ence for the wt relative to the mutant, it correlates
`with the effect of each mutation on the selected
`trait
`(i.e. binding to the anti-tag antibody or
`ErbB2-ECD). Ratios greater than or less than 1
`indicate deleterious
`or beneficial mutations,
`respectively.
`The anti-tag antibody selected for phage variants
`that displayed assembled Fab2C4 fragments con-
`taining both the heavy and light chains. This is
`because the heavy chain was fused directly to a
`bacteriophage coat protein while the epitope tag
`was fused to the light chain N terminus. Thus, the
`anti-tag antibody only binds to phage particles
`that contain a light chain associated with the
`
`Lassen - Exhibit 1050, p. 4
`
`
`
`Shotgun Scanning of Anti-ErbB2 Fab2C4
`
`Table 4. Fab2C4 heavy chain shotgun scan
`
`Wt/mut ratios
`
`Antigen selection
`
`Display selection
`
`Fwt/mut
`
`Residuea Wt/m1 Wt/m2 Wt/m3 Wt/m4 Wt/m1 Wt/m2 Wt/m3 Wt/m4
`
`m1
`
`m2
`
`m3
`
`T28
`T30
`D31
`Y32
`T33
`M34
`D35
`
`D50
`V51
`N52p
`P52ap
`N53p
`S54p
`G55p
`G56p
`S57p
`I58p
`Y59
`N60p
`Q61p
`R62
`F63
`K64
`G65p
`
`N95p
`L96
`G97p
`P98p
`S99p
`F99a
`Y99b
`F100
`D101
`Y102
`
`4.5
`0.33
`170
`. 161
`20
`NDb
`2.8
`
`170
`10
`. 168
`72
`. 166
`84
`14
`0.60
`7.0
`45
`33
`4.8
`2.6
`4.3
`26
`54
`5.8
`
`. 170
`23
`. 78
`. 178
`2.8
`. 75
`. 74
`77
`9.1
`8.3
`
`. 161
`
`. 161
`
`ND
`
`ND
`
`168
`
`166
`
`84
`
`. 166
`
`45
`. 59
`4.4
`0.98
`. 44
`26
`54
`
`21
`. 45
`
`. 75
`74
`. 77
`
`7.5
`
`4.5
`9.8
`120
`1.1
`4.0
`4.6
`6.0
`
`. 170
`0.35
`
`. 75
`74
`77
`
`3.2
`
`0.94
`0.27
`29
`17
`8.9
`2.2
`14
`
`. 91
`1.3
`. 91
`14
`. 91
`. 91
`90
`0.36
`0.47
`2.1
`0.78
`3.0
`0.69
`1.3
`3.2
`0.57
`9.1
`
`21
`1.5
`89
`29
`7.0
`10
`1.7
`17
`. 87
`2.8
`
`0.7
`0.7
`1.4
`2.0
`0.94
`ND
`0.14
`
`0.24
`1.1
`0.41
`6.1
`1.4
`0.33
`0.40
`5.0
`4.4
`0.86
`8.7
`1.2
`0.53
`1.2
`6.6
`4.9
`2.50
`
`1.8
`0.11
`3.3
`1.9
`0.55
`2.4
`0.8
`2.6
`1.1
`2.3
`
`3.1
`
`1.1
`
`ND
`
`ND
`
`0.34
`
`0.97
`
`0.80
`
`2.6
`
`0.95
`10.4
`0.91
`0.42
`15
`2.2
`7.7
`
`2.0
`0.33
`
`5.4
`4.1
`5.9
`
`1.9
`
`0.51
`1.8
`15
`2.0
`0.24
`8.8
`2.7
`
`2.1
`0.19
`
`1.3
`1.7
`1.5
`
`2.1
`
`0.47
`0.39
`1.1
`0.85
`0.38
`0.88
`0.90
`
`0.41
`1.8
`0.83
`0.62
`0.57
`1.1
`2.9
`2.6
`0.86
`0.61
`0.58
`1.8
`0.71
`1.2
`4.0
`0.67
`3.9
`
`3.1
`1.2
`2.1
`0.44
`1.6
`1.1
`0.49
`5.1
`2.5
`0.92
`
`6.4
`0.47
`120
`. 81
`21
`ND
`20
`
`710
`9.4
`. 410
`12 p
`. 120
`260 p
`34 p
`0.12p
`1.6p
`53
`3.8
`4.0
`4.8
`3.6
`4.4
`12
`2.3p
`
`. 98
`210
`. 24 p
`. 94 p
`5.0p
`. 31
`. 93
`30
`8.3
`3.6
`
`. 52
`
`.150
`
`ND
`
`ND
`
`490 p
`
`170 p
`
`110
`
`.64
`
`47
`. 5.7
`4.8p
`2.3p
`2.9
`12
`7.0
`
`11p
`. 140
`
`14
`18
`13
`
`3.9
`
`8.8p
`5.4
`8.0
`0.55
`17
`0.52
`2.2
`
`84
`1.8
`
`58
`44
`51
`
`1.5
`
`419
`
`m4
`
`2.0
`0.69
`26
`20
`23
`2.5
`15
`
`. 220
`0.73
`.110 p
`23 p
`. 160 p
`. 83 p
`31 p
`0.14p
`0.55p
`3.4p
`1.3
`1.7p
`0.97p
`1.0
`0.81
`0.85
`2.4p
`
`6.9p
`1.3
`42 p
`65 p
`4.4p
`9.1
`3.5
`3.3
`. 35
`3.0
`
`For each of the listed heavy chain residues, the effect of each mutation (Table 1) was assessed using data from either the alanine-
`scan libraries (m1, m2, and m3) or the homolog-scan libraries (m4) described in Table 2. The wt/mut ratios were determined from
`the sequences of binding clones isolated after selection for binding to either the ErbB2-ECD (antigen selection) or an anti-tag antibody
`(display selection). The function ratio (Fwt/mut) for each mutation was derived by dividing the antigen selection wt/mut ratio by the
`display selection wt/mut ratio. Fwt/mut provides a quantitative estimate of the effect of each mutation on the binding affinity of
`Fab2C4 for ErbB2-ECD. Deleterious effects are indicated by Fwt/mut values greater than 1.0, and mutations that have large deleterious
`effects (Fwt/mut . 10) are shown in bold text. In cases where a particular mutation was not observed amongst the antigen selection
`sequences, only a lower limit could be defined for the wt/mut ratio and the Fwt/mut (indicated by a greater than sign). Asterisks ( p )
`indicate residues for which the alanine and homolog-scan codons encoded a common substitution.
`a Residues are denoted by the single letter amino acid code and are numbered according to the scheme of Kabat et al.37
`b ND indicates that these values were not determined, because we forgot to include this residue in the alanine-scan libraries.
`
`phage-displayed heavy chain. Most of the wt/mut
`ratios for the display selection were close to 1.0,
`indicating that the mutations did not significantly
`affect Fab2C4 display levels (Tables 3 and 4). How-
`ever, several mutations exhibited wt/mut ratios
`significantly greater than 1.0 (e.g. h P52aA, h Y59A,
`h F63A), suggesting that these mutations reduced
`display. Conversely, for a few mutations, wt/mut
`ratios significantly less than 1.0 suggest that these
`mutations may actually increase display (e.g.
`h D35A, h L96A).
`In the selection for binding to ErbB2-ECD,
`mutations could effect
`the selection either by
`altering the level of Fab2C4 display (as in the
`display selection), or alternatively, by directly or
`indirectly altering the side-chains
`that make
`
`binding contacts with the antigen. In this selection,
`alanine substitutions at three light chain positions
`(Table 3) and 21 heavy chain positions (Table 4)
`exhibited wt/mut ratios greater than 10.
`To obtain a quantitative estimate of each muta-
`tion’s effect on ErbB2-ECD binding affinity, we
`divided the wt/mut ratio from the antigen selec-
`tion by the wt/mut ratio from the display selec-
`tion. This operation corrected for effects on
`Fab2C4 display and provided a number which we
`termed the function ratio (Fwt/mut). As we have
`the Fwt/mut value for each
`shown previously,
`mutation is approximately equal
`to the corre-
`sponding ratio of equilibrium binding constants
`(Ka,wt/Ka,mut),11 and thus,
`it provides a good
`estimate of the effect of each mutation on the
`
`Lassen - Exhibit 1050, p. 5
`
`
`
`420
`
`Shotgun Scanning of Anti-ErbB2 Fab2C4
`
`Figure 2. Fwt/mut values measuring the effects of Fab2C4 CDR mutations on the binding affinity for ErbB2-ECD.
`Values are shown for either alanine (black bars) or homolog (white bars) substitutions. Data for (a) the light chain
`were from Table 3, and data for (b) the heavy chain were from Table 4 (except the mutation hM34A for which the
`EC50,mut/EC50,wt-value from Table 5 was plotted).
`
`equilibrium binding constant between Fab2C4 and
`ErbB2-ECD. Alanine substitutions at three light
`chain positions and 19 heavy chain positions
`exhibited Fwt/Ala values greater than 10, indicating
`that
`side-chains at
`these positions contribute
`significantly to the binding affinity of Fab2C4 for
`ErbB2-ECD (Tables 3 and 4, Figure 2).
`
`Shotgun homolog-scan of Fab2C4
`
`In the shotgun homolog-scan libraries, each
`scanned position was represented by a binomial
`codon that encoded only the wild-type and a
`similar amino acid (Table 1). We constructed two
`
`libraries (HHa and HHb) that together covered 34
`heavy chain CDR residues and a single library
`(LH) that covered all 27 light chain CDR residues
`(Figure 1 and Table 2). As with the alanine-scans,
`the library diversities were sufficient to exceed the
`theoretical diversities by at least 100-fold (Table 2).
`Each library was subjected to separate selections
`for binding to anti-tag antibody or ErbB2-ECD
`and Fwt/mut values were determined for each
`mutation, as described above for shotgun alanine-
`scanning. The Fwt/mut values for many homolog
`substitutions were significantly lower than those
`for the corresponding alanine substitutions; no
`light chain residues and only 13 heavy chain
`
`Lassen - Exhibit 1050, p. 6
`
`
`
`Shotgun Scanning of Anti-ErbB2 Fab2C4
`
`421
`
`leucine, the alanine-scan used tetranomial codons
`that encoded the homolog-scan substitution in
`addition to alanine and wt. Thus, for 26 mutations
`that overlapped in the two scans, we could com-
`pare the Fwt/mut values determined from the
`alanine-scan to those determined from the homo-
`log-scan (asterisks in Tables 3 and 4). For these
`identical mutations, a least squares linear fit of the
`logarithms of the Fwt/mut values from the alanine-
`scan versus the logarithms of the Fwt/mut values
`from the homolog-scan showed a strong corre-
`lation ðr ¼ 0:96Þ; with a slope close to 1.0 and a
`y-intercept close to zero (Figure 3). Thus, it appears
`that identical point mutations in different combi-
`natorial libraries have very similar effects on the
`binding affinity of Fab2C4 for ErbB2-ECD.
`
`Binding activity measurements with Fab2C4
`point mutants
`
`Site-directed mutagenesis was used to construct
`genes encoding Fab2C4 point mutants; the mutants
`were expressed in Escherichia coli and the recombi-
`nant proteins were purified. An enzyme-linked
`immunosorbant assay (ELISA) with immobilized
`ErbB2-ECD was used to measure the binding
`activity of wt Fab2C4 and each mutant protein.
`For each protein, the EC50 was determined as the
`Fab concentration corresponding to the half-
`maximal binding signal. By dividing the EC50 for
`each Fab2C4 point mutant by the EC50 for wt
`Fab2C4, we obtained a measure of
`the fold
`reduction in ErbB2-ECD binding activity due to
`each point mutation (Table 5), and these data were
`in good agreement with the shotgun scanning
`results (Table 4 and Figure 2(b)). Both methods
`indicated that mutations at positions hN52, hN53,
`and hS54 greatly reduced binding affinity for
`ErbB2-ECD, while the mutation hM34L caused
`only a modest 2-fold reduction.
`
`Three-dimensional structure of Fab2C4
`
`The X-ray crystal structure of Fab2C4 was deter-
`mined by the molecular replacement method,
`using as a search model
`the coordinates of
`humanized Fab4D5 version 4 (Fab4D5v4),12 an
`antibody fragment that also binds to ErbB2-ECD
`but recognizes an epitope distinct from that recog-
`nized by Fab2C4.13 The structure was refined at
`1.8 A˚ resolution to Rwork and Rfree values of 19.7%
`and 23.0%, respectively. The details of the structure
`determination and refinement are described in
`Materials and Methods; data collection and refine-
`ment statistics are shown in Table 6. Fab2C4 and
`Fab4D5v4 share 91% sequence identity; most of
`the differences reside in the CDRs, as the frame-
`work regions differ at only seven positions. Thus,
`it is not surprising that the Ca atoms of the two
`structures superimpose with a root mean square
`deviation (rmsd) of 1.5 A˚ , excluding the CDRs.
`Some of
`this difference can be attributed to
`“hinge”-motion between the variable and constant
`
`Figure 3. Correlation between Fwt/mut values deter-
`mined using data from the shotgun alanine (x-axis) or
`homolog-scan (y-axis). The alanine and homolog-scan
`data could be used to determine the Fwt/mut values for
`26 identical point mutations that overlapped in the two
`scans (asterisks in Tables 3 and 4). The logarithms of the
`Fwt/mut values are plotted and the least squares linear fit
`of the data is shown, with the corresponding equation
`and Pearson’s coefficient (r ) given at the top.
`
`residues exhibited Fwt/mut values greater than 10
`(Tables 3 and 4, Figure 2).
`
`Comparison of Fwt/mut values for identical
`mutations in different libraries
`
`While most substitutions in the homolog-scan
`were designed to be different from those in the
`alanine-scan,
`there was some overlap (residues
`with asterisks in Table 1). Glycine, proline and
`serine were substituted with alanine in both scans.
`Furthermore, for asparagine, glutamine, and iso-
`
`Table 5. Relative binding activities for Fab2C4 point
`mutants
`
`Mutant
`
`hM34A
`hM34L
`hN52Aa
`hN52Qa
`hN53Aa
`hN53Da
`hN53Qa
`hS54Aa
`
`EC50,mut/EC50,wt
`
`5.3
`1.8
`. 103
`. 103
`. 103
`. 103
`. 103
`. 103
`
`The binding activities of mutant proteins were assessed as
`EC50 values, and the ratio of EC50,mut/EC50,wt was determined as
`a measure of the fold reduction in ErbB2-ECD binding activity
`due to each point mutation (see Materials and Methods).
`a For extremely deleterious mutations, EC50 values could not
`be determined because binding could not be saturated. Thus,
`only a lower limit (. 103) for fold reduction in ErbB2-ECD
`binding could be estimated for these mutations.
`
`Lassen - Exhibit 1050, p. 7
`
`
`
`422
`
`Shotgun Scanning of Anti-ErbB2 Fab2C4
`
`Table 6. Data collection and refinement statistics for
`Fab2C4
`
`A. Unit cell
`Space group
`a (A˚ )
`b (A˚ )
`c (A˚ )
`b (deg.)
`
`B. Diffraction data
`Resolution (A˚ )
`No. of reflections
`No. of unique reflections
`b
`Rmerge
`Completeness (%)
`I/s(I )
`Redundancy
`
`C. Refinement
`c
`Rwork
`c
`Rfree
`No. of protein atoms
`No. of water molecules
`No. of sulfate ions
`Average Bprotein (A˚ 2)
`Average Bwater molecules (A˚ 2)
`Average Bsulfate (A˚ 2)
`Rmsd bond length (A˚ )
`Rmsd angles (deg.)
`Rmsd bonded Bs (A˚ 2)
`
`P21
`41.97
`64.25
`79.44
`105.44
`
`15– 1.8 (1.9– 1.8)a
`85,734
`36,884
`0.065 (0.328)a
`97.6 (97.6)a
`5.2 (1.4)a
`2.3 (2.3)a
`
`0.197
`0.230
`3323
`382
`2
`22.6
`33.8
`61.1
`0.005
`1.4
`1.8
`
`a Values
`for
`the outer
`resolution shell are given in
`parantheses.
`b Rmerge ¼ ShklðlIhkl 2 kIhklllÞ=ShklkIhkll; where Ihkl is the intensity
`of reflection hkl, and kIhkll is the average intensity of multiple
`observations.
`c Rwork ¼ SlFo 2 Fcl=SFo; where Fo and Fc are the observed
`and calculated structure factor amplitudes, respectively. Rfree is
`the R-factor for a randomly selected 5% of reflections which
`were not used in the refinement.
`
`domains, as the rmsd between the two structures
`decreases to 0.7 A˚ or 0.9 A˚ when the superposition
`is performed using only the constant domains or
`the variable domain frameworks, respectively. All
`residues in the CDRs of Fab2C4 are well ordered,
`with the exception of h L100. The disorder in this
`region appears to be correlated with disorder
`around the immediately adjacent CDR-H1 residue
`h Y32. In comparison with Fab4D5v4, there is a
`single amino acid deletion in the sequence of
`Fab2C4 that occurs in CDR-H3 and is accommo-
`dated by a completely altered backbone trajectory.
`Conservation of a hydrophobic patch in this region
`is maintained by the aromatic ring of h F105 in
`Fab2C4 lying in the same position as h W99 in
`Fab4D5v4, despite the fact that these residues are
`at opposite ends of CDR-H3. The only difference
`in the light chain frameworks occurs at position
`66, where an arginine in Fab4D5v4 is substituted
`by a glycine in Fab2C4, causing the polypeptide
`backbone to undergo a significant rearrangement.
`
`Discussion
`
`Antibody affinity and specificity is predomi-
`nantly dictated by the six CDR loops that together
`
`Figure 4. Mapping of the functional epitopes for bind-
`ing of ErbB2-ECD onto the structure of Fab2C4. The
`functional epitopes defined by (a) shotgun alanine-
`scanning or (b) shotgun homolog-scanning are shown.
`Unscanned light or heavy chain residues are colored
`cyan or blue, respectively. Scanned residues are color-
`coded according to the magnitudes of Fwt/mut values, as
`follows: red, . 30; yellow, 10 – 30; grey, , 10. Labeled resi-
`dues with asterisks ( p ) indicate light chain residues. The
`solvent-exposed Cg2 group of hT33 is located directly
`over a hydrogen bond network involving hT33 and
`several buried side-chains (Figure 5). Data are also
`shown graphically in Figure 2 and were obtained from
`Tables 3 and 4. The Fab2C4 structure is shown in CPK
`representation. This Figure and Figure 5 were generated
`using PyMOL (http://pymol.sourceforge.net).
`
`form the antigen-binding site. With the exception
`of CDR-H3, the conformations of the CDR main-
`chains do not vary greatly within different anti-
`bodies, and they can be classified into a limited
`number of “canonical structures”.14 Thus,
`the
`major determinants of antibody specificity and affi-
`nity are the CDR side-chains. Side-chains can be
`classified as buried or solvent exposed on the
`basis of their solvent accessible surface area, and
`these classifications have implications for
`the
`
`Lassen - Exhibit 1050, p. 8
`
`
`
`Shotgun Scanning of Anti-ErbB2 Fab2C4
`
`423
`
`Table 7. Buried Fab2C4 CDR residues and contacts
`
`Residue
`
`lA25
`lV29
`lG32
`lV33
`lA34
`lA51
`lQ89
`lQ90
`l Y91
`hM34
`h D35
`h D50
`hV51
`h P52a
`h F63
`hN95
`h F100
`
`Contacts
`
`l K24, lS26, lQ27, lV29
`lA25, lQ27, l D28, lS30, lV33
`lV29, l Y91, l Y92
`lV29, lG32, lA34, lQ89, l Y91
`lV33, lQ89, l Y91
`l I31, lV33, lS50, lS52
`lA34, lQ90, l Y96, lT97, h H99a, h Y99b, h F100
`lG32, lQ89, l Y91, l Y92, l I93, l Y96, lT97
`l I31, lG32, lV33, lA34, lS50, lQ90, l Y92, l Y96, hS99, h F99a, h Y99b
`hT33, h D35, hV51
`hM34, h D50, hN95, h F99a, h F100
`hT33, hM34, h D35, hV51, h I58, hN95, h F99a
`hM34, h D50, hN52, hG55, hG56
`hT30, h Y32, hT33, hV51, hN53
`hN60, h R62, h K64
`hT33, h D35, h D50, h L96, hG97, hS99, h F9