HHS Public Access Author manuscript Author Manuscript
J Mol Biol. Author manuscript; available in PMC 2017 January 16. Published in final edited form as: J Mol Biol. 2016 January 16; 428(1): 194–205. doi:10.1016/j.jmb.2015.12.002.
Structure of a TCR mimic antibody with target predicts pharmacogenetics Niloufar Ataie1, Jingyi Xiang2, Neal Cheng2, Elliott J. Brea3, Wenjie Lu1, David A. Scheinberg3,4, Cheng Liu2, and Ho Leung Ng1,5 1University
of Hawaii at Manoa, Department of Chemistry. 2545 McCarthy Mall. Honolulu, HI 96822-2275. USA
Author Manuscript
2Eureka
Therapeutics Inc. 5858 Horton St, Emeryville, CA 94608. USA
3Sloan-Kettering 4Weill
Institute. 1275 York Ave. New York, NY 10065. USA
Cornell Medical College. 1305 York Ave. New York, NY 10021. USA
5University
of Hawaii Cancer Center. 2545 McCarthy Mall. Honolulu, HI. 96822-2275. USA
Abstract
Author Manuscript
Antibody therapies currently target only extracellular antigens. A strategy to recognize intracellular antigens is to target peptides presented by immune HLA receptors. ESK1 is a human, T-cell receptor (TCR)-mimic antibody that binds with sub-nanomolar affinity to the RMF peptide from the intracellular oncoprotein Wilms Tumor 1 (WT1) in complex with HLA-A*02:01. ESK1 is therapeutically effective in mouse models of WT1+ human cancers. TCR-based therapies have been presumed to be restricted to one HLA subtype. The mechanism for the specificity and high affinity of ESK1 is unknown. We show in a crystal structure that ESK1 Fab binds to RMF/HLAA*02:01 in a different mode than TCRs. From the structure, we predict and then experimentally confirm high affinity binding with multiple other HLA-A*02 subtypes, broadening the potential patient pool for ESK1 therapy. Using the crystal structure, we also predict potential off-target binding that we experimentally confirm. Our results demonstrate how protein structure information can contribute to personalized immunotherapy.
Graphical abstract
Author Manuscript
Correspondence: [email protected], phone +1 (808) 956-2014. Conflicts of interest: J. Xiang, N. Cheng, and C. Liu have ownership interest (including patents) in Eureka Therapeutics Inc. D.A. Scheinberg is an inventor of intellectual property related to ESK1 that is owned by Sloan Kettering and licensed to Novartis. No potential conflicts of interest were disclosed by the other authors. Accession numbers Protein Data Bank: Coordinates and structure factors have been deposited with accession code 4WUU. Supplementary data to this article can be found online at http://dx.doi.org/xxxxxx. Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Ataie et al.
Page 2
Author Manuscript Keywords antibodies; MHC; cancer; immunotherapy
Author Manuscript
Introduction Antibody targeted therapies have been highly successful in selectively killing cancer cells with reduced adverse effects. Two dozen cancer therapeutic antibodies are approved for clinical use. Cancer antibody therapies target extracellular proteins, and a major unmet challenge has been targeting intracellular tumor specific, mutated oncogenic proteins, which are not generally present on cell surfaces [1].
Author Manuscript
One strategy to target intracellular antigens is to direct TCR based therapies, either with adoptive T-cell therapy or TCR mimic antibodies, to oncoprotein derived peptides that are displayed on the cell surface by major histocompatibility complex (MHC) class I receptors [2–4]. The Wilms Tumor 1 protein (WT1) is overexpressed in most hematological and solid tumor cancers [5,6], and peptides derived from WT1 are displayed by HLA receptors as 9–10 amino acid T-cell epitopes for presentation to the TCR [7,8]. WT1 is the highest ranked target by the National Cancer Institute pilot project to prioritize cancer immunotherapy antigens for clinical trials based on criteria including the number of patients expressing antigen, specificity of antigen to cancer cells, expression levels of protein, oncogenicity, and immunogenicity [9]. While targeting tumor antigens on MHC-I is a viable therapeutic strategy, investigation into the structure of TCR based therapies has looked primarily at ontarget reactivity, and the nature of the TCR and peptide MHC interaction may allow for offtarget cross-reactivity as well. The ability to predict potential off targets may avoid toxicities of TCR and TCR mimic based therapies and warrants a more systematic evaluation [10].
Author Manuscript
ESK1 is a human, TCR mimic (TCRm) monoclonal antibody (mAb) that we previously engineered to bind the WT1 derived peptide epitope RMFPNAPYL (RMF)/HLA-A*02:01 complex with 0.2 nM affinity and to mediate WT1-restricted cancer cell death in xenograft mouse models of human cancers by antibody-dependent cell-mediated cytotoxicity (ADCC) [11–14]. Toxicity studies in transgenic HLA-A*02:01 expressing mice demonstrate lack of uptake or toxicity in normal tissues [13]. The effectiveness and low toxicity of ESK1 have led to its further development as a drug. There is a need for high resolution characterization
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 3
Author Manuscript
of the binding site and the complementary determining regions (CDRs) that contribute to ESK1 binding to address questions about specificity, possible cross-reactivities and offtargets, self-reactivity, and compatibility with other HLA-A*02 subtypes. While HLAA*02:01 is the most common HLA-A*02 subtype in the United States and Europe, other subtypes bearing different peptide and TCR binding specificities are found across different ethnic groups worldwide [15,16].
Author Manuscript
We solved the crystal structure of the RMF/HLA-A*02:01/ESK1 (Fab) complex to 3.05 Å. The ESK1 variable domains bind the HLA and peptide in a different mode than TCRs and other TCR mimic Fabs [17,18]. ESK1 CDR loops contact regions of HLA that TCRs typically do not reach. Our structure and binding studies also show that RMF acts as an electrostatic key in mediating ESK1 specificity and activity with Arg1 playing a central role. RMF Pro4 also contributes to ESK1 binding. The rest of the binding interface is limited to interactions between ESK1 and a region on the HLA receptor that is conserved between subtypes, suggesting that binding is compatible with other common HLA-A*02 subtypes, which we experimentally confirmed. This potentially broadens the target patient populations for this drug beyond the HLA-A*02:01 subtype found predominantly in Caucasians to multiple other ethnic groups [19]. The crystal structure allowed additional predictions of possible cross reactivity with several human self-peptides, some of which we experimentally confirmed in vitro. This is the first evaluation of cross reactivity of a TCR based therapy in silico, allowing for structural and bioinformatics data to create a pipeline for better predicting specificity. In this way, we report that structural data can be a valuable tool for pre-clinical characterization of antibody pharmacogenetics and toxicology in genetically diverse patient populations.
Author Manuscript
Results and Discussion Overall structure The ESK1 Fab fragment binds the peptide-MHC (pMHC) with the variable domain (Fig. 1a) contacting 160 Å2 of the first five residues of RMF (Fig. 1b). RMF interacts with 873 Å2 of the HLA receptor in a manner closely superimposable with the crystal structure of the HLAA*02:01/RMF complex without bound antibody or TCR with an RMSD of 0.75 Å (PDB 3HPJ) [20]. In total, the ESK1 antibody-HLA surface is 890 Å2 and the total ESK1-pMHC surface is 1050 Å2 (Fig. 1b, c). Electron density quality at the binding interface is excellent for diffraction data of this resolution and unambiguously resolves the side chains (Fig. S1). The structure is well refined with Rfree = 0.25, which is in the upper 50% for crystal structures at this resolution according to the PDB validation report (Table 1).
Author Manuscript
Binding interactions with the CDR loops Figures 1d–e show the interactions of the CDR loops with the pMHC. Four tyrosines on CDR-H2 and CDR-H3 (residues 55, 101, 104, 105), stabilize the interface. Hydrogen bonds with ideal geometry and van der Waals contacts are common as are prominent networks of cation-π interactions (Fig. 2). Cation-π interactions are common in protein interfaces and are among the strongest non-covalent forces within proteins, contributing energies up to 4 kcal/ mol, comparable to hydrogen bonds [21,22]. They are a primary driver of long range
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 4
Author Manuscript
interactions, contributing favorable forces up to 6 Å away and at angles ranging from the favored θ = 0, where the cation is directly above the π ring system, to beyond 60° [23]. CDR-H2 and CDR-H3 make extensive cation-π interactions with Arg108, Arg169, and Arg170 of the HLA. The extensive chains of cation-π interactions in the ESK1-pMHC are consistent with those frequently seen in other antibody-antigen complexes [24]. Comparison to TCRs and other TCR-like antibodies
Author Manuscript
We compared the structure of ESK1 + HLA-A*02 to HLA-A*02 with twenty other TCRs and TCR mimic antibodies [17,18,25–30] to determine the molecular basis for the 1000-fold higher affinity for ESK1 compared to TCRs (Fig. 3). The four TCR mimic antibodies compared differ widely in binding orientations. None of the other three antibodies binds to the same arginine cluster as ESK1. ESK1 CDR loops reach HLA-A*02 regions untouched by TCRs that interact with the HLA arginine cluster. Compared to the sub-nanomolar Kd of ESK1, TCRs typically bind with a maximum Kd of 1 μM [31]. Table 2 compares the contact surface area of three TCR/pHLA structures with that of ESK1/pHLA. ESK1 makes more contacts with the HLA than the peptide compared to three other TCR pMHCs. ESK1 contacts only 160 Å2 of the peptide, which is 15% of the total pMHC contact surface. In contrast, in the other three pMHCs, the peptide contribution is 30–34% of the overall contact surface. While ESK1 does not have the largest contact surface compared to the other pMHCs, it has the highest affinity [32,33]. Peptide binding specificity
Author Manuscript
ESK1 binds the HLA receptor centered on RMF Arg1 (Fig. 4). ESK1 primarily interacts with RMF Arg1 and Pro4. RMF Met2, Phe3, Ala6, Phe7, and Leu9 side chains contact the bottom of the HLA peptide binding region. Asn5 and Tyr8 face the solvent and do not contact ESK1.
Author Manuscript
In order to probe the structural requirements of the ESK1 antibody to the HLA-A*02/RMF complex, we made substitutions along the length of the RMF peptide. ESK1 binding to the peptide/MHC complex was measured in the TAP1/2 deficient T2 lymphoblastoid cell line by pulsing various peptides onto HLA-A*02 (supplementary Fig. S2). Substitution at position 1 with citrulline, sterically similar to arginine, but uncharged, eliminated binding. Lysine substitution at position 1 decreased binding 40-fold. Substitution at position 1 with ornithine, similar to lysine, but shorter by a carbon atom, or with histidine, eliminated ESK1 binding. Overall, the binding data support a highly specific fit of the ESK1 electrostatic lock for arginine. Arg1 is near several polar residues (Fig. 4, supplementary Fig. S2) in an otherwise buried hydrophobic pocket capped by a highly negatively charged loop from the ESK1 heavy chain. This pocket (supplementary Fig. S2e) acts as an electrostatic lock in which the negatively charged dipoles of the backbone carbonyl oxygens of Ser102, Gly103, and the sidechain of Asp106 converge without countercharge, creating a buried space of focused negative charge. Tyrosine at position 1 does not provide the countercharge needed to fit the electrostatic lock. The crystal structure can help reveal the factors driving specificity and cross reactivity of ESK1 toward other peptide/MHC epitopes.
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 5
Author Manuscript
CDR-L3 partially shields RMF Pro4 from solvent (Fig. 4). Asp94 of the light chain is positioned 3.4 Å away from Pro4. Substitution to glycine, alanine, or serine at RMF position 4 does not significantly affect binding, but substitution to cysteine decreases binding tenfold (Fig. 4c). All low energy rotamers of cysteine modeled into position 4 of RMF sterically clashed with Asp94. Surprisingly, the substitution of tryptophan at position 4 showed no loss in binding affinity. Tryptophan modeled into that position forms potentially favorable cation-π interactions with HLA Lys66. Large residues modeled into the fifth position clash with nearby residues. Other large bulky residues at position five decrease binding over tenfold (Fig. 4c).
Author Manuscript
The binding mode of ESK1 differs most from that of the Fab Hyb3 bound to HLA-A*01 receptor with the MAGE-A1 peptide [34]. While ESK1 primarily contacts the N-terminal of WT1, Hyb3 primarily contacts the C-terminal of MAGE-A1 (Fig. 5). The structures of the HLA-receptors are highly conserved between the two structures, with an RMSD of 1.1 Å over all Cα. ESK1 and Hyb3 differ in binding orientation by 58° and represent the diverse binding modes to HLA-peptide complexes available to antibodies, as opposed to the canonical diagonal binding conserved among T-cell receptors. Like ESK1, Hyb3 demonstrates surprisingly high binding specificity in vitro and in vivo despite making close contacts with less than half the peptide, and is capable of differentiating between MAGE-A1 and MAGE-A3 peptides which differ by only three amino acids [34,35]. This is consistent with a model in which small peptide differences affect HLA receptor conformation and thereby, Fab binding [35]. HLA-A*02 subtype specificity
Author Manuscript
An important potential therapeutic advantage of TCRm antibodies over engineered T-cells is that antibody therapy does not require customization on a per patient basis. However, use of a TCRm antibody is still restricted by patient HLA type. Producing TCRm antibodies that are effective across a broad range of HLA types would greatly increase their therapeutic value. We demonstrate that structural information can be used to predict that ESK1 is effective across many of the most common HLA-A*02 subtypes.
Author Manuscript
HLA-A*02 alleles vary widely across different ethnic populations. HLA-A*02:01 is the most commonly found variant, especially in Caucasians in the United States, but other subtypes are more common in other ethnic groups. Six subtypes are found in hundreds of millions of people worldwide that differ from HLA-A*02:01 through 1 to 4 amino acid changes in the side walls of the peptide binding pocket. Models were built of several common HLA-A*02 subtypes, using crystal structure coordinates of the subtype if available (A*02:03, A*02:06, A*02:07), to predict the effect of binding to RMF and ESK1. All residues that differ between A*02:01 and the other six subtypes are not within 4 Å of ESK1 (Fig. 6). Modeling predicts that RMF binding would be retained across these A*02 subtypes due to the lack of interfering steric or electronic interactions. We predicted and experimentally confirmed that these six subtypes bind RMF and ESK1 (Table 3, supplementary Fig. 3) by surface plasmon resonance using immobilized, biotinylated HLA +RMF. The measured Kd of ESK1 to RMF + A*02:01 is 13.2 nM, higher than measured by
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 6
Author Manuscript
the whole cell method. All six HLA variants bind to ESK1 with similar or lower Kd (1.1 nM for A*02:11 to 23.6 nM for A*02:02), suggesting patient compatibility. Structure based prediction of potential ESK1 off-target toxicity A major advantage of antibody over small molecule therapeutics is that toxicity tends to be more predictable and related to mechanism of action due to their higher specificity [36]. Nevertheless, off-target binding is still a serious concern for immunotherapy. Currently, specificity is enforced by counterscreening antibody candidates against non-specific targets such as cell lysate or MHC-receptors bearing untargeted antigen peptides. We demonstrate for the first time that structural knowledge and simple computational analysis can be used to predict off targets and further engineer TCRm specificity.
Author Manuscript Author Manuscript
Given the structural and biochemical data supporting the key binding interactions of the ESK1 antibody to RMF positions Arg1, Phe3, and Pro4, we performed a BLAST search for similar expressed human peptide sequences. The peptides were then subjected to binding affinity analysis using NETMHC [37] to predict peptides with a Kd to HLA-A*02:01 < 500 nM. Using artificial neural networks, NETMHC predicts MHC-peptide binding with 78% accuracy. The two highest ranking peptides were identified to be from MED13L (mediator complex subunit 13-like), RMFPTPPSL, and PIGQ (phosphatidylinositol glycan anchor biosynthesis, class Q), RMFPGEVAL, and tested as potential off targets to validate our hypothesis. ESK1 bound cell surface MHC with either peptide (Fig. 7). Structural modeling supports compatibility of these two peptides with HLA-A*02:01 and ESK1 binding. In silico mutagenesis of the RMF peptide in the ESK1-HLA-RMF crystal structure showed that the sequences in the MED13L and PIGQ peptides could be accommodated with little change to HLA and ESK1 binding surfaces (supplemementary Fig. S4). Studies on normal donor peripheral blood monocyte cells showed potential off target binding of ESK1 to a small subset of healthy donor CD19+ B-cells [11], which were WT1 negative. In this way, it is possible that ESK1 binding to this subset of cells may be attributable to non-WT1 derived epitopes, though we cannot determine the peptide from these studies. Therefore, utilizing structural knowledge of TCRm antibody-MHC complexes provides a novel basis for determining possible off-target self-proteins in advance of clinical trials and could serve as a platform for future development of specific TCRm antibodies to minimize off target effects.
Conclusions Author Manuscript
TCRs with moderate affinity (Kd = 1–100 μM) to antigens are produced in the body and provide the basis for a number of therapeutic agents, including soluble TCR-based constructs, TCR engineered cells, and peptide vaccines. TCR-based drugs have several potentially serious limitations. One is their restriction to a specific HLA subtype, reducing the potential patient population that may benefit from the therapeutic approach to a small fraction of afflicted patients. An additional limitation is cross-reactivity of typically low affinity TCRs to other peptide epitopes presented by MHC molecules that has resulted in severe toxicity in human trials [10]. The current state of the art of antibody engineering is more advanced than with TCRs, allowing for greater structural and binding diversity. Here
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 7
Author Manuscript
we present the first use of structural and biochemical characterization of the interaction of a TCRm antibody to its cognate peptide/MHC epitope in an effort to investigate and overcome both these limitations. An essential feature in the development of a TCRm is high specificity to the molecular complex of both the MHC molecule and presented peptide simultaneously, as mAb reactivity with the MHC alone would confer reactivity with every nucleated cell. Phage display can produce much higher affinity TCR mimic antibodies such as ESK1 than occurs in vivo with TCRs, while simultaneously selecting against antibodies that may cross-react with the MHC complex alone. Second, the high affinity of TCRm antibodies allows pharmacologic dosing and retention of the drug at the cancer cell site in vivo. The TCRm, ESK1, binds with sub-nanomolar affinity, over 1000-fold more tightly than natural TCRs. Our data demonstrate the molecular basis for both the specificity and affinity of ESK1.
Author Manuscript
Surprisingly, the high affinity binding was mainly driven by interactions between the antibody and HLA receptor that are restricted to regions identical or similar between at least six common HLA-A*02 subtypes that vary in different ethnic groups worldwide. This result led us to hypothesize that ESK1 could bind to the same WT1 peptide epitope when presented by various other HLA-A*02 subtypes in addition to HLA-A*02:01. Experimental confirmation of ESK1 binding to these HLA-A*02 variants extends the potential scope of ESK1 therapy to millions of additional patients. In addition to its pharmacologic advantages, this unexpected finding shows that the use of a TCR mimic therapy may be a novel strategy to overcome the limits of HLA subtype restriction of typical TCR-based therapy and therefore have broader therapeutic applications.
Author Manuscript Author Manuscript
Naturally occurring TCR can bind off-target peptides [38,39]. We were able to use the structural information to predict potential cross-reactive peptides of ESK1 in humans. When bound to the HLA receptor, ESK1 primarily interacts with the N-terminal of RMF. As a consequence, other peptides expressed in humans that are homologous to the N terminal of RMF, and bind to HLA-A*02 with the anchor residue at position 9, but that differ at positions 5–8 in the C terminal, might be ESK1 off-targets. We confirmed that such peptides in complex with HLA-A*02:01 can bind to ESK1 on peptide-pulsed T2 cells in vitro. Thus, it may be possible to account for both structure and potential off targets using a bioinformatics approach as we have shown with ESK1 mAb with WT1-HLA-A*02:01. The determination of whether these alternate targets are processed and expressed on the surface in normal tissues would be challenging, but identification of these targets in advance of clinical trials may allow for appropriate prospective clinical observation. In conclusion, we demonstrate for the first time that detailed, integrated structural, genetic, and biochemical analysis can provide a platform for appropriate selection and stratification of patient in efforts to maximize treatment efficacy and minimize toxicity. As such, this structure-based approach may serve as a key step towards personalized and precision medicine.
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 8
Author Manuscript
Materials and methods ESK1 design, construction, purification The antibody ESK1 was prepared as described previously [11]. Briefly, ESK1 was derived from the top clone selected from the Eureka Therapeutics human single-chain variable fragment (scFv) phage library with specific binding for the RMF/HLA-A*02:01 complex. The full-length human IgG1 form of ESK1 antibody was produced in the Chinese Hamster Ovary (CHO) cell line. ESK1 antibody was affinity purified by Protein A chromatography with an AKTA FPLC system. ESK1 flow cytometry binding assay to RMF peptides
Author Manuscript
T2 cells, a TAP1/2 deficient HLA-A*02 expressing cell line that does not present endogenous peptides, were used for pulsing peptides. T2 cells were purchased from ATCC and cultured in RPMI with 10% FBS. Peptides were purchased from Genemed Synthesis (San Antonio, TX USA) with >90% purity. Peptides were resuspended to 20 mg/ml in DMSO and diluted to working stocks in PBS. T2 cells, at 1×106 cells/ml, were pulsed with peptides at various concentrations overnight in RPMI with 10% dialyzed FBS and 10 μg/ml β2-microglobulin. T2 cells were incubated with peptide overnight at 37°C with 5% CO2. Cells were then spun down, washed with PBS, and blocked with human FcR blocking solution for 15 minutes on ice. T2 cells were then stained with ESK1 labeled with APC at 3 μg/ml final antibody for 45 minutes on ice. Cells were then spun down and resuspended in FACS buffer and run on FACS Calibur. Off target binding of ESK1
Author Manuscript
Protein BLAST from NCBI, along with analysis using NetMHC 3.0 for peptides with a binding of 2.0 and CC½ showing statistical significance at 0.1% [44]. The structure was solved by molecular replacement using the structures of HLA-A2 (PDB 3HPJ) and the Fab fragment from antibody HC84-1 (PDB 4JZN) using MOLREP [45] and Phaser [46]. The structure was refined with REFMAC and Phenix.refine using TLS parameters for each chain [47,48] with manual rebuilding with Coot [49] and use of Feature Enhanced Maps [50]. Processing and refinement statistics are shown in Table 1. Structure analysis and modeling
Author Manuscript
Modeling was performed with PyMol, Coot, and Yasara [51]. In silico mutagenesis was performed in Yasara, involving selection of the optimal target residue rotamer, followed by energy minimization with the AMBER03 force field [52]. Interaction surface properties were calculated with Pisa [53].
Supplementary Material Refer to Web version on PubMed Central for supplementary material.
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 10
Author Manuscript
Acknowledgments We thank Aina Cohen (SSRL) for assistance with diffraction data collection. This work was supported by University of Hawaii startup funds (HLN), the Leukemia and Lymphoma Society (DAS), NIH R01CA55349 (DAS) and P01 23766 (DAS), NIH NIGMS MSTP GM007739 (EJB), NIH NIGMS P41GM103393 (DAS), P30 CA008748 (DAS), and the Major Family Fund (DAS). Use of the Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory, is supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under Contract No. DE-AC02-76SF00515. The SSRL Structural Molecular Biology Program is supported by the DOE Office of Biological and Environmental Research.
Abbreviations
Author Manuscript
CDR
complementary determining regions
HLA
human leukocyte antigen
mAb
monoclonal antibody
MHC
major histocompatibility complex
pMHC
peptide-MHC complex
TCR
T-cell receptor
TCRm
T-cell receptor mimic
WT1
Wilms Tumor 1
References
Author Manuscript Author Manuscript
1. Scott AM, Wolchok JD, Old LJ. Antibody therapy of cancer. Nat Rev Cancer. 2012; 12:278– 287.10.1038/nrc3236 [PubMed: 22437872] 2. Dahan R, Reiter Y. T-cell-receptor-like antibodies - generation, function and applications. Expert Rev Mol Med. 2012; 14:e6.10.1017/erm.2012.2 [PubMed: 22361332] 3. Weidanz JA, Hawkins O, Verma B, Hildebrand WH. TCR-like biomolecules target peptide/MHC Class I complexes on the surface of infected and cancerous cells. Int Rev Immunol. 2011; 30:328– 340.10.3109/08830185.2011.604880 [PubMed: 22053972] 4. Yee C, Thompson JA, Byrd D, Riddell SR, Roche P, Celis E, et al. Adoptive T cell therapy using antigen-specific CD8+ T cell clones for the treatment of patients with metastatic melanoma: in vivo persistence, migration, and antitumor effect of transferred T cells. Proc Natl Acad Sci U S A. 2002; 99:16168–16173.10.1073/pnas.242600099 [PubMed: 12427970] 5. Keilholz U, Menssen HD, Gaiger A, Menke A, Oji Y, Oka Y, et al. Wilms’ tumour gene 1 (WT1) in human neoplasia. Leukemia. 2005; 19:1318–1323.10.1038/sj.leu.2403817 [PubMed: 15920488] 6. Yang L, Han Y, Suarez Saiz F, Saurez Saiz F, Minden MD. A tumor suppressor and oncogene: the WT1 story. Leukemia. 2007; 21:868–876.10.1038/sj.leu.2404624 [PubMed: 17361230] 7. Ohminami H, Yasukawa M, Fujita S. HLA class I-restricted lysis of leukemia cells by a CD8(+) cytotoxic T-lymphocyte clone specific for WT1 peptide. Blood. 2000; 95:286–293. [PubMed: 10607714] 8. Oka Y, Elisseeva OA, Tsuboi A, Ogawa H, Tamaki H, Li H, et al. Human cytotoxic T-lymphocyte responses specific for peptides of the wild-type Wilms’ tumor gene (WT1) product. Immunogenetics. 2000; 51:99–107. [PubMed: 10663572] 9. Cheever MA, Allison JP, Ferris AS, Finn OJ, Hastings BM, Hecht TT, et al. The Prioritization of Cancer Antigens: A National Cancer Institute Pilot Project for the Acceleration of Translational Research. Clin Cancer Res. 2009; 15:5323–5337.10.1158/1078-0432.CCR-09-0737 [PubMed: 19723653]
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 11
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
10. Linette GP, Stadtmauer EA, Maus MV, Rapoport AP, Levine BL, Emery L, et al. Cardiovascular toxicity and titin cross-reactivity of affinity-enhanced T cells in myeloma and melanoma. Blood. 2013; 122:863–871.10.1182/blood-2013-03-490565 [PubMed: 23770775] 11. Dao T, Yan S, Veomett N, Pankov D, Zhou L, Korontsvit T, et al. Targeting the intracellular WT1 oncogene product with a therapeutic human antibody. Sci Transl Med. 2013; 5:176ra33.10.1126/ scitranslmed.3005661 12. Dubrovsky L, Pankov D, Brea EJ, Dao T, Scott A, Yan S, et al. A TCR-mimic antibody to WT1 bypasses tyrosine kinase inhibitor resistance in human BCR-ABL+ leukemias. Blood. 2014; 123:3296–3304.10.1182/blood-2014-01-549022 [PubMed: 24723681] 13. Veomett N, Dao T, Liu H, Xiang J, Pankov D, Dubrovsky L, et al. Therapeutic Efficacy of an FcEnhanced TCR-like Antibody to the Intracellular WT1 Oncoprotein. Clin Cancer Res. 2014; 20:4036–4046.10.1158/1078-0432.CCR-13-2756 [PubMed: 24850840] 14. Dao T, Pankov Dmitry, Scott Andrew, Korontsvit Tatyana, Zakhaleva Victoriya, Xu Yiyang, et al. Therapeutic targeting of the intracellular oncoprotein WT1 with a Bi-specific T cell engager antibody. Nat Biotechnol. 2015 In press. 15. Ellis JM, Henson V, Slack R, Ng J, Hartzman RJ, Katovich Hurley C. Frequencies of HLA-A2 alleles in five U.S. population groups. Predominance Of A*02011 and identification of HLAA*0231. Hum Immunol. 2000; 61:334–340. [PubMed: 10689125] 16. Gonzalez-Galarza FF, Christmas S, Middleton D, Jones AR. Allele frequency net: a database and online repository for immune gene frequencies in worldwide populations. Nucleic Acids Res. 2011; 39:D913–919.10.1093/nar/gkq1128 [PubMed: 21062830] 17. Mareeva T, Martinez-Hackert E, Sykulev Y. How a T cell receptor-like antibody recognizes major histocompatibility complex-bound peptide. J Biol Chem. 2008; 283:29053–29059.10.1074/ jbc.M804996200 [PubMed: 18703505] 18. Stewart-Jones G, Wadle A, Hombach A, Shenderov E, Held G, Fischer E, et al. Rational development of high-affinity T-cell receptor-like antibodies. Proc Natl Acad Sci U S A. 2009; 106:5784–5788.10.1073/pnas.0901425106 [PubMed: 19307587] 19. Marsh, SGE.; Parham, P.; Barber, LD. The HLA Facts Book. Academic Press; 1999. 20. Borbulevych OY, Do P, Baker BM. Structures of native and affinity-enhanced WT1 epitopes bound to HLA-A*0201: implications for WT1-based cancer therapeutics. Mol Immunol. 2010; 47:2519–2524.10.1016/j.molimm.2010.06.005 [PubMed: 20619457] 21. Burley SK, Petsko GA. Amino-aromatic interactions in proteins. FEBS Lett. 1986; 203:139–143. [PubMed: 3089835] 22. Dougherty DA. The Cation–π Interaction. Acc Chem Res. 2013; 46:885–893.10.1021/ar300265y [PubMed: 23214924] 23. Marshall MS, Steele RP, Thanthiriwatte KS, Sherrill CD. Potential energy curves for cation-pi interactions: off-axis configurations are also attractive. J Phys Chem A. 2009; 113:13628– 13632.10.1021/jp906086x [PubMed: 19886621] 24. Dalkas GA, Teheux F, Kwasigroch JM, Rooman M. Cation–π, amino–π, π–π, and H-bond interactions stabilize antigen–antibody interfaces. Proteins Struct Funct Bioinforma. 2014; 82:1734–1746.10.1002/prot.24527 25. Ding YH, Smith KJ, Garboczi DN, Utz U, Biddison WE, Wiley DC. Two human T cell receptors bind in a similar diagonal mode to the HLA-A2/Tax peptide complex using different TCR amino acids. Immunity. 1998; 8:403–411. [PubMed: 9586631] 26. Ding YH, Baker BM, Garboczi DN, Biddison WE, Wiley DC. Four A6-TCR/peptide/HLA-A2 structures that generate very different T cell signals are nearly identical. Immunity. 1999; 11:45– 56. [PubMed: 10435578] 27. Garboczi DN, Ghosh P, Utz U, Fan QR, Biddison WE, Wiley DC. Structure of the complex between human T-cell receptor, viral peptide and HLA-A2. Nature. 1996; 384:134–141. [PubMed: 8906788] J Immunol Baltim Md. 1950; 185:6394–6401. 2010. 28. Sami M, Rizkallah PJ, Dunn S, Molloy P, Moysey R, Vuidepot A, et al. Crystal structures of high affinity human T-cell receptors bound to peptide major histocompatibility complex reveal native diagonal binding geometry. Protein Eng Des Sel PEDS. 2007; 20:397–403.10.1093/protein/ gzm033 [PubMed: 17644531]
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 12
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
29. Borbulevych OY, Piepenbrink KH, Baker BM. Conformational melding permits a conserved binding geometry in TCR recognition of foreign and self molecular mimics. J Immunol Baltim Md 1950. 2011; 186:2950–2958.10.4049/jimmunol.1003150 30. Borbulevych OY, Piepenbrink KH, Gloor BE, Scott DR, Sommese RF, Cole DK, et al. T cell receptor cross-reactivity directed by antigen-dependent tuning of peptide-MHC molecular flexibility. Immunity. 2009; 31:885–896.10.1016/j.immuni.2009.11.003 [PubMed: 20064447] 31. Armstrong KM, Insaidoo FK, Baker BM. Thermodynamics of T-cell receptor-peptide/MHC interactions: progress and opportunities. J Mol Recognit JMR. 2008; 21:275–287.10.1002/jmr.896 [PubMed: 18496839] 32. Willcox BE, Gao GF, Wyer JR, Ladbury JE, Bell JI, Jakobsen BK, et al. TCR binding to peptideMHC stabilizes a flexible recognition interface. Immunity. 1999; 10:357–365. [PubMed: 10204491] 33. Davis-Harrison RL, Armstrong KM, Baker BM. Two different T cell receptors use different thermodynamic strategies to recognize the same peptide/MHC ligand. J Mol Biol. 2005; 346:533– 550.10.1016/j.jmb.2004.11.063 [PubMed: 15670602] 34. Hülsmeyer M, Chames P, Hillig RC, Stanfield RL, Held G, Coulie PG, et al. A major histocompatibility complex-peptide-restricted antibody and t cell receptor molecules recognize their target by distinct binding modes: crystal structure of human leukocyte antigen (HLA)-A1MAGE-A1 in complex with FAB-HYB3. J Biol Chem. 2005; 280:2972–2980.10.1074/ jbc.M411323200 [PubMed: 15537658] 35. Chames P, Willemsen RA, Rojas G, Dieckmann D, Rem L, Schuler G, et al. TCR-Like Human Antibodies Expressed on Human CTLs Mediate Antibody Affinity-Dependent Cytolytic Activity. J Immunol. 2002; 169:1110–1118.10.4049/jimmunol.169.2.1110 [PubMed: 12097420] 36. Brennan FR, Morton LD, Spindeldreher S, Kiessling A, Allenspach R, Hey A, et al. Safety and immunotoxicity assessment of immunomodulatory monoclonal antibodies. mAbs. 2010; 2:233– 255.10.4161/mabs.2.3.11782 [PubMed: 20421713] 37. Lundegaard C, Lamberth K, Harndahl M, Buus S, Lund O, Nielsen M. NetMHC-3.0: accurate web accessible predictions of human, mouse and monkey MHC class I affinities for peptides of length 8–11. Nucleic Acids Res. 2008; 36:W509–512.10.1093/nar/gkn202 [PubMed: 18463140] 38. Reiser JB, Darnault C, Grégoire C, Mosser T, Mazza G, Kearney A, et al. CDR3 loop flexibility contributes to the degeneracy of TCR recognition. Nat Immunol. 2003; 4:241–247.10.1038/ni891 [PubMed: 12563259] 39. Birnbaum ME, Mendoza JL, Sethi DK, Dong S, Glanville J, Dobbins J, et al. Deconstructing the peptide-MHC specificity of T cell recognition. Cell. 2014; 157:1073–1087.10.1016/j.cell. 2014.03.047 [PubMed: 24855945] 40. Altman, JD.; Davis, MM. Curr Protoc Immunol. John Wiley & Sons, Inc; 2001. MHC-Peptide Tetramers to Visualize Antigen-Specific T Cells. http://onlinelibrary.wiley.com/doi/ 10.1002/0471142735.im1703s53/abstract [accessed November 22, 2014] 41. Soltis SM, Cohen AE, Deacon A, Eriksson T, González A, McPhillips S, et al. New paradigm for macromolecular crystallography experiments at SSRL: automated crystal screening and remote data collection. Acta Crystallogr D Biol Crystallogr. 2008; 64:1210–1221.10.1107/ S0907444908030564 [PubMed: 19018097] 42. Otwinowski Z, Minor Wladek. Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 1997; 276:307–326. 43. Kabsch W. XDS. Acta Crystallogr D Biol Crystallogr. 2010; 66:125–132.10.1107/ S0907444909047337 [PubMed: 20124692] 44. Karplus PA, Diederichs K. Linking Crystallographic Model and Data Quality. Science. 2012; 336:1030–1033.10.1126/science.1218231 [PubMed: 22628654] 45. Vagin A, Teplyakov A. Molecular replacement with MOLREP. Acta Crystallogr D Biol Crystallogr. 2010; 66:22–25.10.1107/S0907444909042589 [PubMed: 20057045] 46. McCoy AJ, Grosse-Kunstleve RW, Adams PD, Winn MD, Storoni LC, Read RJ. Phaser crystallographic software. J Appl Crystallogr. 2007; 40:658–674.10.1107/S0021889807021206 [PubMed: 19461840]
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 13
Author Manuscript Author Manuscript
47. Murshudov GN, Vagin AA, Dodson EJ. Refinement of Macromolecular Structures by the Maximum-Likelihood Method. Acta Crystallogr D Biol Crystallogr. 1997; 53:240–255.10.1107/ S0907444996012255 [PubMed: 15299926] 48. Adams PD, Afonine PV, Bunkóczi G, Chen VB, Davis IW, Echols N, et al. PHENIX : a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr D Biol Crystallogr. 2010; 66:213–221.10.1107/S0907444909052925 [PubMed: 20124702] 49. Emsley P, Cowtan K. Coot: model-building tools for molecular graphics. Acta Crystallogr D Biol Crystallogr. 2004; 60:2126–2132.10.1107/S0907444904019158 [PubMed: 15572765] 50. Afonine PV, Moriarty NW, Mustyakimov M, Sobolev OV, Terwilliger TC, Turk D, et al. FEM: feature-enhanced map. Acta Crystallogr D Biol Crystallogr. 2015; 71:646–666.10.1107/ S1399004714028132 [PubMed: 25760612] 51. Krieger E, Vriend G. YASARA View—molecular graphics for all devices—from smartphones to workstations. Bioinformatics. 2014; 30:2981–2982.10.1093/bioinformatics/btu426 [PubMed: 24996895] 52. Duan Y, Wu C, Chowdhury S, Lee MC, Xiong G, Zhang W, et al. A point-charge force field for molecular mechanics simulations of proteins based on condensed-phase quantum mechanical calculations. J Comput Chem. 2003; 24:1999–2012.10.1002/jcc.10349 [PubMed: 14531054] 53. Krissinel E, Henrick K. Inference of macromolecular assemblies from crystalline state. J Mol Biol. 2007; 372:774–797.10.1016/j.jmb.2007.05.022 [PubMed: 17681537]
Author Manuscript Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 14
Author Manuscript
Highlights 1.
We describe the first application of crystal structure information to provide pharmacogenetic insight into a therapeutic T-cell receptor antibody.
2.
The crystal structure reveals a new binding mode responsible for the subnanomolar affinity of a TCR-like antibody for an HLA receptor-peptide complex, previously shown to be potent in mouse cancer models.
3.
Structural information was used for the first time to predict compatibility of a cancer immunotherapy with multiple HLA types.
4.
We describe a structure-based strategy to predict off-target binding which may lead to toxicity.
Author Manuscript Author Manuscript Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 15
Author Manuscript Author Manuscript Author Manuscript
Fig. 1. Overall complex structure
Author Manuscript
(A) The cartoon backbone rendering of the ESK1/pMHC complex is shown. HLAA*02:01and β2-microglobulin are colored blue and magenta. HLA helices α1 and α2 in the RMF binding site are red. The light and heavy antibody Fab fragments are cyan and orange respectively and are labeled by their variable (VL, VH,) or constant (CL, CH) domains. (B) The surface of the variable domains of the antibody is shown where the heavy chain contact surface is colored olive and the light chain contact surface is lime. (C) The top down view of the HLA surface is shown with the HLA contact surface colored yellow and the RMF contact surface colored pink. (D) The variable loop regions of the light and heavy chains are shown at the antigen binding site. VL loops L1, L2, and L3 are colored blue, purple, and magenta, respectively. VH loops H1, H2, and H3 are colored yellow, red, and green respectively. (E) The top-down view shows the RMF binding site with the variable loops circled.
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 16
Author Manuscript Author Manuscript
Fig. 2. Cation-π interaction network at binding interface
The high affinity-binding site features two networks of cation-π interactions at the interface between ESK1 and HLA-A*02:01. Tyr101 and Tyr105 from ESK1 CDR-H3 (orange), straddle HLA (blue) Arg170. HLA Arg108 and Arg169 straddle ESK1 CDR-H3 Tyr55.
Author Manuscript Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 17
Author Manuscript Author Manuscript Author Manuscript Fig. 3. Structural comparison between ESK1 and other TCRs and TCR-like antibodies
Author Manuscript
(A) Structural alignment between ESK1 and twenty TCR/HLA-A*02:01/peptide structures shows that ESK1 CDR loops reach regions of HLA α2 that TCR loops do not. This includes the circled arginine rich region: Arg108, Arg169, and Arg170. The blue ESK1/pHLA structure has VH domains H1, H2, and H3, colored yellow, red, and green respectively. (B) Structural alignment of TCR-mimic-antibody/HLA-A*02 structures shows that antibodies are more diverse in binding position and angle than TCRs, though none bind to HLA via the arginine rich ESK1 binding site. (C–F) The footprints of three TCRs and ESK1 show that TCRs bind to the mid-helical region, in contact with the peptide, whereas ESK1 reaches beyond the canonical TCR binding site.
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 18
Author Manuscript Author Manuscript Author Manuscript
Fig. 4. Peptide interactions
Author Manuscript
(A) RMF binds HLA with Met2, Phe3, Pro7, and Leu9 facing the HLA receptor and Arg1, Pro4, Asn5, and Tyr8 facing solvent and/or antibody. The heavy and light chains are in orange and cyan. Arg1, Pro4 and Asn5 are within van der Waals distance to ESK1. (B) RMF Arg1 is located near hydrogen bond distance of Ser102. Arg1 also forms potential hydrogen bonds to HLA Thr163 and Trp167 and is within van der Waals distance to ESK1 Tyr104. Also shown is the proximity of RMF Pro4 to ESK1 light chain Asp94. (C) Binding avidity of WT1 substituted peptides relative to RMF, as determined by flow cytometry. EW is a peptide (QLQNPSYDK) unrelated to RMF, isolated from Ewing sarcoma, used here as a negative control. Results were normalized to RMF binding. Scrambled corresponds to a positive control peptide which binds HLA-A*02 but not ESK1. O corresponds to ornithine and Ct to citrulline; all other amino acid abbreviations are standard.
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 19
Author Manuscript Author Manuscript Author Manuscript Fig. 5. Different binding modes of ESK1 and Hyb3 to HLA-peptide complexes
ESK1 (blue) and Hyb3 (green) differ in binding orientation to HLA-peptide complexes by 58°. ESK1 contacts the N-terminal portion of the antigen peptide. Hyb3 contacts the Cterminal portion of the antigen peptide.
Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 20
Author Manuscript Author Manuscript
Fig. 6. Mapped HLA-A*02 subtype differences
Mapped positional differences between the subtypes are colored red: A*02:02, orange: A*02:03, green: A*02:05, blue: A*02:06, magenta: A*02:07, and yellow: A*02:11. The ESK1 footprint is shown in light blue. None of the residues that vary between A*02 subtypes contacts ESK1.
Author Manuscript Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 21
Author Manuscript Author Manuscript
Fig. 7. Validation of predictions of potential ESK1 off-target epitopes
ESK1 binding was measured by flow cytometry. Relative binding was normalized to 25 μg/ml RMF. Relative binding of a scrambled control peptide or two identified peptides derived from MED13L and PIGQ are shown.
Author Manuscript Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 22
Table 1
X-ray diffraction data collection and refinement statistics
Author Manuscript
Data were collected at the Stanford Synchrotron Radiation Lightsource Beamline 12-2 on a Pilatus detector with wavelength 0.979500 Å.
Author Manuscript
Space group
P212121
Cell dimensions a, b, c (Å)
70.1, 118.3, 126.3
Resolution (Å)
37.6 – 3.05
Rsym*
0.158 (1.11)
I/σ*
13.9 (2.1)
CC½
99.8 (77.6)
Completeness (%)*
99.2 (96.6)
Redundancy*
11.8 (10.3)
Rwork/Rfree
0.198 / 0.254
Average B factor
74.8
RMS deviation, bond lengths (Å)
0.004
RMS deviation, bond angles (°)
0.74
*
Highest resolution shell is shown in parenthesis.
Author Manuscript Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 23
Table 2
Comparison of TCR interactions
Author Manuscript
A comparison between interface interactions of ESK1/pHLA-A*02:01 and three TCR/HLA-A*02:01 structures showing the contact surface area and binding free energy calculated from experimental dissociation constants. ESK1 has the strongest binding free energy despite having a total contact surface area comparable to TCRs.
TCR/HLA-A*02/peptide
Author Manuscript
MHC/TCR contact surface Å2 (PISA)
Peptide/TCR contact surface Å2 (PISA)
Total TCR contact surface Å2 (PISA)
ΔG (kcal/mol)
555 (α) + 198 (β) = 753
182 (α) + 214 (β) = 396
1149
−7.7
548 (α) + 171 (β) = 719
164 (α) + 173 (β) = 337
1056
−8.0
206 (α) + 435 (β) = 641
80 (α) + 199 (β) = 279
920
−7.1
585 (h)+ 305 (I)= 890
70 (h)+ 90 (I) = 160
1050
−14
Author Manuscript Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Author Manuscript Table 3
Author Manuscript
Author Manuscript
1.40E-02
1.06E-02
5.79E-03
1.91E-03
1.50E-03
9.30E-04
A*02:03
A*02:05
A*02:06
A*02:07
A*02:11
1.35E-02
A*02:01
A*02:02
kd [1/s]
Subtype
1.80E-05
3.00E-05
2.30E-05
1.10E-04
2.70E-04
4.10E-04
6.70E-04
Error in kd [1/s]
8.40E+05
9.53E+05
9.32E+05
1.03E+06
5.24E+05
5.91E+05
1.02E+06
ka [1/Ms]
5.50E+03
1.00E+04
7.50E+03
1.90E+04
3.10E+04
4.30E+04
4.80E+04
Error in ka [1/Ms]
1.1
1.6
2.1
5.6
20.3
23.6
13.2
KD [nM]
0.02
0.04
0.03
0.2
1.3
1.9
0.90
Error in KD [nM]
14.0
14.0
12.2
10.9
17.5
9.2
10.6
Chi2 [RU2]
145.1
92.8
124.3
199
61.4
42.2
115.8
Rmax [RU2]
simulate the bimolecular mechanism in its entirety. Chi2 and Rmax were calculated using the BIAcore X100 evaluation software.
Binding parameters were measured by surface plasmon resonance on a Biacore X100. Despite the presence of a bivalent analyte, the 1:1 binding model was used since the purpose of this section is to provide a quantitative comparison between the binding between the HLA-A*02 variants, rather than to
Author Manuscript
Characterization of ESK1 binding to HLA-A*02 variants with RMF
Ataie et al. Page 24
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
J Mol Biol. Author manuscript; available in PMC 2017 January 16. Published in final edited form as: J Mol Biol. 2016 January 16; 428(1): 194–205. doi:10.1016/j.jmb.2015.12.002.
Structure of a TCR mimic antibody with target predicts pharmacogenetics Niloufar Ataie1, Jingyi Xiang2, Neal Cheng2, Elliott J. Brea3, Wenjie Lu1, David A. Scheinberg3,4, Cheng Liu2, and Ho Leung Ng1,5 1University
of Hawaii at Manoa, Department of Chemistry. 2545 McCarthy Mall. Honolulu, HI 96822-2275. USA
Author Manuscript
2Eureka
Therapeutics Inc. 5858 Horton St, Emeryville, CA 94608. USA
3Sloan-Kettering 4Weill
Institute. 1275 York Ave. New York, NY 10065. USA
Cornell Medical College. 1305 York Ave. New York, NY 10021. USA
5University
of Hawaii Cancer Center. 2545 McCarthy Mall. Honolulu, HI. 96822-2275. USA
Abstract
Author Manuscript
Antibody therapies currently target only extracellular antigens. A strategy to recognize intracellular antigens is to target peptides presented by immune HLA receptors. ESK1 is a human, T-cell receptor (TCR)-mimic antibody that binds with sub-nanomolar affinity to the RMF peptide from the intracellular oncoprotein Wilms Tumor 1 (WT1) in complex with HLA-A*02:01. ESK1 is therapeutically effective in mouse models of WT1+ human cancers. TCR-based therapies have been presumed to be restricted to one HLA subtype. The mechanism for the specificity and high affinity of ESK1 is unknown. We show in a crystal structure that ESK1 Fab binds to RMF/HLAA*02:01 in a different mode than TCRs. From the structure, we predict and then experimentally confirm high affinity binding with multiple other HLA-A*02 subtypes, broadening the potential patient pool for ESK1 therapy. Using the crystal structure, we also predict potential off-target binding that we experimentally confirm. Our results demonstrate how protein structure information can contribute to personalized immunotherapy.
Graphical abstract
Author Manuscript
Correspondence: [email protected], phone +1 (808) 956-2014. Conflicts of interest: J. Xiang, N. Cheng, and C. Liu have ownership interest (including patents) in Eureka Therapeutics Inc. D.A. Scheinberg is an inventor of intellectual property related to ESK1 that is owned by Sloan Kettering and licensed to Novartis. No potential conflicts of interest were disclosed by the other authors. Accession numbers Protein Data Bank: Coordinates and structure factors have been deposited with accession code 4WUU. Supplementary data to this article can be found online at http://dx.doi.org/xxxxxx. Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Ataie et al.
Page 2
Author Manuscript Keywords antibodies; MHC; cancer; immunotherapy
Author Manuscript
Introduction Antibody targeted therapies have been highly successful in selectively killing cancer cells with reduced adverse effects. Two dozen cancer therapeutic antibodies are approved for clinical use. Cancer antibody therapies target extracellular proteins, and a major unmet challenge has been targeting intracellular tumor specific, mutated oncogenic proteins, which are not generally present on cell surfaces [1].
Author Manuscript
One strategy to target intracellular antigens is to direct TCR based therapies, either with adoptive T-cell therapy or TCR mimic antibodies, to oncoprotein derived peptides that are displayed on the cell surface by major histocompatibility complex (MHC) class I receptors [2–4]. The Wilms Tumor 1 protein (WT1) is overexpressed in most hematological and solid tumor cancers [5,6], and peptides derived from WT1 are displayed by HLA receptors as 9–10 amino acid T-cell epitopes for presentation to the TCR [7,8]. WT1 is the highest ranked target by the National Cancer Institute pilot project to prioritize cancer immunotherapy antigens for clinical trials based on criteria including the number of patients expressing antigen, specificity of antigen to cancer cells, expression levels of protein, oncogenicity, and immunogenicity [9]. While targeting tumor antigens on MHC-I is a viable therapeutic strategy, investigation into the structure of TCR based therapies has looked primarily at ontarget reactivity, and the nature of the TCR and peptide MHC interaction may allow for offtarget cross-reactivity as well. The ability to predict potential off targets may avoid toxicities of TCR and TCR mimic based therapies and warrants a more systematic evaluation [10].
Author Manuscript
ESK1 is a human, TCR mimic (TCRm) monoclonal antibody (mAb) that we previously engineered to bind the WT1 derived peptide epitope RMFPNAPYL (RMF)/HLA-A*02:01 complex with 0.2 nM affinity and to mediate WT1-restricted cancer cell death in xenograft mouse models of human cancers by antibody-dependent cell-mediated cytotoxicity (ADCC) [11–14]. Toxicity studies in transgenic HLA-A*02:01 expressing mice demonstrate lack of uptake or toxicity in normal tissues [13]. The effectiveness and low toxicity of ESK1 have led to its further development as a drug. There is a need for high resolution characterization
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 3
Author Manuscript
of the binding site and the complementary determining regions (CDRs) that contribute to ESK1 binding to address questions about specificity, possible cross-reactivities and offtargets, self-reactivity, and compatibility with other HLA-A*02 subtypes. While HLAA*02:01 is the most common HLA-A*02 subtype in the United States and Europe, other subtypes bearing different peptide and TCR binding specificities are found across different ethnic groups worldwide [15,16].
Author Manuscript
We solved the crystal structure of the RMF/HLA-A*02:01/ESK1 (Fab) complex to 3.05 Å. The ESK1 variable domains bind the HLA and peptide in a different mode than TCRs and other TCR mimic Fabs [17,18]. ESK1 CDR loops contact regions of HLA that TCRs typically do not reach. Our structure and binding studies also show that RMF acts as an electrostatic key in mediating ESK1 specificity and activity with Arg1 playing a central role. RMF Pro4 also contributes to ESK1 binding. The rest of the binding interface is limited to interactions between ESK1 and a region on the HLA receptor that is conserved between subtypes, suggesting that binding is compatible with other common HLA-A*02 subtypes, which we experimentally confirmed. This potentially broadens the target patient populations for this drug beyond the HLA-A*02:01 subtype found predominantly in Caucasians to multiple other ethnic groups [19]. The crystal structure allowed additional predictions of possible cross reactivity with several human self-peptides, some of which we experimentally confirmed in vitro. This is the first evaluation of cross reactivity of a TCR based therapy in silico, allowing for structural and bioinformatics data to create a pipeline for better predicting specificity. In this way, we report that structural data can be a valuable tool for pre-clinical characterization of antibody pharmacogenetics and toxicology in genetically diverse patient populations.
Author Manuscript
Results and Discussion Overall structure The ESK1 Fab fragment binds the peptide-MHC (pMHC) with the variable domain (Fig. 1a) contacting 160 Å2 of the first five residues of RMF (Fig. 1b). RMF interacts with 873 Å2 of the HLA receptor in a manner closely superimposable with the crystal structure of the HLAA*02:01/RMF complex without bound antibody or TCR with an RMSD of 0.75 Å (PDB 3HPJ) [20]. In total, the ESK1 antibody-HLA surface is 890 Å2 and the total ESK1-pMHC surface is 1050 Å2 (Fig. 1b, c). Electron density quality at the binding interface is excellent for diffraction data of this resolution and unambiguously resolves the side chains (Fig. S1). The structure is well refined with Rfree = 0.25, which is in the upper 50% for crystal structures at this resolution according to the PDB validation report (Table 1).
Author Manuscript
Binding interactions with the CDR loops Figures 1d–e show the interactions of the CDR loops with the pMHC. Four tyrosines on CDR-H2 and CDR-H3 (residues 55, 101, 104, 105), stabilize the interface. Hydrogen bonds with ideal geometry and van der Waals contacts are common as are prominent networks of cation-π interactions (Fig. 2). Cation-π interactions are common in protein interfaces and are among the strongest non-covalent forces within proteins, contributing energies up to 4 kcal/ mol, comparable to hydrogen bonds [21,22]. They are a primary driver of long range
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 4
Author Manuscript
interactions, contributing favorable forces up to 6 Å away and at angles ranging from the favored θ = 0, where the cation is directly above the π ring system, to beyond 60° [23]. CDR-H2 and CDR-H3 make extensive cation-π interactions with Arg108, Arg169, and Arg170 of the HLA. The extensive chains of cation-π interactions in the ESK1-pMHC are consistent with those frequently seen in other antibody-antigen complexes [24]. Comparison to TCRs and other TCR-like antibodies
Author Manuscript
We compared the structure of ESK1 + HLA-A*02 to HLA-A*02 with twenty other TCRs and TCR mimic antibodies [17,18,25–30] to determine the molecular basis for the 1000-fold higher affinity for ESK1 compared to TCRs (Fig. 3). The four TCR mimic antibodies compared differ widely in binding orientations. None of the other three antibodies binds to the same arginine cluster as ESK1. ESK1 CDR loops reach HLA-A*02 regions untouched by TCRs that interact with the HLA arginine cluster. Compared to the sub-nanomolar Kd of ESK1, TCRs typically bind with a maximum Kd of 1 μM [31]. Table 2 compares the contact surface area of three TCR/pHLA structures with that of ESK1/pHLA. ESK1 makes more contacts with the HLA than the peptide compared to three other TCR pMHCs. ESK1 contacts only 160 Å2 of the peptide, which is 15% of the total pMHC contact surface. In contrast, in the other three pMHCs, the peptide contribution is 30–34% of the overall contact surface. While ESK1 does not have the largest contact surface compared to the other pMHCs, it has the highest affinity [32,33]. Peptide binding specificity
Author Manuscript
ESK1 binds the HLA receptor centered on RMF Arg1 (Fig. 4). ESK1 primarily interacts with RMF Arg1 and Pro4. RMF Met2, Phe3, Ala6, Phe7, and Leu9 side chains contact the bottom of the HLA peptide binding region. Asn5 and Tyr8 face the solvent and do not contact ESK1.
Author Manuscript
In order to probe the structural requirements of the ESK1 antibody to the HLA-A*02/RMF complex, we made substitutions along the length of the RMF peptide. ESK1 binding to the peptide/MHC complex was measured in the TAP1/2 deficient T2 lymphoblastoid cell line by pulsing various peptides onto HLA-A*02 (supplementary Fig. S2). Substitution at position 1 with citrulline, sterically similar to arginine, but uncharged, eliminated binding. Lysine substitution at position 1 decreased binding 40-fold. Substitution at position 1 with ornithine, similar to lysine, but shorter by a carbon atom, or with histidine, eliminated ESK1 binding. Overall, the binding data support a highly specific fit of the ESK1 electrostatic lock for arginine. Arg1 is near several polar residues (Fig. 4, supplementary Fig. S2) in an otherwise buried hydrophobic pocket capped by a highly negatively charged loop from the ESK1 heavy chain. This pocket (supplementary Fig. S2e) acts as an electrostatic lock in which the negatively charged dipoles of the backbone carbonyl oxygens of Ser102, Gly103, and the sidechain of Asp106 converge without countercharge, creating a buried space of focused negative charge. Tyrosine at position 1 does not provide the countercharge needed to fit the electrostatic lock. The crystal structure can help reveal the factors driving specificity and cross reactivity of ESK1 toward other peptide/MHC epitopes.
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 5
Author Manuscript
CDR-L3 partially shields RMF Pro4 from solvent (Fig. 4). Asp94 of the light chain is positioned 3.4 Å away from Pro4. Substitution to glycine, alanine, or serine at RMF position 4 does not significantly affect binding, but substitution to cysteine decreases binding tenfold (Fig. 4c). All low energy rotamers of cysteine modeled into position 4 of RMF sterically clashed with Asp94. Surprisingly, the substitution of tryptophan at position 4 showed no loss in binding affinity. Tryptophan modeled into that position forms potentially favorable cation-π interactions with HLA Lys66. Large residues modeled into the fifth position clash with nearby residues. Other large bulky residues at position five decrease binding over tenfold (Fig. 4c).
Author Manuscript
The binding mode of ESK1 differs most from that of the Fab Hyb3 bound to HLA-A*01 receptor with the MAGE-A1 peptide [34]. While ESK1 primarily contacts the N-terminal of WT1, Hyb3 primarily contacts the C-terminal of MAGE-A1 (Fig. 5). The structures of the HLA-receptors are highly conserved between the two structures, with an RMSD of 1.1 Å over all Cα. ESK1 and Hyb3 differ in binding orientation by 58° and represent the diverse binding modes to HLA-peptide complexes available to antibodies, as opposed to the canonical diagonal binding conserved among T-cell receptors. Like ESK1, Hyb3 demonstrates surprisingly high binding specificity in vitro and in vivo despite making close contacts with less than half the peptide, and is capable of differentiating between MAGE-A1 and MAGE-A3 peptides which differ by only three amino acids [34,35]. This is consistent with a model in which small peptide differences affect HLA receptor conformation and thereby, Fab binding [35]. HLA-A*02 subtype specificity
Author Manuscript
An important potential therapeutic advantage of TCRm antibodies over engineered T-cells is that antibody therapy does not require customization on a per patient basis. However, use of a TCRm antibody is still restricted by patient HLA type. Producing TCRm antibodies that are effective across a broad range of HLA types would greatly increase their therapeutic value. We demonstrate that structural information can be used to predict that ESK1 is effective across many of the most common HLA-A*02 subtypes.
Author Manuscript
HLA-A*02 alleles vary widely across different ethnic populations. HLA-A*02:01 is the most commonly found variant, especially in Caucasians in the United States, but other subtypes are more common in other ethnic groups. Six subtypes are found in hundreds of millions of people worldwide that differ from HLA-A*02:01 through 1 to 4 amino acid changes in the side walls of the peptide binding pocket. Models were built of several common HLA-A*02 subtypes, using crystal structure coordinates of the subtype if available (A*02:03, A*02:06, A*02:07), to predict the effect of binding to RMF and ESK1. All residues that differ between A*02:01 and the other six subtypes are not within 4 Å of ESK1 (Fig. 6). Modeling predicts that RMF binding would be retained across these A*02 subtypes due to the lack of interfering steric or electronic interactions. We predicted and experimentally confirmed that these six subtypes bind RMF and ESK1 (Table 3, supplementary Fig. 3) by surface plasmon resonance using immobilized, biotinylated HLA +RMF. The measured Kd of ESK1 to RMF + A*02:01 is 13.2 nM, higher than measured by
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 6
Author Manuscript
the whole cell method. All six HLA variants bind to ESK1 with similar or lower Kd (1.1 nM for A*02:11 to 23.6 nM for A*02:02), suggesting patient compatibility. Structure based prediction of potential ESK1 off-target toxicity A major advantage of antibody over small molecule therapeutics is that toxicity tends to be more predictable and related to mechanism of action due to their higher specificity [36]. Nevertheless, off-target binding is still a serious concern for immunotherapy. Currently, specificity is enforced by counterscreening antibody candidates against non-specific targets such as cell lysate or MHC-receptors bearing untargeted antigen peptides. We demonstrate for the first time that structural knowledge and simple computational analysis can be used to predict off targets and further engineer TCRm specificity.
Author Manuscript Author Manuscript
Given the structural and biochemical data supporting the key binding interactions of the ESK1 antibody to RMF positions Arg1, Phe3, and Pro4, we performed a BLAST search for similar expressed human peptide sequences. The peptides were then subjected to binding affinity analysis using NETMHC [37] to predict peptides with a Kd to HLA-A*02:01 < 500 nM. Using artificial neural networks, NETMHC predicts MHC-peptide binding with 78% accuracy. The two highest ranking peptides were identified to be from MED13L (mediator complex subunit 13-like), RMFPTPPSL, and PIGQ (phosphatidylinositol glycan anchor biosynthesis, class Q), RMFPGEVAL, and tested as potential off targets to validate our hypothesis. ESK1 bound cell surface MHC with either peptide (Fig. 7). Structural modeling supports compatibility of these two peptides with HLA-A*02:01 and ESK1 binding. In silico mutagenesis of the RMF peptide in the ESK1-HLA-RMF crystal structure showed that the sequences in the MED13L and PIGQ peptides could be accommodated with little change to HLA and ESK1 binding surfaces (supplemementary Fig. S4). Studies on normal donor peripheral blood monocyte cells showed potential off target binding of ESK1 to a small subset of healthy donor CD19+ B-cells [11], which were WT1 negative. In this way, it is possible that ESK1 binding to this subset of cells may be attributable to non-WT1 derived epitopes, though we cannot determine the peptide from these studies. Therefore, utilizing structural knowledge of TCRm antibody-MHC complexes provides a novel basis for determining possible off-target self-proteins in advance of clinical trials and could serve as a platform for future development of specific TCRm antibodies to minimize off target effects.
Conclusions Author Manuscript
TCRs with moderate affinity (Kd = 1–100 μM) to antigens are produced in the body and provide the basis for a number of therapeutic agents, including soluble TCR-based constructs, TCR engineered cells, and peptide vaccines. TCR-based drugs have several potentially serious limitations. One is their restriction to a specific HLA subtype, reducing the potential patient population that may benefit from the therapeutic approach to a small fraction of afflicted patients. An additional limitation is cross-reactivity of typically low affinity TCRs to other peptide epitopes presented by MHC molecules that has resulted in severe toxicity in human trials [10]. The current state of the art of antibody engineering is more advanced than with TCRs, allowing for greater structural and binding diversity. Here
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 7
Author Manuscript
we present the first use of structural and biochemical characterization of the interaction of a TCRm antibody to its cognate peptide/MHC epitope in an effort to investigate and overcome both these limitations. An essential feature in the development of a TCRm is high specificity to the molecular complex of both the MHC molecule and presented peptide simultaneously, as mAb reactivity with the MHC alone would confer reactivity with every nucleated cell. Phage display can produce much higher affinity TCR mimic antibodies such as ESK1 than occurs in vivo with TCRs, while simultaneously selecting against antibodies that may cross-react with the MHC complex alone. Second, the high affinity of TCRm antibodies allows pharmacologic dosing and retention of the drug at the cancer cell site in vivo. The TCRm, ESK1, binds with sub-nanomolar affinity, over 1000-fold more tightly than natural TCRs. Our data demonstrate the molecular basis for both the specificity and affinity of ESK1.
Author Manuscript
Surprisingly, the high affinity binding was mainly driven by interactions between the antibody and HLA receptor that are restricted to regions identical or similar between at least six common HLA-A*02 subtypes that vary in different ethnic groups worldwide. This result led us to hypothesize that ESK1 could bind to the same WT1 peptide epitope when presented by various other HLA-A*02 subtypes in addition to HLA-A*02:01. Experimental confirmation of ESK1 binding to these HLA-A*02 variants extends the potential scope of ESK1 therapy to millions of additional patients. In addition to its pharmacologic advantages, this unexpected finding shows that the use of a TCR mimic therapy may be a novel strategy to overcome the limits of HLA subtype restriction of typical TCR-based therapy and therefore have broader therapeutic applications.
Author Manuscript Author Manuscript
Naturally occurring TCR can bind off-target peptides [38,39]. We were able to use the structural information to predict potential cross-reactive peptides of ESK1 in humans. When bound to the HLA receptor, ESK1 primarily interacts with the N-terminal of RMF. As a consequence, other peptides expressed in humans that are homologous to the N terminal of RMF, and bind to HLA-A*02 with the anchor residue at position 9, but that differ at positions 5–8 in the C terminal, might be ESK1 off-targets. We confirmed that such peptides in complex with HLA-A*02:01 can bind to ESK1 on peptide-pulsed T2 cells in vitro. Thus, it may be possible to account for both structure and potential off targets using a bioinformatics approach as we have shown with ESK1 mAb with WT1-HLA-A*02:01. The determination of whether these alternate targets are processed and expressed on the surface in normal tissues would be challenging, but identification of these targets in advance of clinical trials may allow for appropriate prospective clinical observation. In conclusion, we demonstrate for the first time that detailed, integrated structural, genetic, and biochemical analysis can provide a platform for appropriate selection and stratification of patient in efforts to maximize treatment efficacy and minimize toxicity. As such, this structure-based approach may serve as a key step towards personalized and precision medicine.
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 8
Author Manuscript
Materials and methods ESK1 design, construction, purification The antibody ESK1 was prepared as described previously [11]. Briefly, ESK1 was derived from the top clone selected from the Eureka Therapeutics human single-chain variable fragment (scFv) phage library with specific binding for the RMF/HLA-A*02:01 complex. The full-length human IgG1 form of ESK1 antibody was produced in the Chinese Hamster Ovary (CHO) cell line. ESK1 antibody was affinity purified by Protein A chromatography with an AKTA FPLC system. ESK1 flow cytometry binding assay to RMF peptides
Author Manuscript
T2 cells, a TAP1/2 deficient HLA-A*02 expressing cell line that does not present endogenous peptides, were used for pulsing peptides. T2 cells were purchased from ATCC and cultured in RPMI with 10% FBS. Peptides were purchased from Genemed Synthesis (San Antonio, TX USA) with >90% purity. Peptides were resuspended to 20 mg/ml in DMSO and diluted to working stocks in PBS. T2 cells, at 1×106 cells/ml, were pulsed with peptides at various concentrations overnight in RPMI with 10% dialyzed FBS and 10 μg/ml β2-microglobulin. T2 cells were incubated with peptide overnight at 37°C with 5% CO2. Cells were then spun down, washed with PBS, and blocked with human FcR blocking solution for 15 minutes on ice. T2 cells were then stained with ESK1 labeled with APC at 3 μg/ml final antibody for 45 minutes on ice. Cells were then spun down and resuspended in FACS buffer and run on FACS Calibur. Off target binding of ESK1
Author Manuscript
Protein BLAST from NCBI, along with analysis using NetMHC 3.0 for peptides with a binding of 2.0 and CC½ showing statistical significance at 0.1% [44]. The structure was solved by molecular replacement using the structures of HLA-A2 (PDB 3HPJ) and the Fab fragment from antibody HC84-1 (PDB 4JZN) using MOLREP [45] and Phaser [46]. The structure was refined with REFMAC and Phenix.refine using TLS parameters for each chain [47,48] with manual rebuilding with Coot [49] and use of Feature Enhanced Maps [50]. Processing and refinement statistics are shown in Table 1. Structure analysis and modeling
Author Manuscript
Modeling was performed with PyMol, Coot, and Yasara [51]. In silico mutagenesis was performed in Yasara, involving selection of the optimal target residue rotamer, followed by energy minimization with the AMBER03 force field [52]. Interaction surface properties were calculated with Pisa [53].
Supplementary Material Refer to Web version on PubMed Central for supplementary material.
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 10
Author Manuscript
Acknowledgments We thank Aina Cohen (SSRL) for assistance with diffraction data collection. This work was supported by University of Hawaii startup funds (HLN), the Leukemia and Lymphoma Society (DAS), NIH R01CA55349 (DAS) and P01 23766 (DAS), NIH NIGMS MSTP GM007739 (EJB), NIH NIGMS P41GM103393 (DAS), P30 CA008748 (DAS), and the Major Family Fund (DAS). Use of the Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory, is supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under Contract No. DE-AC02-76SF00515. The SSRL Structural Molecular Biology Program is supported by the DOE Office of Biological and Environmental Research.
Abbreviations
Author Manuscript
CDR
complementary determining regions
HLA
human leukocyte antigen
mAb
monoclonal antibody
MHC
major histocompatibility complex
pMHC
peptide-MHC complex
TCR
T-cell receptor
TCRm
T-cell receptor mimic
WT1
Wilms Tumor 1
References
Author Manuscript Author Manuscript
1. Scott AM, Wolchok JD, Old LJ. Antibody therapy of cancer. Nat Rev Cancer. 2012; 12:278– 287.10.1038/nrc3236 [PubMed: 22437872] 2. Dahan R, Reiter Y. T-cell-receptor-like antibodies - generation, function and applications. Expert Rev Mol Med. 2012; 14:e6.10.1017/erm.2012.2 [PubMed: 22361332] 3. Weidanz JA, Hawkins O, Verma B, Hildebrand WH. TCR-like biomolecules target peptide/MHC Class I complexes on the surface of infected and cancerous cells. Int Rev Immunol. 2011; 30:328– 340.10.3109/08830185.2011.604880 [PubMed: 22053972] 4. Yee C, Thompson JA, Byrd D, Riddell SR, Roche P, Celis E, et al. Adoptive T cell therapy using antigen-specific CD8+ T cell clones for the treatment of patients with metastatic melanoma: in vivo persistence, migration, and antitumor effect of transferred T cells. Proc Natl Acad Sci U S A. 2002; 99:16168–16173.10.1073/pnas.242600099 [PubMed: 12427970] 5. Keilholz U, Menssen HD, Gaiger A, Menke A, Oji Y, Oka Y, et al. Wilms’ tumour gene 1 (WT1) in human neoplasia. Leukemia. 2005; 19:1318–1323.10.1038/sj.leu.2403817 [PubMed: 15920488] 6. Yang L, Han Y, Suarez Saiz F, Saurez Saiz F, Minden MD. A tumor suppressor and oncogene: the WT1 story. Leukemia. 2007; 21:868–876.10.1038/sj.leu.2404624 [PubMed: 17361230] 7. Ohminami H, Yasukawa M, Fujita S. HLA class I-restricted lysis of leukemia cells by a CD8(+) cytotoxic T-lymphocyte clone specific for WT1 peptide. Blood. 2000; 95:286–293. [PubMed: 10607714] 8. Oka Y, Elisseeva OA, Tsuboi A, Ogawa H, Tamaki H, Li H, et al. Human cytotoxic T-lymphocyte responses specific for peptides of the wild-type Wilms’ tumor gene (WT1) product. Immunogenetics. 2000; 51:99–107. [PubMed: 10663572] 9. Cheever MA, Allison JP, Ferris AS, Finn OJ, Hastings BM, Hecht TT, et al. The Prioritization of Cancer Antigens: A National Cancer Institute Pilot Project for the Acceleration of Translational Research. Clin Cancer Res. 2009; 15:5323–5337.10.1158/1078-0432.CCR-09-0737 [PubMed: 19723653]
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 11
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
10. Linette GP, Stadtmauer EA, Maus MV, Rapoport AP, Levine BL, Emery L, et al. Cardiovascular toxicity and titin cross-reactivity of affinity-enhanced T cells in myeloma and melanoma. Blood. 2013; 122:863–871.10.1182/blood-2013-03-490565 [PubMed: 23770775] 11. Dao T, Yan S, Veomett N, Pankov D, Zhou L, Korontsvit T, et al. Targeting the intracellular WT1 oncogene product with a therapeutic human antibody. Sci Transl Med. 2013; 5:176ra33.10.1126/ scitranslmed.3005661 12. Dubrovsky L, Pankov D, Brea EJ, Dao T, Scott A, Yan S, et al. A TCR-mimic antibody to WT1 bypasses tyrosine kinase inhibitor resistance in human BCR-ABL+ leukemias. Blood. 2014; 123:3296–3304.10.1182/blood-2014-01-549022 [PubMed: 24723681] 13. Veomett N, Dao T, Liu H, Xiang J, Pankov D, Dubrovsky L, et al. Therapeutic Efficacy of an FcEnhanced TCR-like Antibody to the Intracellular WT1 Oncoprotein. Clin Cancer Res. 2014; 20:4036–4046.10.1158/1078-0432.CCR-13-2756 [PubMed: 24850840] 14. Dao T, Pankov Dmitry, Scott Andrew, Korontsvit Tatyana, Zakhaleva Victoriya, Xu Yiyang, et al. Therapeutic targeting of the intracellular oncoprotein WT1 with a Bi-specific T cell engager antibody. Nat Biotechnol. 2015 In press. 15. Ellis JM, Henson V, Slack R, Ng J, Hartzman RJ, Katovich Hurley C. Frequencies of HLA-A2 alleles in five U.S. population groups. Predominance Of A*02011 and identification of HLAA*0231. Hum Immunol. 2000; 61:334–340. [PubMed: 10689125] 16. Gonzalez-Galarza FF, Christmas S, Middleton D, Jones AR. Allele frequency net: a database and online repository for immune gene frequencies in worldwide populations. Nucleic Acids Res. 2011; 39:D913–919.10.1093/nar/gkq1128 [PubMed: 21062830] 17. Mareeva T, Martinez-Hackert E, Sykulev Y. How a T cell receptor-like antibody recognizes major histocompatibility complex-bound peptide. J Biol Chem. 2008; 283:29053–29059.10.1074/ jbc.M804996200 [PubMed: 18703505] 18. Stewart-Jones G, Wadle A, Hombach A, Shenderov E, Held G, Fischer E, et al. Rational development of high-affinity T-cell receptor-like antibodies. Proc Natl Acad Sci U S A. 2009; 106:5784–5788.10.1073/pnas.0901425106 [PubMed: 19307587] 19. Marsh, SGE.; Parham, P.; Barber, LD. The HLA Facts Book. Academic Press; 1999. 20. Borbulevych OY, Do P, Baker BM. Structures of native and affinity-enhanced WT1 epitopes bound to HLA-A*0201: implications for WT1-based cancer therapeutics. Mol Immunol. 2010; 47:2519–2524.10.1016/j.molimm.2010.06.005 [PubMed: 20619457] 21. Burley SK, Petsko GA. Amino-aromatic interactions in proteins. FEBS Lett. 1986; 203:139–143. [PubMed: 3089835] 22. Dougherty DA. The Cation–π Interaction. Acc Chem Res. 2013; 46:885–893.10.1021/ar300265y [PubMed: 23214924] 23. Marshall MS, Steele RP, Thanthiriwatte KS, Sherrill CD. Potential energy curves for cation-pi interactions: off-axis configurations are also attractive. J Phys Chem A. 2009; 113:13628– 13632.10.1021/jp906086x [PubMed: 19886621] 24. Dalkas GA, Teheux F, Kwasigroch JM, Rooman M. Cation–π, amino–π, π–π, and H-bond interactions stabilize antigen–antibody interfaces. Proteins Struct Funct Bioinforma. 2014; 82:1734–1746.10.1002/prot.24527 25. Ding YH, Smith KJ, Garboczi DN, Utz U, Biddison WE, Wiley DC. Two human T cell receptors bind in a similar diagonal mode to the HLA-A2/Tax peptide complex using different TCR amino acids. Immunity. 1998; 8:403–411. [PubMed: 9586631] 26. Ding YH, Baker BM, Garboczi DN, Biddison WE, Wiley DC. Four A6-TCR/peptide/HLA-A2 structures that generate very different T cell signals are nearly identical. Immunity. 1999; 11:45– 56. [PubMed: 10435578] 27. Garboczi DN, Ghosh P, Utz U, Fan QR, Biddison WE, Wiley DC. Structure of the complex between human T-cell receptor, viral peptide and HLA-A2. Nature. 1996; 384:134–141. [PubMed: 8906788] J Immunol Baltim Md. 1950; 185:6394–6401. 2010. 28. Sami M, Rizkallah PJ, Dunn S, Molloy P, Moysey R, Vuidepot A, et al. Crystal structures of high affinity human T-cell receptors bound to peptide major histocompatibility complex reveal native diagonal binding geometry. Protein Eng Des Sel PEDS. 2007; 20:397–403.10.1093/protein/ gzm033 [PubMed: 17644531]
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 12
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
29. Borbulevych OY, Piepenbrink KH, Baker BM. Conformational melding permits a conserved binding geometry in TCR recognition of foreign and self molecular mimics. J Immunol Baltim Md 1950. 2011; 186:2950–2958.10.4049/jimmunol.1003150 30. Borbulevych OY, Piepenbrink KH, Gloor BE, Scott DR, Sommese RF, Cole DK, et al. T cell receptor cross-reactivity directed by antigen-dependent tuning of peptide-MHC molecular flexibility. Immunity. 2009; 31:885–896.10.1016/j.immuni.2009.11.003 [PubMed: 20064447] 31. Armstrong KM, Insaidoo FK, Baker BM. Thermodynamics of T-cell receptor-peptide/MHC interactions: progress and opportunities. J Mol Recognit JMR. 2008; 21:275–287.10.1002/jmr.896 [PubMed: 18496839] 32. Willcox BE, Gao GF, Wyer JR, Ladbury JE, Bell JI, Jakobsen BK, et al. TCR binding to peptideMHC stabilizes a flexible recognition interface. Immunity. 1999; 10:357–365. [PubMed: 10204491] 33. Davis-Harrison RL, Armstrong KM, Baker BM. Two different T cell receptors use different thermodynamic strategies to recognize the same peptide/MHC ligand. J Mol Biol. 2005; 346:533– 550.10.1016/j.jmb.2004.11.063 [PubMed: 15670602] 34. Hülsmeyer M, Chames P, Hillig RC, Stanfield RL, Held G, Coulie PG, et al. A major histocompatibility complex-peptide-restricted antibody and t cell receptor molecules recognize their target by distinct binding modes: crystal structure of human leukocyte antigen (HLA)-A1MAGE-A1 in complex with FAB-HYB3. J Biol Chem. 2005; 280:2972–2980.10.1074/ jbc.M411323200 [PubMed: 15537658] 35. Chames P, Willemsen RA, Rojas G, Dieckmann D, Rem L, Schuler G, et al. TCR-Like Human Antibodies Expressed on Human CTLs Mediate Antibody Affinity-Dependent Cytolytic Activity. J Immunol. 2002; 169:1110–1118.10.4049/jimmunol.169.2.1110 [PubMed: 12097420] 36. Brennan FR, Morton LD, Spindeldreher S, Kiessling A, Allenspach R, Hey A, et al. Safety and immunotoxicity assessment of immunomodulatory monoclonal antibodies. mAbs. 2010; 2:233– 255.10.4161/mabs.2.3.11782 [PubMed: 20421713] 37. Lundegaard C, Lamberth K, Harndahl M, Buus S, Lund O, Nielsen M. NetMHC-3.0: accurate web accessible predictions of human, mouse and monkey MHC class I affinities for peptides of length 8–11. Nucleic Acids Res. 2008; 36:W509–512.10.1093/nar/gkn202 [PubMed: 18463140] 38. Reiser JB, Darnault C, Grégoire C, Mosser T, Mazza G, Kearney A, et al. CDR3 loop flexibility contributes to the degeneracy of TCR recognition. Nat Immunol. 2003; 4:241–247.10.1038/ni891 [PubMed: 12563259] 39. Birnbaum ME, Mendoza JL, Sethi DK, Dong S, Glanville J, Dobbins J, et al. Deconstructing the peptide-MHC specificity of T cell recognition. Cell. 2014; 157:1073–1087.10.1016/j.cell. 2014.03.047 [PubMed: 24855945] 40. Altman, JD.; Davis, MM. Curr Protoc Immunol. John Wiley & Sons, Inc; 2001. MHC-Peptide Tetramers to Visualize Antigen-Specific T Cells. http://onlinelibrary.wiley.com/doi/ 10.1002/0471142735.im1703s53/abstract [accessed November 22, 2014] 41. Soltis SM, Cohen AE, Deacon A, Eriksson T, González A, McPhillips S, et al. New paradigm for macromolecular crystallography experiments at SSRL: automated crystal screening and remote data collection. Acta Crystallogr D Biol Crystallogr. 2008; 64:1210–1221.10.1107/ S0907444908030564 [PubMed: 19018097] 42. Otwinowski Z, Minor Wladek. Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 1997; 276:307–326. 43. Kabsch W. XDS. Acta Crystallogr D Biol Crystallogr. 2010; 66:125–132.10.1107/ S0907444909047337 [PubMed: 20124692] 44. Karplus PA, Diederichs K. Linking Crystallographic Model and Data Quality. Science. 2012; 336:1030–1033.10.1126/science.1218231 [PubMed: 22628654] 45. Vagin A, Teplyakov A. Molecular replacement with MOLREP. Acta Crystallogr D Biol Crystallogr. 2010; 66:22–25.10.1107/S0907444909042589 [PubMed: 20057045] 46. McCoy AJ, Grosse-Kunstleve RW, Adams PD, Winn MD, Storoni LC, Read RJ. Phaser crystallographic software. J Appl Crystallogr. 2007; 40:658–674.10.1107/S0021889807021206 [PubMed: 19461840]
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 13
Author Manuscript Author Manuscript
47. Murshudov GN, Vagin AA, Dodson EJ. Refinement of Macromolecular Structures by the Maximum-Likelihood Method. Acta Crystallogr D Biol Crystallogr. 1997; 53:240–255.10.1107/ S0907444996012255 [PubMed: 15299926] 48. Adams PD, Afonine PV, Bunkóczi G, Chen VB, Davis IW, Echols N, et al. PHENIX : a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr D Biol Crystallogr. 2010; 66:213–221.10.1107/S0907444909052925 [PubMed: 20124702] 49. Emsley P, Cowtan K. Coot: model-building tools for molecular graphics. Acta Crystallogr D Biol Crystallogr. 2004; 60:2126–2132.10.1107/S0907444904019158 [PubMed: 15572765] 50. Afonine PV, Moriarty NW, Mustyakimov M, Sobolev OV, Terwilliger TC, Turk D, et al. FEM: feature-enhanced map. Acta Crystallogr D Biol Crystallogr. 2015; 71:646–666.10.1107/ S1399004714028132 [PubMed: 25760612] 51. Krieger E, Vriend G. YASARA View—molecular graphics for all devices—from smartphones to workstations. Bioinformatics. 2014; 30:2981–2982.10.1093/bioinformatics/btu426 [PubMed: 24996895] 52. Duan Y, Wu C, Chowdhury S, Lee MC, Xiong G, Zhang W, et al. A point-charge force field for molecular mechanics simulations of proteins based on condensed-phase quantum mechanical calculations. J Comput Chem. 2003; 24:1999–2012.10.1002/jcc.10349 [PubMed: 14531054] 53. Krissinel E, Henrick K. Inference of macromolecular assemblies from crystalline state. J Mol Biol. 2007; 372:774–797.10.1016/j.jmb.2007.05.022 [PubMed: 17681537]
Author Manuscript Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 14
Author Manuscript
Highlights 1.
We describe the first application of crystal structure information to provide pharmacogenetic insight into a therapeutic T-cell receptor antibody.
2.
The crystal structure reveals a new binding mode responsible for the subnanomolar affinity of a TCR-like antibody for an HLA receptor-peptide complex, previously shown to be potent in mouse cancer models.
3.
Structural information was used for the first time to predict compatibility of a cancer immunotherapy with multiple HLA types.
4.
We describe a structure-based strategy to predict off-target binding which may lead to toxicity.
Author Manuscript Author Manuscript Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 15
Author Manuscript Author Manuscript Author Manuscript
Fig. 1. Overall complex structure
Author Manuscript
(A) The cartoon backbone rendering of the ESK1/pMHC complex is shown. HLAA*02:01and β2-microglobulin are colored blue and magenta. HLA helices α1 and α2 in the RMF binding site are red. The light and heavy antibody Fab fragments are cyan and orange respectively and are labeled by their variable (VL, VH,) or constant (CL, CH) domains. (B) The surface of the variable domains of the antibody is shown where the heavy chain contact surface is colored olive and the light chain contact surface is lime. (C) The top down view of the HLA surface is shown with the HLA contact surface colored yellow and the RMF contact surface colored pink. (D) The variable loop regions of the light and heavy chains are shown at the antigen binding site. VL loops L1, L2, and L3 are colored blue, purple, and magenta, respectively. VH loops H1, H2, and H3 are colored yellow, red, and green respectively. (E) The top-down view shows the RMF binding site with the variable loops circled.
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 16
Author Manuscript Author Manuscript
Fig. 2. Cation-π interaction network at binding interface
The high affinity-binding site features two networks of cation-π interactions at the interface between ESK1 and HLA-A*02:01. Tyr101 and Tyr105 from ESK1 CDR-H3 (orange), straddle HLA (blue) Arg170. HLA Arg108 and Arg169 straddle ESK1 CDR-H3 Tyr55.
Author Manuscript Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 17
Author Manuscript Author Manuscript Author Manuscript Fig. 3. Structural comparison between ESK1 and other TCRs and TCR-like antibodies
Author Manuscript
(A) Structural alignment between ESK1 and twenty TCR/HLA-A*02:01/peptide structures shows that ESK1 CDR loops reach regions of HLA α2 that TCR loops do not. This includes the circled arginine rich region: Arg108, Arg169, and Arg170. The blue ESK1/pHLA structure has VH domains H1, H2, and H3, colored yellow, red, and green respectively. (B) Structural alignment of TCR-mimic-antibody/HLA-A*02 structures shows that antibodies are more diverse in binding position and angle than TCRs, though none bind to HLA via the arginine rich ESK1 binding site. (C–F) The footprints of three TCRs and ESK1 show that TCRs bind to the mid-helical region, in contact with the peptide, whereas ESK1 reaches beyond the canonical TCR binding site.
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 18
Author Manuscript Author Manuscript Author Manuscript
Fig. 4. Peptide interactions
Author Manuscript
(A) RMF binds HLA with Met2, Phe3, Pro7, and Leu9 facing the HLA receptor and Arg1, Pro4, Asn5, and Tyr8 facing solvent and/or antibody. The heavy and light chains are in orange and cyan. Arg1, Pro4 and Asn5 are within van der Waals distance to ESK1. (B) RMF Arg1 is located near hydrogen bond distance of Ser102. Arg1 also forms potential hydrogen bonds to HLA Thr163 and Trp167 and is within van der Waals distance to ESK1 Tyr104. Also shown is the proximity of RMF Pro4 to ESK1 light chain Asp94. (C) Binding avidity of WT1 substituted peptides relative to RMF, as determined by flow cytometry. EW is a peptide (QLQNPSYDK) unrelated to RMF, isolated from Ewing sarcoma, used here as a negative control. Results were normalized to RMF binding. Scrambled corresponds to a positive control peptide which binds HLA-A*02 but not ESK1. O corresponds to ornithine and Ct to citrulline; all other amino acid abbreviations are standard.
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 19
Author Manuscript Author Manuscript Author Manuscript Fig. 5. Different binding modes of ESK1 and Hyb3 to HLA-peptide complexes
ESK1 (blue) and Hyb3 (green) differ in binding orientation to HLA-peptide complexes by 58°. ESK1 contacts the N-terminal portion of the antigen peptide. Hyb3 contacts the Cterminal portion of the antigen peptide.
Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 20
Author Manuscript Author Manuscript
Fig. 6. Mapped HLA-A*02 subtype differences
Mapped positional differences between the subtypes are colored red: A*02:02, orange: A*02:03, green: A*02:05, blue: A*02:06, magenta: A*02:07, and yellow: A*02:11. The ESK1 footprint is shown in light blue. None of the residues that vary between A*02 subtypes contacts ESK1.
Author Manuscript Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 21
Author Manuscript Author Manuscript
Fig. 7. Validation of predictions of potential ESK1 off-target epitopes
ESK1 binding was measured by flow cytometry. Relative binding was normalized to 25 μg/ml RMF. Relative binding of a scrambled control peptide or two identified peptides derived from MED13L and PIGQ are shown.
Author Manuscript Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 22
Table 1
X-ray diffraction data collection and refinement statistics
Author Manuscript
Data were collected at the Stanford Synchrotron Radiation Lightsource Beamline 12-2 on a Pilatus detector with wavelength 0.979500 Å.
Author Manuscript
Space group
P212121
Cell dimensions a, b, c (Å)
70.1, 118.3, 126.3
Resolution (Å)
37.6 – 3.05
Rsym*
0.158 (1.11)
I/σ*
13.9 (2.1)
CC½
99.8 (77.6)
Completeness (%)*
99.2 (96.6)
Redundancy*
11.8 (10.3)
Rwork/Rfree
0.198 / 0.254
Average B factor
74.8
RMS deviation, bond lengths (Å)
0.004
RMS deviation, bond angles (°)
0.74
*
Highest resolution shell is shown in parenthesis.
Author Manuscript Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Ataie et al.
Page 23
Table 2
Comparison of TCR interactions
Author Manuscript
A comparison between interface interactions of ESK1/pHLA-A*02:01 and three TCR/HLA-A*02:01 structures showing the contact surface area and binding free energy calculated from experimental dissociation constants. ESK1 has the strongest binding free energy despite having a total contact surface area comparable to TCRs.
TCR/HLA-A*02/peptide
Author Manuscript
MHC/TCR contact surface Å2 (PISA)
Peptide/TCR contact surface Å2 (PISA)
Total TCR contact surface Å2 (PISA)
ΔG (kcal/mol)
555 (α) + 198 (β) = 753
182 (α) + 214 (β) = 396
1149
−7.7
548 (α) + 171 (β) = 719
164 (α) + 173 (β) = 337
1056
−8.0
206 (α) + 435 (β) = 641
80 (α) + 199 (β) = 279
920
−7.1
585 (h)+ 305 (I)= 890
70 (h)+ 90 (I) = 160
1050
−14
Author Manuscript Author Manuscript J Mol Biol. Author manuscript; available in PMC 2017 January 16.
Author Manuscript Table 3
Author Manuscript
Author Manuscript
1.40E-02
1.06E-02
5.79E-03
1.91E-03
1.50E-03
9.30E-04
A*02:03
A*02:05
A*02:06
A*02:07
A*02:11
1.35E-02
A*02:01
A*02:02
kd [1/s]
Subtype
1.80E-05
3.00E-05
2.30E-05
1.10E-04
2.70E-04
4.10E-04
6.70E-04
Error in kd [1/s]
8.40E+05
9.53E+05
9.32E+05
1.03E+06
5.24E+05
5.91E+05
1.02E+06
ka [1/Ms]
5.50E+03
1.00E+04
7.50E+03
1.90E+04
3.10E+04
4.30E+04
4.80E+04
Error in ka [1/Ms]
1.1
1.6
2.1
5.6
20.3
23.6
13.2
KD [nM]
0.02
0.04
0.03
0.2
1.3
1.9
0.90
Error in KD [nM]
14.0
14.0
12.2
10.9
17.5
9.2
10.6
Chi2 [RU2]
145.1
92.8
124.3
199
61.4
42.2
115.8
Rmax [RU2]
simulate the bimolecular mechanism in its entirety. Chi2 and Rmax were calculated using the BIAcore X100 evaluation software.
Binding parameters were measured by surface plasmon resonance on a Biacore X100. Despite the presence of a bivalent analyte, the 1:1 binding model was used since the purpose of this section is to provide a quantitative comparison between the binding between the HLA-A*02 variants, rather than to
Author Manuscript
Characterization of ESK1 binding to HLA-A*02 variants with RMF
Ataie et al. Page 24
J Mol Biol. Author manuscript; available in PMC 2017 January 16.
