Structural Basis for Recognition of CD20 by Therapeutic Antibody Rituximab*

Rituximab is a widely used monoclonal antibody drug for treating certain lymphomas and autoimmune diseases. To understand the molecular mechanism of recognition of human CD20 by Rituximab, we determined the crystal structure of the Rituximab Fab in complex with a synthesized peptide comprising the CD20 epitope (residues 163-187) at 2.6-Å resolution. The combining site of the Fab consists of four complementarity determining regions that form a large, deep pocket to accommodate the epitope peptide. The bound peptide assumes a unique cyclic conformation that is constrained by a disulfide bond and a rigid proline residue (Pro172). The 170ANPS173 motif of CD20 is deeply embedded into the pocket on the antibody surface and plays an essential role in the recognition and binding of Rituximab. The antigen-antibody interactions involve both hydrogen bonds and van der Waals contacts and display a high degree of structural and chemical complementarity. These results provide a molecular basis for the specific recognition of CD20 by Rituximab as well as valuable information for development of improved antibody drugs with better specificity and higher affinity.

CD20 is a pan-B cell marker expressed from pre-B cells until B cells are differentiated into plasma cells (1). It is a tetraspan membrane protein that is predicted to contain a large extracellular loop (about residues 142 to 182) and to form oligomers on the cell surface (2)(3)(4). Although the precise function of CD20 remains unclear, biochemical and cell biological data have shown that it seems to form or regulate a voltage-independent calcium channel (3,5). Despite the limited knowledge about its function, several lines of evidence have clearly demonstrated that CD20 is an ideal target for passive immunotherapy of B-cell lymphoma: it is highly expressed in more than 80% of the B-cell lymphomas but not in stem cells, pro-B cells, normal plasma cells, or other normal tissues; it remains on the cell surface without substantial internalization after cross-linking with antibodies; and it is not shed to the circulation to inhibit the antibody therapy (6 -8).
The CD20-targeted chimeric monoclonal antibody (mAb) 3 Rituximab (Rituxan, IDEC-C2B8) was the first Food and Drug Administration approved mAb drug for the treatment of malignancy. Although it was originally used for treating low-grade non-Hodgkin lymphoma, Rituximab has been proven to be also effective against other types of lymphomas (9,10) and some autoimmune diseases (11)(12)(13). Multiple mechanisms have been proposed for the action of Rituximab in the depletion of B cells including its ability to mediate complement-dependent cytotoxicity and antibody-dependent cellular cytotoxicity, and to induce cell apoptosis (reviewed in Refs. 14 and 15). With the expansion of the clinical application of Rituximab in the treatment of lymphoproliferative diseases, it has been noticed that intensity of CD20 expression on B cells varies in patients that may affect the binding and efficacy of Rituximab therapy (16,17). Therefore, it is rational to expect that new antibodies with higher affinity and better specificity developed based on Rituximab might be beneficial in clinical use, especially for patients who have low expression levels of CD20.
Besides Rituximab and Zevalin (the prototype of Rituximab 2B8 attached by a radioactive substance 90 Y), another mAb specific to human CD20, namely B1, in both native and radioderivative forms (Bexxar, Tositumomab and 131 I-Tositumomab), was approved by the Food and Drug Administration for the treatment of non-Hodgkin lymphoma in 2003 (18). These antibody drugs along with some other mAbs against CD20 such as 1F5, AT80, and 2H7 were suggested to most likely recognize the same region (Tyr 165 to Tyr 182 ) of the large extracellular loop of human CD20 with fine specificities (2,19). However, these antibodies vary considerably in their functional activities. For example, treatment of B cells with most mAbs promotes segregation of CD20 into detergent-insolu-ble lipid raft, whereas B1 is the exception. This property seems to be correlated with the ability of the antibodies to mediate complement-dependent cytotoxicity but irrespective of activation of cell apoptosis pathways (20 -22).
To understand the functional diversity of CD20 mAbs and the underlying mechanisms, identification of the epitope of CD20 recognized by Rituximab and other CD20-targeted mAbs has raised great interest in recent years. Sequence comparison of human and murine CD20 reveals that although the two species share 73% sequence identity, the large extracellular loop is less conserved as 16 of the approximate 43 amino acids are different (19). Mutagenesis studies by exchanging variant residues of the large extracellular loop between human CD20 and mouse CD20 at the equivalent position indicate that residues Ala 170 and Pro 172 of human CD20 are critical determinants for the CD20 epitope (19). Using a process of biopanning of a phage display peptide library consisting of randomized 7-mer cyclic peptides, it has been shown that an NPS motif corresponding to 171 NPS 173 of human CD20 is essential for Rituximab binding and Ala 170 can be substituted by Ser (4). Similar studies have also defined a discontinuous epitope that comprises 170 ANPS 173 and 182 YCYSI 185 of CD20 joined together spatially by a disulfide bond between Cys 167 and Cys 183 (23). However, the underlying mechanism of the recognition and binding of Rituximab with the CD20 epitope remains unclear.
We report here the crystal structure of the Rituximab Fab fragment in complex with an epitope peptide of the large extracellular loop (residues 163-187) of CD20 that provides the molecular basis for the antigen-antibody recognition and binding. Analysis of the complex structure explains very well biochemical data from the epitope-mapping studies and provides useful hints for the design and development of improved antibody drugs.

EXPERIMENTAL PROCEDURES
Preparation of Antibody and Peptide-Rituximab was purchased from Roche. The Fab fragment was obtained by papain digestion of Rituximab and purified by cation exchange chromatography using SP-Sepharose FF (GE Healthcare) followed by hydrophobic interaction chromatography using phenyl-Sepharose HP (GE Healthcare). The purity and homogeneity of the Fab fragment was characterized by SDS-PAGE and dynamic light scattering analyses. The protein sample was concentrated to 8 mg/ml and then exchanged into a stock buffer (100 mM NaCl and 10 mM Tris-HCl, pH 8.0) for crystallization. The amino acid sequence of the Fab fragment was determined according to United States patent 5843439 (42).
A 25-mer cyclic peptide (NIYNCEPANPSEKNSPSTQYCY-SIQ) corresponding to residues 163-187 of the large extracellular loop of human CD20 was synthesized in which an intrachain disulfide bond was introduced between Cys 167 and Cys 183 (Shanghai Science Peptide Biological Technology). The quality of the peptide was determined by analytical reverse-phase chromatography and mass spectral analysis with a purity of greater than 95%.
Crystallization and Diffraction Data Collection-Initial crystallization trials of the Rituximab Fab fragment itself yielded large crystals that, however, did not diffract x-ray beyond 7-Å resolution and could not be used for structure determination. For co-crystallization experiments, the purified Rituximab Fab and the epitope peptide were mixed at a molar ratio of 1:5 at 4°C for 12 h. Co-crystallization was carried out using the hanging drop vapor diffusion method by mixing equal volumes of the protein/peptide mixture solution and a reservoir solution (0.2 M calcium acetate, 0.1 M sodium cacodylate, pH 6.5, and 18% PEG8000). Square shaped crystals grew to a maximum size of 0.1 ϫ 0.1 ϫ 0.2 mm 3 at 4°C in 2 weeks. For diffraction data collection, crystals were cryostabilized by Paratone-N (Hampton Research) and then flash-cooled to Ϫ170°C. Diffraction data were collected to 2.6-Å resolution at beamline NW12 of Photon Factory, Japan, and processed using suite HKL2000 (24). Statistics for diffraction data are summarized in Table 1.
Structure Determination, Refinement, and Analysis-The structure of the Rituximab Fab in complex with human CD20 epitope peptide was solved using the molecular replacement method as implemented in Phaser (25). The structure of the Fab fragment of human mAb A5B7 (26) (Protein Data Bank code 1AD0) was used as the search model. Structure refinement was carried out by CNS using standard protocols consisting of conjugate-gradient energy minimization, torsion-constrained molecular dynamics simulated annealing, group B factor refinement, and individual B factor refinement (27). Free R factor was calculated using 5% randomly selected reflections. After several cycles of manual model building using program O (28), the electron density was further improved and clear enough for tracing the epitope peptide without ambiguity. The stereochemistry of the structure model was analyzed with Procheck (29). Statistics of the structure refinement are also summarized in Table 1. Structural analysis was mainly performed using CNS (27) and programs in the CCP4 suite (29). The elbow angle of the Fab fragment was calculated with the method described by Stanfield et al. (30). Figures were prepared using programs Ribbons (31) and Pymol (32).

RESULTS AND DISCUSSION
Overall Structure of the Rituximab Fab-CD20 Epitope Peptide Complex-Recent studies have identified the epitope of CD20 recognized by Rituximab being a sequence motif located at the large extracellular loop of CD20 consisting of 170 ANPS 173 (4,19,23). To understand the structural basis of the recognition of CD20 by Rituximab, we synthesized a 25-mer peptide mimic of the epitope of CD20 comprising the CD20 sequence from Asn 163 to Gln 187 (numbered according to the CD20 sequence). To mimic the conformation of the CD20 epitope, an intrachain disulfide bond was introduced between residues Cys 167 and Cys 183 of the synthesized peptide because such linkage has been found in human CD20 expressed in Escherichia coli and probably exists naturally (33). This disulfide linkage has also been implicated to play an important role in the recognition and binding of the epitope by Rituximab because disruption of the disulfide bond abolishes the binding of CD20 to Rituximab and reconstruction of the disulfide bond can partially restore the binding (33). The peptide was conjugated to a protein carrier keyhole limpet hemocyanin and shown to have similar reactivity with Rituximab as the keyhole limpet hemocyanin-conjugated cyclic peptides Rp15-C and Rp3-C (4) by the enzyme-linked immunosorbent assay binding assay (supplementary Fig. S1A). Specificity of the reactivity was also confirmed by peptide blocking assays with immunofluorescence and complement-dependent cytotoxicity using Raji cells (supplementary Fig. S1, B and C).
The CD20 epitope peptide was co-crystallized in complex with the Rituximab Fab fragment. The structure of the complex was solved using molecular replacement and refined to a resolution of 2.6 Å with an R factor of 23.9% and a free R factor of 29.6% (Table 1). There are two Fab-peptide complexes (A and B) in the crystallographic asymmetric unit. The two complexes adopt very similar conformation (root mean square deviation of 0.96 Å for 449 C␣ atoms) except that the heavy chain in complex A contains three more residues at the C terminus than in complex B and two extra residues can be visualized at the N terminus of the peptide chain in complex B than in complex A. Due to the better electron density quality, the structure model of complex A will be used for further structural analysis and discussion. Chain identifiers L, H, and P are designated to identify the light chain and heavy chain of the Fab fragment and the epitope peptide, respectively (Fig. 1).
Overall the Rituximab Fab has very good electron density, except residues from Ser H135 to Gly H141 of a loop in the constant region that have high B factors above 80 Å 2 compared with the average B factor of 49 Å 2 for the whole model. The Rituximab Fab has a canonical immunoglobulin fold consisting of four ␤-barrel domains (Fig. 1A). The light chain comprises residues Leu 1 to Leu 213 that fold into the V L and C L domains, and heavy chain residues His 1 to His 224 that fold into the V H and C H domains. The elbow angle made by the two pseudo 2-fold axes that define the relative disposition of V H to V L and C H to C L is about 139°. There are four intra-domain disulfide bonds between Cys L23 and Cys L87 , Cys L133 and Cys L193 , Cys H22 and Cys H96 , and Cys H148 and Cys H204 , and one inter-domain disulfide bond between Cys L213 and Cys H224 . Like other Fab structures, residue Thr 50 of the light chain (Thr L50 ) lies in the disallowed region of the Ramachandran plot and forms part of the classic ␥-turn (34). The complementarity determining regions (CDRs) of the Rituximab Fab have ordinary length without unusual residues according to Kabat sequence data base searching (35). The CDR loops L1, L2, L3, H1, and H2 belong to Chothia canonical classes (34) 1, 1, 1, 1, and 2, respectively. The CDR loops L3, H1, H2, and H3 together form a large, deep pocket to accommodate the epitope peptide, whereas loops L1 and L2 are located behind loops L3 and H3 (Fig. 1A).
The electron density for the bound epitope peptide is well defined without ambiguity in the positioning of the main chains and side chains (Fig. 1B). Residues Cys P167 and Cys P183 of the peptide form a disulfide bond that drags the termini of the peptide together covalently and makes the peptide to adopt a unique cyclic conformation (Fig. 1C). The N-terminal part of the peptide (Cys P167 to Asn P171 ) forms a short coil that is stabilized by the tense restraints of the disulfide bond and the hydrogen-bonding interactions between residues of the coil and the CDRs of the Rituximab Fab. The middle part of the peptide forms a short 3 10 helix (Pro P172 to Glu P174 ) and a small loop (Lys P175 to Ser P177 ) that are stabilized by hydrogen-bonding interactions with the other regions (Table 2 and Fig. 1C). The C-terminal part of the peptide (Pro P178 to Tyr P184 ) forms a short ␣-helix of hydrophobic nature (Tyr P182 , Tyr P184 , and Ile P186 ). Considering it is at the end of the extracellular loop and must be close to the cell membrane, this short ␣-helix might be the extension of a long transmembrane ␣-helix of CD20 as predicted by PredictProtein Server (36).
Interactions between the Rituximab Fab and the Epitope Peptide-Rituximab can bind to CD20 on B cells with a binding affinity of 5 nM (37). In the complex structure, the epitope peptide of human CD20 is bound at the large pocket formed by CDR loops L3, H1, H2, and H3 of the Rituximab Fab ( Fig. 2A), which is consistent with the observation that the heavy chain CDRs usually make more contributions than the light chain CDRs in antigen binding especially when Fab binds with a small antigen (38,39). The binding of the epitope peptide with the Fab buries a solvent accessible surface area of about 440 Å 2 (calculated with a probe radius of 1.4 Å) which is about 23% of the peptide surface (1911 Å 2 ) or 2.3% of the Fab surface (19403 Å 2 ). Although the buried surface area is within the average range of the protein-peptide complexes (400 -700 Å 2 ) (40), the peptide fits the CDR regions of the Rituximab Fab quite well with a high degree of structural and chemical complementarity (Fig. 2B) as indicated by the high shape complementarity value (Sc) of 0.83 (calculated with default parameters) compared with the average Sc value of 0.64 -0.68 for antibody-antigen complexes (41).
The CDR loops L3, H1, H2, and H3 of the Fab participate in interactions primarily with four residues, 170 ANPS 173 , of the epitope peptide that have been shown to be a critical motif on

Structure of Rituximab Fab-Epitope Peptide Complex
the CD20 surface for antibody recognition (2,4,23). Residues of the motif form a network of hydrogen bonds with residues of the surrounding CDR loops (Table 3 and Fig. 2, C and D). Specifically, the side chain of Asn P171 forms two hydrogen bonds with the side chains of Ser H99 and Trp H106 of the H3 loop, respectively. The main chain carbonyl of Pro P172 forms a hydrogen bond with the side chain of Asn H55 of loop H2 via a water molecule; and the main chain amide and side chain O ␥ of Ser P173 make two hydrogen bonds with the side chain of Asn H33 of loop H1, respectively. Moreover, the residues flanking the motif including Glu P168 , Pro P169 , and Lys P175 also contribute to the interactions of the peptide with the Fab by forming hydrogen bonds with Asn L93 , Ser H59 , and Thr H58 , respectively (Table  3 and Fig. 2, C and D). In addition to these hydrogen-bonding interactions, extensive van der Waals contacts are observed between residues 168 and 179 of the peptide and the Fab (Table  4). In particular, the 170 ANPS 173 motif contributes 79 of the 97 van der Waals contacts between the peptide and the Fab (Table  4). These interactions can partially explain the high affinity of Rituximab with human CD20. In addition, Pro P172 of the 170 ANPS 173 motif that has been shown to play an essential role in the antigen-antibody recognition (4,19), is located at the bottom of the CDR pocket (Fig.  2B) formed by residues Ala H50 , Tyr H52 , Asp H57 , and Ser H59 of loop H2 of the Fab and residue Asn H33 of loop H1, and has both hydrophilic and hydrophobic interactions with the surrounding residues of the Fab. The special position of Pro P172 at the turn of the 3 10 helix and the rigid conformation of proline might play an important role in maintaining the unique conformation of the peptide and hence in the recognition and binding of Rituximab.
Structural Basis for the Specificity of Rituximab-Human CD20 is an important drug target for the treatment of lymphomas. Rituximab is the first Food and Drug Administration approved mAb drug against CD20 for the treatment of B cell non-Hodgkin lymphoma, and now has been used to treat autoimmune diseases and reduce the alloreaction in organic transplantations. Our crystal structure of the Rituximab Fab in complex with a 25-mer peptide mimicking the epitope on the large extracellular loop of human CD20 has provided a molecular basis for understanding the underlying mechanisms of the recognition and binding of CD20 with its antibodies and will be valuable in the modification of Rituximab for development of more effective mAb drugs against non-Hodgkin lymphoma and other diseases.
Screenings of phage display peptide libraries that can express 7-mer cyclic and 7-/12-mer linear peptides have shown that the sequence motif A(S)NPS corresponding to 170 ANPS 173 of the large extracellular loop of human CD20 is the most important region for Rituximab recognition and binding (4,23). In particular, Ala 170 and Pro 172 are the most critical ones determined by means of phage display, mutagenesis, and peptide scanning (2,4,19). In our structure of the Rituximab Fab-CD20 epitopepeptide complex, four residues 170 ANPS 173 of the motif are deeply buried in the pocket formed by four CDR loops (L3, H1, H2, and H3). Ala 170 is located in a hydrophobic cavity formed by Trp L47 , Trp L90 , Asn L93 , and Pro L95 of the Fab. The space of the cavity is too narrow to accommodate other residues with a large side chain except serine due to steric conflict (Fig. 2B), providing an explanation for the result that only serine is interchangeable with Ala 170 (4). Similarly, Pro 172 is positioned at the tip of the 3 10 helix and is bound at the bottom of the CDR pocket (Fig. 2B). Modeling study indicates that a serine residue could fit at the same site. However, the relaxation of the rigid conformation of Pro 172 might disrupt the conformational constraint of the 3 10 helix and hence reduce the specificity and binding affinity, which explains the result that when Pro 172 was substituted by serine in the cyclic 7-mer peptide, the mutant peptide could not bind to Rituximab (4).
The importance and requirement of the two other residues, Asn 171 and Ser 173 , of the motif are also supported by our crystal structure showing that each residue makes two of total eight hydrogen bonds as well as extensive van der Waals contacts with the Rituximab Fab. In the structure model, six residues (Glu 168 , Pro 169 , Asn 171 , Pro 172 , Ser 173 , and Lys 175 ) of the epitope peptide make hydrogen-bonding interactions, and additionally Ala 170 , Glu 174 , Asn 176 , and Ser 179 make van der Waals contacts with the Rituximab Fab. Although only 170 ANPS 173 of CD20 were documented to be involved in the Rituximab binding, the other residues might also play some roles in the recognition and binding of CD20 by Rituximab.
Based on phage display results, it was suggested that fragment 182 YCYSI 186 at the C terminus of the large extracellular loop of CD20 is also involved in Rituximab binding (23). In our structure, this region forms part of the C-terminal ␣-helix and has no direct interaction with the Fab. However, residues Cys P183 and Cys P167 of the epitope peptide form a disulfide bond that makes the peptide adopt a unique cyclic conformation. In search for the epitope of CD20 with phage display peptide libraries, a series of cyclic and linear peptides were obtained; however, only the cyclic peptides match the sequence of human CD20 (4). Biochemical data have shown that disruption of the disulfide bond on the large extracellular loop of FIGURE 1. Overall structure of the Rituximab Fab-CD20 epitope-peptide complex. A, overall structure of the complex. The Rituximab Fab is colored with the light chain in yellow and the heavy chain in green, and the CD20 epitope peptide in cyan. B, a stereoview of a composite-omit electron density map at 2.6-Å resolution for the bound epitope peptide contoured at 1.0-level. The atomic coordinates of the peptide residues are shown in ball and stick models. C, structure of the bound epitope peptide. The epitope peptide consists of a short N-terminal coil (residues 167-171), a 3 10 helix (residues 172-174), a small loop (residues 175-177), and a short C-terminal ␣-helix (residues 178 -184). The intra-peptide hydrogen-bonding interactions between residues of the middle part (the 3 10 helix and the small loop) and the other parts of the peptide are indicated with dashed lines.  CD20 completely ablates the binding of CD20 with Rituximab (33). It is very likely that the involvement of the C-terminal 182 YCYSI 186 fragment of the large extracellular loop in the recognition and binding of Rituximab is through the formation of the disulfide bond and the constraint and stabilization of the cyclic conformation of the epitope rather than direct interaction with the antibody.
Analysis of the crystal structure of the Rituximab Fab in complex with its epitope peptide also provides valuable information for modification of the antibody to improve the specificity and binding affinity. Changes of residues on the CDR loops of the Rituximab Fab that could generate more favorable interactions with residues of the epitope of CD20 might increase its binding affinity and specificity with CD20. For instance, substitution of Ser H59 with a polar residue having a slightly longer side chain could introduce favorable hydrophilic interactions with the side chains of Glu P168 and/or Lys P175 . Change of Asp H57 to Glu could generate more hydrophilic interactions with the side chain of Lys P175 and/or Asn P176 . Mutation of Asn H55 to Gln might form potential hydrogen bonds with the side chain of Asn P176 . Similarly, substitution of Tyr H102 with a basic residue such as Lys might form a salt bridge with the side chain of Glu P174 and/or hydrophilic interaction with the side chain of Ser P179 . Substitution of Trp H106 with a basic residue like Lys might form new hydrophilic interactions with the side chain of Asn P171 and Glu P174 .
In summary, we report here the crystal structure of the Fab fragment of therapeutic antibody Rituximab in complex with its epitope peptide of human CD20. Structural analysis reveals the molecular basis of the specific recognition and binding of CD20 by Rituximab. Specifically, the most important epitope region 170 ANPS 173 on the large extracellular loop of CD20 is bound at a pocket formed by four CDR loops of the Rituximab Fab and recognized by residues of the CDR loops through a network of hydrogen-bonding interactions and extensive van der Waals contacts. The unique cyclic conformation of the epitope peptide, which is attributed to the formation of a disulfide bond between Cys P167 and Cys P183 and the presence of a rigid Pro P172 , forms the basis of the specificity of Rituximab. Our structural results also provide useful hints for the development of new therapeutic antibodies with higher binding affinity and better specificity for the treatment of non-Hodgkin lymphoma. a Hydrogen bond mediated by water molecule. The number before the slash is the distance between the Fab atom and the water molecule, and the number after the slash is the distance between the water molecule and the peptide atom.