Identification of Their Epitope Reveals the Structural Basis for the Mechanism of Action of the Immunosuppressive Antibodies Basiliximab and Daclizumab

Interleukin-2 (IL-2) and its receptor (IL-2R) play a pivotal role as signaling molecules in cellular immune responses. Therefore, this ligand-receptor interaction has been considered a very good target in the clinical treatment of acute allograft rejection, graft-versus-host disease, T-cell–mediated autoimmune disease, and certain hematologic malignancies (1–5).

Binding of IL-2 to the IL-2R expressed on T cells elicits cellular changes primarily related to proliferation and differentiation. The IL-2R is composed of the α, β, and γ_c subunits. On activation, they assemble to form the intermediate-affinity IL-2R (βγ_c) or the high-affinity IL-2R (αβγ_c; ref. 6), in which the IL-2 interaction with the α chain initiates the sequential recruitment of the β and γ_c subunits (7). T cells express the α chain (CD25) only on activation. Therefore, blocking of this receptor efficiently and specifically interferes with the cellular immune response, whereas other components of the immune system remain unaffected. An anti-CD25 antibody [initially referred to as “anti-Tac” (8)] has been modified to facilitate its application in patients for activated T-cell suppression. At present, two such anti-Tac derivatives are used clinically: basiliximab, a chimeric antibody containing human constant regions and murine variable regions, and daclizumab, a humanized antibody whose only remaining murine parts are the CDR regions. These therapeutic antibodies are used for IL-2 signaling suppression to prevent (5) or treat (1) acute allograft rejections, acute graft-versus-host disease (2, 3), autoimmune disorders (9), and malignancies, such as T-cell leukemias (4), with high response rates and very few side effects.

The mechanism of action of these antibodies is mediated primarily by inhibition of IL-2 binding to CD25 (10). Thus, downstream events mediating IL-2/IL-2R signaling are impeded. In addition, these antibodies may also act through antibody-directed cellular cytolysis and complement-mediated cytotoxicity in T cells (11).

The epitope within the CD25 molecule recognized by anti-Tac and its derivatives basiliximab and daclizumab has been unknown thus far. Phage display library screenings are a valuable tool for the exploration of epitopes recognized by antibodies (12–14). Phage displaying a peptide that mimics the epitope of a given antibody can be selectively enriched through panning on antibody-coated matrices (15, 16). In this study, we used such phage peptide library screenings to analyze the unknown epitope of the monoclonal antibody basiliximab. A striking peptide sequence motif was recovered, which is homologous to a peptide string within CD25. We substantiate the hypothesis that this is the epitope recognized by basiliximab and show that this epitope overlaps with the previously characterized interaction site of CD25 and IL-2, revealing the structural basis for the inhibition of IL-2 binding to its receptor by this group of immunosuppressive antibodies.

Tissue culture. 293T cells were used with kind permission of David Baltimore (California Institute of Technology, Pasadena, CA) and maintained in DMEM containing 10% fetal bovine serum and 1% penicillin and streptomycin.

Phage display library screening. The fUSE5 vector and K91Kan bacteria were from George Smith (University of Missouri, Columbia, MO). fUSE5-based phage peptide libraries displaying seven random residues flanked by two cysteines (CX₇C) were made as described (12). The degenerate oligonucleotide was TGT(NNB)₇TGT. Second strand was synthesized using sequenase (Amersham) and the primer 5′-TTCGGCCCCAGCGGCCCC-3′. The fUSE5 plasmid was digested with SfiI, and the insert was digested with BglI. Plasmid and insert were ligated and transformed into MC1061. Phage was purified from the bacterial culture using polyethylene glycol-NaCl precipitation. The library was screened on basiliximab (Novartis, Basel, Switzerland) basically as described previously (13). Briefly, polystyrene 96-well plates were coated with 0.5 mg/mL basiliximab (Roche, Basel, Switzerland) in PBS. The wells were blocked with 3% bovine serum albumin and incubated with 10¹⁰ transducing units of the library for 1 h. In the third and fourth round of selection, the library was precleared on rituximab. Bound phage was eluted by adding a K91Kan culture and amplified by overnight growth. Randomly selected clones from the fourth panning round were sequenced (GATC, Konstanz, Germany).

Single clone binding assays. Individual phage clones were tested for binding to basiliximab, daclizumab (Roche), or control proteins as indicated following the protocol of the initial screening. The number of bound phage was determined by growing the recovered phage-bacteria mixture on Luria-Bertani agar containing 40 μg/mL tetracycline.

Glutathione S-transferase fusion proteins. The oligonucleotide encoding the phage-derived peptide was synthesized, digested with BamHI and EcoRI, and ligated into pGEX-2TK (Amersham). Proteins were expressed in BL21 and purified following the manufacturer's instructions.

Mutation of CD25. The pReceiverM02_CD25 plasmid encoding human CD25 was obtained from RZPD (Heidelberg, Germany). CD25 variants with point mutations were generated using the QuikChange Site-Directed Mutagenesis kit (Stratagene) as described (17). Five mutants were established as indicated (Fig. 3B) using suitable primers (Supplementary Table S1).

Transfection and fluorescence-activated cell sorting analysis. 293T cells were transfected with wild-type or mutant CD25 plasmids as indicated using PolyFect (Qiagen). After 24 h, cells were stained using basiliximab at 10 μg/mL or a polyclonal goat anti-human IL-2Rα antibody (R&D Systems) at 7 μg/mL for 30 min. Secondary antibodies were FITC-labeled rabbit anti-human (Jackson) at 2.3 μg/mL or rabbit anti-goat antibodies (Dianova) at 13 μg/mL. Stained cells were analyzed using FACSCalibur (BD Biosciences).

Phage library biopanning on basiliximab yields specific peptides. To identify the epitope of the anti-Tac monoclonal antibody and its derivatives, we screened a random CX₇C phage display peptide library (C = cysteine, X = any amino acid) on immobilized basiliximab. In the fourth selection round, phage bound 525× stronger to basiliximab than to the rituximab control, indicating that basiliximab-specific clones had been enriched (Fig. 1). Sequencing of the inserts of 18 clones recovered from the basiliximab wells of this selection round revealed their striking similarity. The consensus motif contains two aromatic amino acids separated by a single E or H and flanked either upstream or downstream by the branched side chain amino acids L or V. This consensus motif can be summarized as ^F/_Y-^E/_H-W-^L/_V (Fig. 2A). The amino acid N between the aromatic amino acids occurred numerically as often as H but was still not considered to be part of the motif as it only occurred in one single clone.

The selected phage clones specifically bind basiliximab and daclizumab. All of the selected clones, but not insertless fd-tet control phage, bound specifically and strongly to basiliximab but not to the rituximab control (Fig. 2B). Specificity of the phage clone CWYHYIWEC was confirmed using an independent control antibody (infliximab; Supplementary Fig. S1). To verify that the selected peptides bind basiliximab at the paratope region of the antibody, binding of the CWYHYIWEC phage clone was also evaluated for daclizumab, as both therapeutic antibodies derive from the same parental murine antibody (8). Like basiliximab, the CWYHYIWEC phage bound specifically to daclizumab but not to control antibodies (Fig. 2C). This suggests that the selected peptides structurally mimic the common epitope recognized by daclizumab and basiliximab.

CWYHYIWEC phage binding to basiliximab can be competitively blocked by the cognate peptide. To confirm that the peptides displayed on the enriched phage bind the antibody independently of other phage components, competition experiments were done using recombinant peptides. Therefore, the peptide CWYHYIWEC displayed on the phage clone that had shown the strongest antibody binding was produced as a glutathione S-transferase (GST) fusion protein and tested for its ability to inhibit phage binding to basiliximab. GST-CWYHYIWEC, but not GST alone, inhibited phage binding to basiliximab in a dose-dependent manner (Fig. 2D; Supplementary Fig. S2). This indicates that the CWYHYIWEC phage mimics the basiliximab epitope by means of the displayed library peptide and that the conformation of this peptide is independent of the phage capsid protein context.

The selected phage displayed peptides are homologous to CD25. The selected consensus motif ^F/_Y-^E/_H-W-^L/_V shows sequence homology to the peptide string Y-H-F-V comprising amino acid positions 119 to 122 within the extracellular domain of CD25 (SwissProt database). In fact, the amino acids flanking the YHFV peptide string within CD25 contain additional homologies to at least some of the selected phage peptides (Fig. 2A). Taking these additional residues into account, we hypothesized that the epitope recognized by basiliximab consists of the strand ERIYHFV (amino acid positions 116–122) within CD25. In congruence with this, the strongest basiliximab-binding phage CWYHYIWEC showed a remarkably high degree of sequence similarity to CD25, especially when read from COOH terminus to NH₂ terminus (Fig. 2A). There are possible explanations for the reversed order of the amino acids in the CWYHYIWEC clone compared with the original sequence of the presumed epitope. Importantly, the peptide sequence theoretically having optimal binding characteristics may not have been present in the phage library in the forward orientation at all. Only ∼1‰ to 1% of all conceivable amino acid combinations are represented in a CX₇C state of the art-made phage library. If the N-to-C-direction of the sequence is not crucial for the binding characteristics of the peptide, the CWYHYIWEC sequence may have been enriched, as it possessed closest sterical resemblance to the original peptide string within CD25. Of note, 50% of all enriched peptide sequences could be read in both forward and reverse orientation without losing their homology to the 116 to 122 segment of CD25 (Fig. 2A), supporting this assumption. This suggests that the orientation of the peptide backbone does not fundamentally affect the binding characteristics of the peptides enriched in our screenings on basiliximab.

Interestingly, the selected peptides contained a considerable number of aromatic amino acids, and in defined positions, the amino acids tyrosine (Y), tryptophane (W), and phenylalanine (F) seemed to be interchangeable. Such predominance of aromatic amino acids in the selected peptide motifs has been reported in previous epitope mapping studies using phage display (13, 18, 19). In fact, aromatic amino acids and hydroxylated amino acids, such as serine, are particularly suited for molecular recognition patterns in protein-protein interactions (20, 21). For antibody-antigen interactions, tyrosine seems to be especially important (22). Although this phenomenon has been described for paratopes of antibodies, it is remarkable that a similar pattern seems to apply for epitopes as well, at least when they are optimized by random library biopanning. Peptides with additional aromatic residues not present in the natural epitope may have been enriched in the panning because these residues enhance affinity. The replacement of functionally important amino acids by slightly different equivalents may thus improve binding of a peptide, as it may lead to a better structural mimicry of the epitope. A similar finding has recently been described for paratopes of synthetic antibodies by Fellouse et al. (22). This suggests that for epitope mapping approaches using phage display functional classes of amino acids rather than single specific residues have to be considered.

Phage displaying the natural CD25 sequence specifically bind basiliximab. Although all peptides of the enriched phage clones showed a certain degree of sequence similarity to the CD25 segment comprising amino acids 116 to 122, none of the selected peptides had sequence identity to this presumed epitope region. To further explore if the presumed CD25 epitope binds to basiliximab, we designed phage expressing the natural CD25 sequence, flanked by cysteines like in the library-derived peptides. This CERIYHFVC phage (termed “CD25 phage”), but not fd-tet control phage, specifically bound to basiliximab (Fig. 3A), although binding was weaker compared with the phage selected from the random library. This may be due to improved structural epitope mimicry by amino acid exchange as discussed above.

Mutation of the presumed epitope within the extracellular domain of CD25 abrogates binding of basiliximab. To prove that the ERIYHFV peptide string within CD25 is the epitope recognized by basiliximab, we tested basiliximab binding to cells expressing CD25 with mutations in the suspected epitope region or an unrelated downstream region, respectively. We transfected 293T cells, which are CD25 negative, with the plasmid pReceiverM02_CD25 encoding CD25 or the mutants thereof (Fig. 3B). CD25 expression and integration into the membrane was verified by staining with a polyclonal CD25 antibody. Twenty-four hours after transfection, >70% of the transfected cells were positive for CD25 as assessed by fluorescence-activated cell sorting (FACS) analysis (Fig. 3C,, white columns). Binding of basiliximab to cells expressing wild-type CD25 was approximately the same as to cells expressing CD25 with control mutations surrounding amino acid position 210 (mutants M4 and M5). Likewise, exchange of the amino acid in position 119 (Y→D; mutant M1) had no negative effect on basiliximab binding. In contrast, exchange of two or more amino acids in this region (mutants M2 and M3) totally abolished basiliximab binding to the CD25-transfected cells (Fig. 3C , black columns). This confirms that the peptide string surrounding amino acid position 120 within the extracellular domain of CD25 is critical for the binding of basiliximab.

The binding sites of IL-2 and basiliximab in CD25 overlap. As IL-2 has been previously crystalized in complex with the IL-2R (7), it is possible to compare the binding site of IL-2 to IL-2Rα (CD25) with the localization of the basiliximab epitope identified in our study. Figure 4 illustrates the structure of IL-2 in complex with the three receptor subunits. The IL-2/IL-2Rα interface consists of 20 IL-2 residues and 21 IL-2Rα residues (Fig. 4, yellow) on different loops of the molecule. The basiliximab epitope characterized here comprises the seven amino acids (116) E-R-I-Y-H-F-V (122) (Fig. 4, red and orange), which include three of the IL-2 contact residues (118) I-Y-H (120) (Fig. 4, orange; ref. 7). We therefore conclude that this overlap in contact sites is the structural basis for the basiliximab-mediated inhibition of IL-2 binding to IL-2Rα. As a consequence of this inhibition, the assembly of the IL-2R and thus downstream signaling events leading to proliferation and differentiation are prevented in vivo.

Taken together, we identified the epitope targeted by the immunosuppressive antibody basiliximab within the extracellular domain of CD25 and showed that the structural basis for inhibition of IL-2 binding to IL-2Rα is given by a topic overlap of binding sites of IL-2 and basiliximab.

Grant support: Deutsche José Carreras Leukämie-Stiftung grant R03/08 (M. Trepel).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Florian Otto for helpful discussions, Kerstin Schmidt for technical support, and George Smith for generously supplying the fUSE5 plasmid and K91Kan bacteria.