`
`Three-Dimensional Structure of a Fluorescein-Fab
`Complex Crystallized in 2-Methyl-2,4-pentanediol
`James N. Herron,1 Xiao-min He.' Martha L. Mason, 1 Edward W. Voss, Jr.,2 and Allen B. Edmundson 1
`1Departrrumt of Biology, University of Utah, Salt Lake City, Utah 84112 and 2Department of Microbiology,
`University of Illinois, Urbana-Champaign, Illirwis 61081
`
`ABSTRACT
`The crystal structure of a flu-
`orescein-Fab (4-4-20) complex was determined
`at 2.7 A resolution by molecular replacement
`methods. The starting model was the refined 2.7
`A structure of unliganded Fab from an au(cid:173)
`toantibody (BV04-01) with specificity for single(cid:173)
`stranded DNA. In the 4-4-20 complex fluores(cid:173)
`cein fits tightly into a relatively deep slot
`formed by a network of tryptophan and ty(cid:173)
`rosine side chains. The planar xanthonyl ring of
`the hapten is accommodated at the bottom of
`the slot while the phenylcarboxyl group inter(cid:173)
`faces with solvent. Tyrosine 37 (light chain) and
`tryptophan 33 (heavy chain) flank the xantho(cid:173)
`nyl group and tryptophan 101 (light chain) pro(cid:173)
`vides the floor of the combining site. Tyrosine
`103 (heavy chain) is situated near the phenyl
`ring of the hapten and tyrosine 102 (heavy
`chain) forms part of the boundary of the slot.
`Histidine 31 and arginine 39 of the light chain
`are located in positions adjacent to the two
`enolic groups at opposite ends of the xanthonyl
`ring, and thus account for neutralization of one
`of two negative charges in the haptenic di anion.
`Formation of an enol-arginine ion pair in a re(cid:173)
`gion of low dielectric constant may account for
`an incremental increase in affinity of 2-3 orders
`of magnitude in the 4-4-20 molecule relative to
`other members of an idiotypic family of mono(cid:173)
`clonal antifluorescyl antibodies. The phenyl
`carboxyl group offluorescein appears to be hy(cid:173)
`drogen bonded to the phenolic hydroxyl group
`of tyrosine 37 of the light chain. A molecule of
`2-methyl-2,4-pentanediol (MPD), trapped in the
`interface of the variable domains just below the
`fluorescein binding site, may be partly respon(cid:173)
`sible for the decrease in affinity for the hapten
`inMPD.
`
`Key words: antifluorescyl monoclonal anti(cid:173)
`body, high-affinity binding site, ef(cid:173)
`fects of MPD on bapten binding
`
`INTRODUCTION
`The antifluorescein system is well suited for
`studying the molecular basis of antigenic specificity
`because it offers both a wide range of binding affm(cid:173)
`ities (105- 1010 M- 1), and a variety of experimental
`
`© 1989 ALAN R. LISS, INC.
`
`techniques for correlating antigen binding affini(cid:173)
`ties, kinetics, and thermodynamics.1- 9 Further(cid:173)
`more, it has been possible to develop a family of
`idiotypically cross-reactive antibodies in which indi(cid:173)
`vidual monoclonals vary in affinity over a 1000-fold
`range.9 Amino acid sequences recently determined
`for eight of these antibodies in one of our laborato(cid:173)
`ries (E.W.V.) show that at least six were derived
`from the same germline variable genes. Thus, the
`antifluorescein idiotype family can be used to fur(cid:173)
`ther the understanding of both idiotypy and affinity
`maturation. In this report, we describe the three(cid:173)
`dimensional structure of a complex of dianionic flu(cid:173)
`orescein with the antigen-binding fragment from
`the antibody (4-4-20) with the highest affinity in
`this idiotype family.
`The 4-4-20 monoclonal is an IgG2,. (K) antibody
`that binds fluorescein with an association constant
`of 3.4 x 1010 M - 1 in aqueous solution. This affinity
`in 47% (v/v) 2-methyl-2,4-
`decreases 300-fold
`pentanediol (MPD), the solvent used for cocrystalli(cid:173)
`zation of the 4-4-20 Fab with fluorescein hapten.5
`The antibody is a highly specialized molecule that
`does not cross-react with rhodamine compounds. 1•6
`In the formation of a complex with the 4-4-20 anti(cid:173)
`body, fluorescein satisfied criteria for a site-filling
`ligand, with the xanthonyl ring behaving as the
`"immunodominant" moiety.6 Despite its relatively
`large size and distinctive chemical features, fluores(cid:173)
`cein induces a diverse immune response when in(cid:173)
`jected as a conjugate with keyhole limpet hemocya(cid:173)
`nin (KLH).1 •7- 9 Interpretation of binding studies
`have sometimes been ambiguous with heteroge(cid:173)
`neous populations of antibodies. It therefore seemed
`appropriate to determine the mode of binding in a
`single molecular species, particularly one with high
`affinity for fluorescein.
`This crystal system affords a rare opportunity to
`consider both the structural features responsible for
`high-affinity binding and the effecl<> of solvent in
`lowering that affinity. The complex will also be ex-
`
`Received January 20, 1989; revision accepted March 24,
`1989.
`Address reprint requests to A.B. Edmundson, Department of
`Biology, University of Utah, Salt Lake City, UT 84112.
`James N. Herron's present address is Department of Phar(cid:173)
`maceutics, University of Utah, Salt Lake City, UT 84112.
`
`1 of 10
`
`Celltrion, Inc., Exhibit 1086
`
`
`
`272
`
`J . N. HERRON ET AL.
`
`amined to assess the influence of the carrier protein
`(KLH) on the location and orientation of the hapten
`in the combining site. In the preparation of the im(cid:173)
`munogen, molecules of the isothiocyanate derivative
`of fluorescein amine (isomer I) were coupled to
`KLH, with the principal reactions presumed to in(cid:173)
`volve the E-amino group of lysine side chains. The
`4-4-20 hybridoma was obtained by fusion of hyper(cid:173)
`immune splenocytes stimulated with the fl uorescein(cid:173)
`KLH conjugate.
`The present work adds to previous crystallo(cid:173)
`graphic studies of complexes in which small mole(cid:173)
`cules were diffused into crystals of human and mu(cid:173)
`rine immunoglobulin fragments with binding sites
`shaped like cavities or shallow depressions.10- 15
`These studies have helped define the molecular ba(cid:173)
`sis of low- and medium-affinity interactions and
`have provided the background to broaden our under(cid:173)
`standing of antigenic specificity. Block-end types of
`interactions over large external surfaces have been
`studied in cocrystals of Fab molecules and protein
`antigens
`like
`lysozyme and
`influenza neur(cid:173)
`aminidase.16-19 Antibodies binding DNA have
`grooves as potential combining sites, and cocrystals
`of Fahs and oligodeoxynucleotides are currently be(cid:173)
`ing subjected to X-ray analyses.20-23 The structure
`of an unliganded Fab (J539) with specificity for ga(cid:173)
`lactan in carbohydrates has also been determined.24
`A model for the binding of the ligand was proposed
`on the basis of solution studies and the structure of
`the combining site.
`Recently, a single-chain antigen-binding pro(cid:173)
`tein25 was constructed to simulate the Fv fragment
`of the 4-4-20 antibody (an Fv fragment consists of
`the "variable" domains of the heavy and light
`chains). The three-dimensional structure of the 4-
`4-20 Fab should prove very useful in assessing the
`properties and applications of the single chain pro(cid:173)
`tein.
`
`MATERIALS AND METHODS
`Preparation of Fluorescein-Fab Complex
`Procedures for the isolation and purification of the
`4-4-20 monoclonal antibody and fluorescein-Fab
`complex were described in a previous article.5 In
`outline the Fa b fragments were prepared by hydro(cid:173)
`lysis of fluorescein- IgG complexes with papain.
`Liganded Fahs were purified by chromatofocusing
`on 15 ml Pharmacia PBE 94 columns, with a linear
`pH gradient of 9 to 6 formed with Pharmacia Poly(cid:173)
`buffer 96. Because of potential destructive effects of
`dye-sensitized photooxidation by free fluorescein, it
`was necessary to protect the liganded protein from
`light at all stages of the procedures. Columns, tubes,
`and vessels used in affinity chromatography, chro(cid:173)
`matofocusing, dialysis, enzymatic hydrolysis, and
`crystallization trials were all covered with alumi(cid:173)
`num foil. All manipulations and transfers were car(cid:173)
`ried out in reduced light. Under such precautions
`
`the liganded Fab was eluted from the chromatofo(cid:173)
`cusing column as a single component with a pl of
`7.2.
`After dialysis against 50 mM sodium phosphate,
`pH 7.2, the solution of the fluorescein- Fab complex
`was concentrated to 25 mg/ml by ultrafiltration (the
`acceptable range for crystallization was 10-30
`mg/ml). The complex was crystallized by a batch
`method in an environment with strict light and tem(cid:173)
`perature (12-14°C) control. Graded aliquots ofMPD
`were added to 40 µ.l samples of the liganded protein
`in flat-bottomed glass vials. Crystals appeared in 2
`days with final MPD concentrations of 38- 60%
`(v/v), the optimum being - 47%. Bladed crystals
`suitable for X-ray analysis grew to dimensions of0.6
`x 0.35 x 0.3 mm (I x w x d) in 1- 2 months. Green
`fluorescence, attributable to the dissociation of flu(cid:173)
`orescein from the complex in MPD, was observed in
`each crystallization tube.
`
`Collection of X-Ray Diffraction Data
`The fluorescein-Fab complex crystallized in the
`triclinic space group Pl, with a = 58.3, b = 43.9,
`and c = 42.5 A; a = 82.1, 13 = 87.3, and 'Y = 84.6°.5
`Crystals disintegrated at temperatures > 22°C, but
`were mechanically stable even in the X-ray beam
`when the ambient temperature was maintained at
`12- 14°C. Temperature instability of the complex in
`MPD had also been noted in solution.5 For example,
`irreversible increases in the standard free-energy
`changes (~G 0) in the liganded lgG and Fab mole(cid:173)
`cules occurred in 40% MPD at relatively low tem(cid:173)
`peratures (transition temperature of 30°C).
`A single crystal was used to collect X-ray diffrac(cid:173)
`tion data to 2.7 A resolution with a Nicolet P21 dif(cid:173)
`fractometer operated at 40 kV and 35 mA (CuK., ra(cid:173)
`diation). The data set included 11,116 unique
`reflections, of which 9120 (82.0%) were observed at
`intensity levels > 1.5 standard deviations (based on
`counting statistics).
`
`Determination of the Three-Dimensional
`Structure of the Liganded Fab
`The fluorescein-Fab complex crystallized in t he
`same space group (Pl) as the unliganded Fab of the
`BV04-0l IgG2b autoantibody, with specificity for
`single-stranded DNA.5•26 Unit cell dimensions for
`the two crystals were nearly identical and one of
`us (X-M.H.) found by molecular replacement meth(cid:173)
`ods27-30 that the proteins were in the same orienta(cid:173)
`tions in these unit cells. With the refined 2.7 A
`structure of the BV04-0l Fab as starting model, the
`orientation of the 4-4-20 Fab was determined more
`accurately with rotation function programs.
`Crystallographic refinement31•32 of the structure
`was initiated with the X-ray diffraction data5 for the
`4-4-20 Fab and the atomic model of the BV04-0l
`Fab.22•23 After 30 cycles of refinement with 2.7-6.0
`A data (8304 reflections), the amino acid sequences
`
`2 of 10
`
`Celltrion, Inc., Exhibit 1086
`
`
`
`STRUCTURE OF FLUORESCEIN-FAB COMPLEX
`
`273
`
`TABLE I. Refinement Data
`
`Target
`value
`
`Actual rms deviation
`46.6
`0.215
`15.5
`8304
`
`Average !:iF*
`R-factor
`< B > in A2t
`No. of reflections, I > 1.5 a (!)+
`Root mean square (rms) deviations
`From ideal distance (A )
`Bond distance
`Angle distance
`Planar 1-4 distance
`Hydrogen bond distance§
`rms deviation from planarity (Al
`rms deviation from ideal chirality <A 3)
`rms deviation from permitted contact distances (A)
`Single torsion contacts
`Multiple torsion contacts
`Possible hydrogen bond
`rms deviation from ideal torsion angles (0
`15.0
`30.8
`For prespecified <!>, \fl angles
`For planar group (0 or 180)
`6.3
`3.0
`26.9
`15.0
`For staggered group (:t 60 or 180)
`28.1
`For orthonormal group (:t 90)
`15.0
`*The weight for the structure factor refinement was obtained from the equation "' = ( l/a)2, where cr =
`40 - 278 x [sin(6)/>. -
`l/6J.
`t Average temperature factor.
`:l:Resolution limits of 2.7- 6.0 A.
`~Explicit hydrogen bonds for J3·pleated sheets.
`
`0.028
`0.055
`0.032
`0.132
`0.014
`0.210
`
`0.280
`0.400
`0.409
`
`0.03
`0.04
`0.03
`0.05
`0.025
`0.150
`
`0.500
`0.500
`0.500
`
`)
`
`of the light and heavy chains were altered to corre(cid:173)
`spond to those of the 4-4-20 Fab. 34 The model of the
`4-4-20 Fab was improved through alternating cycles
`of refinement and interactive model building on an
`Evans and Sutherland PS300 graphics system with
`the FRODO program. 35·36 Polypeptide backbones
`and amino acid side chains were fitted to '2F0 - Fc
`maps, in which F 0 and Fe were observed and calcu(cid:173)
`lated structure factors. Fluorescein was located in a
`6F (F0 - FJ map, for which the phase angles were
`calculated from the refined atomic model of the pro(cid:173)
`tein (atomic coordinates for fluorescein were omit(cid:173)
`ted). The ligand-protein complex was subjected to
`additional cycles of refinement until the R factor
`(2. II F 0 1-1 F eII 12. I Fo I ) began to plateau at its cur(cid:173)
`rent value of 0.180 (before idealization of bond
`lengths and angles; 0.215 after).
`
`RESULTS
`Description of the Three-Dimensional
`Structure of the Fluorescein-Fab Complex
`
`The results of the crystallographic refinement of
`the complex are presented in Table I. This structure
`could be refined very quickly because of the great
`similarities with the structure of the unliganded
`BV04-01 Fab.
`Figure 1 contains the three-dimensional "cage"
`electron density to which a skeletal model of fluo(cid:173)
`rescein was fitted by interactive computer graphics.
`The torsion angle measured between the xanthonyl
`and benzoyl rings was 73° in the bound hapten.
`
`An o:C skeletal model of the 4-4-20 Fab is shown
`as a stereo pair in Figure 2, with a model of fluores(cid:173)
`cein codisplayed in the binding site. Tracings of the
`o:C chains of the 4-4-20 and BV04-0l Fabs are su(cid:173)
`perimposed in Figure 3. Details of the structures of
`the ligand and combining site are presented in Fig(cid:173)
`ure 4 (skeletal models). Solvent-accessible sur(cid:173)
`faces37 are illustrated in Figure 5. Amino acid
`sequences34 for the hypervariable regions are listed
`in Figure 6.
`Polypeptide chains could be traced unambigu(cid:173)
`ously in both the light and heavy chains of the 4-
`4-20 complex. Significantly, the third hypervariable
`loop, which was difficult to follow in the heavy chain
`of the BV04-0l Fab, was well defined in the fluores(cid:173)
`cein- Fab complex. In the presence of ligand, constit(cid:173)
`uents of this loop were found to have small temper(cid:173)
`ature factors (B values), characteristic of regions
`with low mobility.
`
`Comparison of the Stru ctures of the 4-4-20 and
`BV04·01 Fahs
`
`The 4-4-20 Fab is an extended molecule in which
`the pseudotwofold axes between the pairs of variable
`and constant domains are nearly colinear (i.e., the
`measured "elbow bend" angle between the two
`pseudodiads is 171°). Except for the third hypervari(cid:173)
`able loops, in which the 4-4-20 heavy chain is
`shorter than the BV04-01 sequence by three resi(cid:173)
`dues, the o:C tracings are remarkably similar in the
`two proteins. This similarity is readily understand-
`
`3 of 10
`
`Celltrion, Inc., Exhibit 1086
`
`
`
`274
`
`J. N. HERRON ET AL.
`
`1
`
`2
`
`Rg. 1. Three-dimensional electron density (orange) corre·
`sponding to fluorescein (green) in the combining site of the 4·4·20
`Fab. This electron density was obtained in a ditterence Fourier
`map alter crystallographic refinement of the ligand-protein com·
`plex. A skeletal model of fluoresoein was fitted to the electron·
`density by interactive computer graphics.
`
`Fig. 2. Stereo diagram of o: C tracings of the light (blue) and
`heavy (red) chains of the 4·4·20 Fab, with fluorescein (green) in
`the combining site. The "variable" domains are at the top and the
`"constant" domains are at the bottom. Disulfide bonds are repre·
`sented by yellow bars.
`
`4 of 10
`
`Celltrion, Inc., Exhibit 1086
`
`
`
`STRUCTURE OF FLUORESCEIN- FAB COMPLEX
`
`275
`
`Fluorescein Binding Site
`Fluorescein fits into a slot lined by constituents of
`both the light and heavy chains (see Figs. 4 and 5),
`and participates in a network of interacting aro(cid:173)
`matic groups with the protein. There are 68 pairs of
`atoms of the ligand and protein separated by < 4 A.
`Of the potential interactions, 42 are associated with
`the xanthonyl moiety and 26 with the benzoyl group
`of fluorescein. This distribution is consistent. with
`the assignment of the xanthonyl moiety as the "im(cid:173)
`munodominant" portion of the hapten.
`Tryptophan LlOl forms the bottom of the s lot and
`tryptophan H33 and tyrosine L37 provide the sides.
`The width of the slot, as measured from the distance
`between the centroids of the rings oftryptoph~n H33
`and tryosine L37, is about 9 A. The phenolic ring of
`tyrosine L37 is stacked with the xanthonyl moiety of
`fluorescein. Tryptophan H33 interacts with the hap(cid:173)
`ten to produce an aromatic pair of the type described
`by Burley and Petsko.38 One enolic group (oxygen l )
`on the xanthonyl ring is only 3.0 A from an imid(cid:173)
`azolium nitrogen of histidine 131. The other enolic
`group (oxygen 3) is sufficiently close (2.8 A) to the
`guanido group of arginine 139 for an electrostatic
`interaction and is also within hydrogen bonding dis(cid:173)
`tance of serine L96. Thus one of the ligand's two
`negative charges is neutralized by the antibody. Ox(cid:173)
`ygen 2, in the ether linkage at the top of the middle
`ring of the xanthonyl moiety, does not participate in
`interactions with the protein.
`Surprisingly, the second negative charge on fluo(cid:173)
`rescein (the carboxylic acid) is not formally neutral(cid:173)
`ized by a protein substituent. Instead, this group
`faces the solvent at an open end of the slot. Partial
`charge compensation is achieved by the formation of
`a hydrogen bond between the phenolic hydroxyl
`group of tyrosine 137 and one of the carboxylic acid
`oxygen atoms (the electron density in a 2F0 - Fc map
`is continuous between the two sets of atoms).
`Tyrosine 103 extends toward the phenyl ring of
`fluorescein and tyrosine 102 contributes to the
`structural integrity of the binding site without ac(cid:173)
`tually being in contact with the hapten (see Fig. 4).
`There is a clear solvent channel for a lysine side
`chain to be connected to an isothiocyanate group in
`the para position of the phenyl ring, as would be
`expected for a hapten coupled to the KLH carrier
`protein in the imrnunogen. This solvent channel is
`shown in Figure 5.
`Glutamine H50, which replaces an important
`arginine residue in the putative binding site of the
`BV04-0l Fab, is mostly buTied in the presence of
`fluorescein in the 4-4-20 Fab. However, the polar
`end of the side chain (the carbonyl oxygen of the
`amide group) is within hydrogen bonding distance of
`the indole nitrogen of tryptophan H33. Arginine
`H52, another key residue in the BV04-0l binding
`region, is salt bridged to glutamic acid H59 (an ala(cid:173)
`nine in BV04-01 ).
`
`Fig. 3. Superimposed aC tracings of the 4-4-20 and BV04-01
`Fabs. the latter having specificity for single-stranded DNA. The
`two chains of the 4-4-20 Fab are colored red (heavy) and blue
`(light). The 4-4-20 Fab was cocrystallized with fluorescein in 2-
`methyl-2,4-pentanediol and the BV04-01 Fab was crystallized
`without ligand in ammonium sulfate. Yet the two crystals were
`nearly isomorphous and the structures of the Fabs were very
`similar.
`
`able in the light chains, which differ in only six
`amino acid positions.:i4 However, there are 42 re(cid:173)
`placements in the variable domain of the 4-4-20
`heavy chain. Substitutions that appear to be critical
`for the binding of fluorescein are arginine 39 for
`histidine and tryptophan 101 for leucine in the light
`chain; tryptophan 33 for alanine and glutamine 50
`for arginine in the heavy chain. Because of differ(cid:173)
`ences in the lengths of the third hypervariable loops,
`tyrosine 102 and 103 have no direct counterparts in
`the BV04-01 Fab. Interestingly, arginine 39 is re(cid:173)
`placed by histidine in other members of the antiflu(cid:173)
`orescyl idiotype family and by glutamine in the Meg
`Bence-Jones dimer. This glutamine is located out(cid:173)
`side the regions available for ligand binding. 12
`
`5 of 10
`
`Celltrion, Inc., Exhibit 1086
`
`
`
`276
`
`J . N. HERRON ET AL.
`
`Fig. 4. Stereo diagram of the hapten binding site in the 4·4·20
`Fab. light (L) chain constituents are blue, heavy (H) chain com(cid:173)
`ponents are red, and fluorescein is green. The xanthonyl (three
`ring) group of fluorescein is flanked by tyrosine L37 on the left and
`tryptophan H33 on the right, with tryptophan L101 forming the
`
`bottom of the slot. Enolic oxygen atoms on opposite corners of the
`xanthonyl group point toward histidine L31 (left) and arginine L39
`(right). The phenyl carboxyl group of fluorescein (single ring) is
`located below tyrosines H103 (left) and H102 (right).
`
`of the unliganded BV04-0l Fab. Light chain constit(cid:173)
`uents of the 4-4-20 Fab were found to be involved in
`a large proportion (23 of the 26 examples in the sur(cid:173)
`vey) of major packing interactions. The light chain
`is also prominent in crystal packing of the unli(cid:173)
`ganded BV04-0l. The close similarities in amino
`acid sequences and conformations of the two light
`chains are apparently reflected in the packing of the
`parent Fabs, despite the differences in crystallizing
`media.
`
`Sequestering of an MPD Molecule in the
`V L -V 8 Interface
`
`In the interface of the variable domains below the
`fluorescein binding site (closest distance ~4.6 A),
`there was a module of electron density that could not
`be assigned to either protein or hapten. This module
`had the size and shape expected for a molecule of
`MPD, and persisted in maps calculated after ad(cid:173)
`vanced stages of the crystallographic refinement.
`The cage electron density for the putative solvent
`molecule is shown with the surrounding protein con(cid:173)
`stituents in Figure 7. MPD was omitted from the
`model used to calculate the phases for the difference
`Fourier map. The binding site for MPD was lined by
`both polar and apolar residues, such as threonine 99,
`serine 101, and tryptophans 4 7 and 108 of the heavy
`chain, and arginine 39, tyrosine 41, and tryptophan
`101, and phenylalanine 103 of the light chain. Hy(cid:173)
`droxyl groups on threonine H99 and tyrosine L41
`were located within hydrogen bonding distances of
`the 2- and 4-hydroxyl groups of MPD.
`
`Fig. 5. Solvent-accessible surface dot representation37 of the
`hapten combining site, with a skeletal model of fluorescein super(cid:173)
`imposed. Note the free space above the para position of the phe(cid:173)
`nylcarboxyl group of fluorescein. In the immunogen used to elicit
`the production of the 4-4-20 antibody, a lysine side chain on a
`carrier protein would be linked to an isothiocyanate group in the
`para position of the hapten. Lysine would presumably occupy the
`free space when the hapten entered the antibody combining site.
`
`Crystal Packing Interactions in the 4-4-20 and
`BV04-01 Systems
`Analyses of packing interactions revealed struc(cid:173)
`tural features that are conducive to the formation of
`4-4-20 Fab crystals nearly isomorphous with those
`
`6 of 10
`
`Celltrion, Inc., Exhibit 1086
`
`
`
`STRUCTURE OF FLUORESCEIN- F AB COMPLEX
`
`277
`
`LIGHT CHAIN
`
`CDR1
`CRSSOSLVHSNGNTYLRW
`30
`35
`40
`25
`lt
`25 27 o b c d e 28 30
`35
`t
`
`HEAVY CHAIN
`
`CD R1
`SOY WMN
`"30
`35
`t 30
`35
`
`CDR2
`YKVS NR FSG
`lt 55
`60
`t 50
`55
`
`CDR3
`CSOS TH VPWT
`*
`95
`100
`90
`95
`
`CDR2
`AOIRNKPYNYE TYYS O SVK G R
`lt 50
`55
`60
`65
`t 50 52 0 b c 53 55
`
`60
`
`65
`
`COR3
`GSYYGMDYW
`* 100
`105
`100 0 b c d 101
`t
`
`Fig. 6. Amino acid sequences of the hypervariable regions
`(complementarity-determining regions or CDR45
`) of the light and
`heavy chains of the 4-4-20 Fab. These sequences were deter-
`
`mined by Bedzyk et al.3 4 (')The strict sequential numbering sys(cid:173)
`tem used in our computers and graphics terminals. (t) The more
`widely used numbering scheme of Kabat et al.45
`
`DISCUSSION
`A simple, yet restricted combining site seems well
`suited for a specialized high-affinity antibody for
`fluorescein. Space available for binding in the site is
`limited. While geometrically optimized, the aro(cid:173)
`matic, electrostatic, and hydrogen bonding interac(cid:173)
`tions are relatively few in number. We suspect that
`the formation of such a tight complex was accompa(cid:173)
`nied by conformational adjustments in the protein.
`Recent immunological studies support this view. For
`example, antibodies elicited to the liganded site of
`the 4-4-20 antibody were not reactive with the non(cid:173)
`liganded, idiotypic state. 39 These studies suggest the
`generation of new epitopes on ligand binding as a
`consequence of induced conformational changes.
`It is interesting to ask what factors contribute to a
`high-affinity site. There is a red shift in the absorp(cid:173)
`tion spectrum when fluorescein is bound to the
`4-4-20 and other monoclonal antifluorescyl anti(cid:173)
`bodies.2·40 This shift has been attributed to hydro(cid:173)
`phobic effects involving tryptophan side chains. The
`model of the complex supports such a view, since two
`trytophan residues are located in close juxtaposition
`to the ligand in the 4-4-20 combining site. However,
`strategically placed tryptophans are not sufficient to
`account for the high affinity of 4-4-20. For example,
`the 9-40 antifluorescyl antibody has tryptophan res(cid:173)
`idues in homologous positions in its amino acid se(cid:173)
`quence and a lower affinity for the ligand than
`4-4-20.41
`Quenching of fluorescence is characteristic of an(cid:173)
`tifluorescyl antibodies. The maximum quenching
`constant for the 4-4-20 molecule is greater than the
`Qmax values of some antibodies (e.g., 20-4-4), but
`slightly smaller than those for other antibodies (e.g.,
`
`20-20-3) with lower affinities for fluorescein.2·40
`Moreover, Qmax for 4-4-20 actually increases in
`MPD while the ligand affinity decreases.5 These re(cid:173)
`sults indicate that there is not a strong correlation
`between fluorescence quenching and high affinity.
`Interactions resulting in quenching appear to be
`very diverse. Tryptophan has been invoked as a pos(cid:173)
`sible participant in such interactions and in one case
`(20-20-3) histidine was suggested to be involved in
`the protonation of the enolic group of bound
`fluorescein. 2 The 4-4-20 active site contains both
`tryptophan and histidine residues in positions that
`should be favorable for fluorescence quenching.
`Model-based hypotheses seem to be especially use(cid:173)
`in explaining deuterium fluorescence en(cid:173)
`ful
`hancement2,42 and iodine quenching43 of bound
`ligand in antifluorescyl antibodies. The liganded 4-
`4-20 molecule shows about 188% fluorescence en(cid:173)
`hancement in deuterium oxide (relative to water).
`Such experiments afford a measure of the degree of
`hydrogen bonding between ligand and protein. The
`enhancement for 4-4-20 was 10 times greater than
`values for four other antifluorescyl monoclonal an(cid:173)
`tibodies and hydrogen bonding was therefore consid(cid:173)
`ered to he very important for ligand binding.2 The
`structure of the complex is consistent with this pro(cid:173)
`posal and can be used to identify the most prominent
`hydrogen bonds (enol with histidine L31 and phenyl
`carboxyl group with tyrosine L37). Only 2.5%
`quenching of bound fluorescein was noted when the
`liganded 4-4-20 IgG was titrated with potassium
`iodide.9·43 This observation indicated that fluores(cid:173)
`cein bound in the 4-4-20 active site was essentially
`inaccessible to iodide. Quenching was higher (8.4-
`15.5%) in four other antibodies with lower affinities
`
`7 of 10
`
`Celltrion, Inc., Exhibit 1086
`
`
`
`278
`
`J . N. HERRON ET AL.
`
`Fig. 7. Stereo diagram of the cage electron density (lavender)
`corresponding to a molecule of MPD in its binding site just below
`that of fluorescein in the V domain interface. Key residues of the
`light (blue) and heavy chains (red) are labeled with three-letter
`
`abbreviations. The hydroxyl groups of threonine H99 and tyrosine
`L41 may form hydrogen bonds with the 2-and 4-hydroxyl groups
`of MPD.
`
`for fluorescein.9 Figures 4 and 5 show a very tight fit
`of the ligand in the site, and the presence of tyrosine
`102 and 103 further shields the bound ligand from
`additional reactants. These tyrosine residues are not
`present in the sequence of the 9-40 antibody, in
`which bound fluorescein is more accessible to
`iodide.9
`The interactions of tyrosine L37 with fluorescein
`appear to be important in orienting the ligand in the
`high-affinity site. The stacking of the phenyl group
`of tyrosine with the xanthonyl ring was further sta(cid:173)
`bilized by a hydrogen bond between the phenolic hy(cid:173)
`droxyl group and the phenylcarboxyl moiety of f1u(cid:173)
`orescein. The dual interaction of fluorescein with
`tyrosine also played a major role in fixing the tor(cid:173)
`sion angle between the xanthonyl and phenyl rings,
`which would have greater freedom to rotate in solu(cid:173)
`tion.
`Studies of the effects of pH on lifetimes of dissoci·
`ation indicated that ionizable groups (pK > 8.0) in
`
`the 4-4-20 active site contribute to the binding of
`fluorescein.2 Lifetime maxima at pH 6.5-7.0 sug(cid:173)
`gested that the xanthonyl moiety is negatively
`charged when bound to the 4-4-20 (the pK of the
`enolic group is 6.714
`). In the 3-D structure of the
`complex the two enolic groups of fluorescein are
`bracketed by the ionizable side chains of histidine
`L31 and arginine 139. At the pH used for crystalli(cid:173)
`zation the histidine side chain would be expected to
`be in the uncharged form. The formation of an ion
`pair with arginine probably makes a very signifi(cid:173)
`cant contribution to the binding energy, particularly
`in a region of low dielectric constant. In the 9-40
`antibody this arginine is replaced by histidine,41 a
`substitution that could partially account for the de(cid:173)
`crease in affinity relative to 4-4-20.
`Rhodamine 110 and rhodamine B, structural an(cid:173)
`alogs of fluorescein with amino (110) and diethyl(cid:173)
`amino (BJ groups substituted for the enolic oxygen
`atoms, do not bind to 4-4-20.2 Acylamino groups
`
`8 of 10
`
`Celltrion, Inc., Exhibit 1086
`
`
`
`STRUCTURE OF FLUORESCEIN- FAB COMPLEX
`
`279
`
`would have formal positive charges at relatively low
`pH values, ( < 5), at which there would be electro(cid:173)
`static repulsion by like charges on the histidine L31
`and arginine L39 side chains. At pH 8 the only
`charge on a rhodamine molecule would reside on the
`phenylcarboxyl group. A monoanionic ligand would
`be incompatible with the pH profile of binding,
`which indicates a strong dependence on the presence
`of a dianion. In the case of rhodamine B, there is a n
`additional steric factor to consider. For example, tet(cid:173)
`ramethyl rhodamine formed a tight complex with
`the Meg Bence-Jones dimer, but the tetraethyl de(cid:173)
`rivative (rhodamine B) could not be accommodated
`in the binding site. Exposure to rhodamine B led to
`destruction of a crystal of the Meg dimer in 5
`hours. 13
`The crystal structure of the fl uorescein-Mcg com(cid:173)
`plex provides further insight into the minimal re(cid:173)
`quirements for low-affinity binding of the ligand. As
`in 4-4-20, the ligand was accommodated in a net(cid:173)
`work of aromatic residues. The xanthonyl ring of
`fluorescein was more deeply immersed in the cavity
`than the benzoyl moiety and participated in more
`interactions (14 vs 6) with the protein. A tyrosine
`side chain was wedged between the xanthonyl and
`benzoyl groups and largely dictated the torsion an(cid:173)
`gle of 57° between these two rings. At the pH (6.2) of
`the crystals fluorescein was a monoanion and the
`charged group (phenylcarboxyl moiety) was oriented
`toward bulk solvent at the entrance of the binding
`cavity. In summary, the low-affinity site was more
`voluminous and there were fewer ligand- protein in(cid:173)
`teractions. Aromatic rings did not stack, and poten(cid:173)
`tial negative charges were not balanced by basic
`side chains on the protein.
`Thermodynamic studies indicated that the tem(cid:173)
`perature stabilities of both the intact 4-4-20 anti(cid:173)
`body and Fab were significantly decreased in the
`presence of 40% (v/v) MPD.5 These studies also sug(cid:173)
`gested that the decrease in affinity for fluorescein in
`MPD was mainly attributable to conformational
`changes in the protein