Synergy between trastuzumab and pertuzumab for human epidermal growth factor 2 (Her2) from colocalization: an in silicobased mechanism
- 14k Downloads
Human epidermal growth factor 2 (Her2), a receptor tyrosine kinase, is overexpressed in breast cancers. It has been successfully targeted by small molecule kinase inhibitors and by antibodies. Recent clinical data show a synergistic response in patients when two antibodies, trastuzumab and pertuzumab, are given in combination.
This unexpected effect is rationalized through computer models and molecular dynamic simulations by hypothesizing that the two antibodies can co-localize on the same molecule of the Her2 extracellular domain.
Simulations suggest that the clinical synergism observed for the two antibodies arises partly from enhanced affinity that originates in cooperative interactions between these two antibodies when they are co-localized on Her2 and "clamp" it; this may inhibit dimerization and possibly higher oligomerizations with neighboring receptors. In the presence of trastuzumab, the receptor becomes highly plastic, especially domains I to III, and this appears to promote increased association with pertuzumab. Further, the presence of pertuzumab evokes novel interactions between the receptor and trastuzumab. Indeed, splicing out of this region in silico results in a big reduction in the interactions of the antibody with the receptor.
If validated, these findings will bring about a new direction in the design of antibodies whereby different epitopes on the same antibody may be targeted to lead to synergistic/cooperative inhibition and contribute to generate more potent therapeutics and to increase clinical efficacy.
KeywordsTrastuzumab Molecular Dynamic Simulation Lapatinib Ternary Complex Pertuzumab
binding free energy
van der Waals energy
complementary determining regions
epidermal growth factor receptor
fragment antigen binding
human epidermal growth factor 2
insulin-like growth factor
linear combination of pairwise overlaps
molecular mechanics/Generalized-Born surface Area
Principal component analysis
particle mesh Ewald
receptor tyrosine kinase
solvent-accessible surface area
We have used molecular modelling to develop a mechanism that may partly explain this cooperative effect. A simple model is that cooperativity could arise from the physical co-localization of the two antibodies on the same Her2 molecule, as has been previously demonstrated for the insulin-like growth factor receptor . Models of the binary structures of Her2:trastuzumab and Her2:pertuzumab were combined to create a model of the ternary complex of Her2 and the two Fab fragments of these antibodies. This system was subjected to structural/energetic refinement followed by molecular dynamics (MD) simulations (Additional file 1).
Building the models
The binary associations of Her2 and the two antibodies have been experimentally characterized and are available in the Protein Data Base (pdb entries: 1n8z and 1s78 ). The electronic density of the C-terminal region of the receptor is the Her2:pertuzumab complex is missing, due probably to the high inherent flexibility of this part of the structure. For the coordinates of the residues in this region, we have assumed that this stretch will adopt a conformation similar to the one found in the complex with trastuzumab and have been so modeled. A few other missing residues, located in loops, have been modeled based on homologue structures. The ternary model of the complex Her2 and the two Fab domains of these antibodies, trastuzumab and pertuzumab, were manually constructed using the coordinates of the previous models, after superposition on the backbone of the receptor ectodomain.
MD simulations and binding energy calculations
We have used seven different systems for this study (Additional file 1); the complete 3D model structure of the complex (Her2-P-T), the receptor bound to trastuzumab (Her2-T) and pertuzumab (Her2-P), the apo-receptor (Her2), the two independent antibodies (pert and trast) and both antibodies co-localized as they are found in the complex (Abs). All these systems have been refined using several energy minimization and short MD simulation cycles in order to equilibrate the structures after which they were subjected to the production phases of MD simulation (around 18 to 20 ns each) in explicit water, as detailed. The AMBER (Assisted Model Building with Energy Refinement) 9 package  together with the ff99 force field  was used. The protonation states of the residues were assigned using PDB2PQR , after which the systems were surrounded with a box of TIP3 or Transferrable Interaction Potential 3 water molecules that result in an 8 Å layer of water molecules around the proteins and neutralized by the necessary numbers of Cl-or Na+ counter-ions using the t-leap module of AMBER 9. The solvated systems were then subjected to a minimization phase, mainly by the steepest descend method and small number of conjugate gradient steps, in order to remove bad contacts. After that a 50 ps heating phase was carried out, where the proteins were heated from 0 to 300 K, followed by another 50 ps phase density equilibration at a constant pressure of 1 atm. A final 2 ns unconstrained MD simulation was carried out. The production runs were extended to around 20 ns at constant temperature using a Langevin thermostat  and constant pressure (1 atm) by coupling to a Berendsen barostat These two queries do not need to be address since these are mathematical methods and not devices . The particle mesh Ewald (PME) method  was used with a direct-space non-bonded cutoff of 8 Å to account for long range interactions. Covalent bond lengths to hydrogen atoms were constrained using the SHAKE algorithm , which allowed a time step of 2 fs for all the MD simulations. Snapshots in the interval of 8 to 16 ns were extracted for further processing. The analysis was carried out using functionalities in the ptraj module of AMBER 9, the module Bio3D  of R and Dyndom . The PyMOL molecular viewer, a software package  was used for visualizing and generating the figures.
The binding free energy ΔGbind of the different systems was calculated according to the so-called Molecular Mechanics/Generalized-Born Surface Area (MM/GBSA) procedure . Here ΔGbind denotes the contributions from the molecular mechanics (MM) energies of the molecules, and originates in contributions from internal energies, electrostatics (ΔGELE) and van der Waals (ΔGVDW) contributions. These terms have been calculated using the mm_pbsa module of AMBER 9. The polar component was evaluated with the Generalized-Born (GB) approach. In this approach, the non-polar contribution to the solvation energy was calculated as ΔGNP = γ (SASA) + β, in which γ = 0.00542 kcal/Å, β = 0.92 kcal/mol and SASA is the solvent-accessible surface area estimated with LCPO (linear combination of pairwise overlaps) method  as implemented in AMBER. While the real free energy of a system also includes contributions from the entropies, given the large sizes of the systems studied here, computations of entropies were prohibitive and hence could not be carried out. However, the free energies computed here have been averaged over several conformations generated through MD simulations, and hence the effects of entropy to some extent are implicitly included in the current calculations.
Essential dynamics analysis and Dyndom analysis of Her2 and its antibodies
Essential dynamics [27, 28] have been widely used for filtering large scale concerted fluctuations from an ensemble of conformations, like those obtained from MD simulations . The method is comparable to a principal component analysis, and it is based on the construction of the diagonalized covariance matrix containing the atomic fluctuations. We have performed a standard essential dynamics (or Principal Component Analysis, PCA) analysis in order to identify the most relevant motions occurring in the receptor in the different bound states considered on the MD simulations carried out in this study (Additional file 2); all the covariance matrices have been calculated taking the apo receptor as a reference structure. This yields a set of eigenvectors in a 3N-dimensional space, and eigenvalues, reflecting the magnitude of the fluctuations along the corresponding eigenvectors. Each eigenvector describing a collective motion can be represented by two structures, one at the maximum value, the other at the minimum value of the collective variable when the trajectory is projected onto it. These two structures thus generated for the first eigenvector of each trajectory were used as input for the program DynDom , using default parameters, in order to determine dynamic domains, hinge axes and hinge-bending regions.
Results and discussion
Conformational space of Her2 in different bound forms
Simulations show that the ectodomain of Her2 adopts various conformations, depending on its bound state. The binary states (when complexed to trastuzumab or pertuzumab) are more stable compared to the apo or the ternary complex (Additional file 3). This could stem from the high intrinsic flexibility of the apo receptor or/and the ternary complex. Although a complete understanding of the behaviour of such a receptor demands the full-length structure, we believe that the models used here represent valid starting structures for the type of analyses carried out in this study. As expected, the presence of an antibody immobilizes its corresponding epitope (Additional file 4). The largest fluctuations in the pertuzumab binding region (ranging from 240 to 300) and in the trastuzumab epitope (around 550 to 600) are seen in the apo state or in the Her2:trastuzumab or Her2:pertuzumab complexes respectively. Interestingly the dimerization arm, a β hairpin in domain II that protrudes from the rest of Her2, together with the spatially adjacent regions, undergoes enhanced fluctuations when trastuzumab binds. This may reinforce the fact that the Her2:trastuzumab complex can actually promote the association of Her2 with other members of the ErbB family.
We have also investigated the flexibility of the antibodies and the stability of their binding interfaces with Her2. In general, the complementary determining regions (CDR) exhibit smaller fluctuations in comparison to the rest of the antibody; whereas regions far from the receptor's influence remain extremely flexible. In the ternary complex, the C-terminal half of the light chain of pertuzumab shows an increased flexibility, induced by the presence of trastuzumab (Additional file 4). Trastuzumab shows a more complicated pattern. When only trastuzumab is present in the complex, the whole antibody shows a high flexibility, with the exception of the CDR for the light chain. In the ternary complex, the light chain undergoes the largest fluctuations in the region interacting with the receptor binding epitope; however the most relevant difference is an attenuation in the mobility of the N-terminal half of the heavy chain due to the proximity of one of the disulphide-bonded modules of domain II, which in turn shows lower flexibility (residues 310 to 325; see section below on new epitopes). We can conclude that the dynamics of pertuzumab seem to be regulated mainly by the binding to the receptor; while for trastuzumab, the co-binding of the second antibody has a significant effect. When the two antibodies are analysed in the absence of the receptor, trastuzumab per se seems more flexible than pertuzumab, which can be the origin of the more complex dynamics observed for the systems including this antibody. These intricate changes and regulation between the different components of the system reflect the importance of flexibility and could point to "allosteric" changes in explaining the in vivo synergistic effect.
Uncovering new epitopes in Her2
The analysis of the interactions between Her2 and the two antibodies reveals that the perturbation of the crystallographically observed interfacial region of Her2 is small during the simulations of the binary complexes (Additional file 5). However, the co-localization of the two Abs on Her2 surprisingly triggers the formation of additional and novel contacts of Her2 with trastuzumab; these lie outside of its epitopes (Figure 1). This β-hairpin loop region, termed the pertuzumab-induced trastuzumab epitope, is located in domain II and is in proximity to the dimerization arm. The requirement of the co-localization of the two Abs for the formation of this new epitope is revealed in the minimum distance matrices for all the Cα atoms for the receptor and trastuzumab in the simulations of the binary and ternary complexes (see Figure 1C upper and lower diagonal of the matrix, respectively). This can also be observed from the reported distances between two residues in this novel epitope and residues at the interface in trastuzumab (Additional file 6). For this epitope to emerge in the Her2:trastuzumab interface, the presence of pertuzumab is strictly required. This novel association is mainly driven by long-range interactions, and could be the origin of the observed synergism.
Quantification of synergism between pertuzumab and trastuzumab
In silico binding free energy for the association of Her2 ectodomain and the two antibodies.
Her2-T < -Her2 + trast
Her2-P < -Her2 + pert
Her2-P-T < -Her2 + trast + pert
Her2-P-T < -Her2 + Abs
Her2-trunc-P-T < -Her2-trunc + trast + pert
- trastuzumab has a higher affinity for apo Her2 than pertuzumab (Figure 2A); this arises from stronger electrostatic and van der Waals interactions between trastuzumab and Her2 (the net charge on Her2 is -15, while on trastuzumab and pertuzumab it is +6 and +5 respectively); the buried area at the interface is approximately 1,675 and approximately 1,530 Å2 for trastuzumab and pertuzumab, respectively
- a marked gain in affinity when both Abs are bound simultaneously to the receptor (synergism). This effect arises mainly from the interactions between these two antibodies upon co-localization, with a difference of approximately 27 kcal/mol between isolated and simultaneously bound antibodies (Figure 2C).
Abs-induced bending of the Her2 ectodomain
Principal Component Analysis (Additional file 2) combined with dynamical domain motion analysis using DynDom suggests that domains I to III (residues 1 to 480) generally behave as a rigid body, with hinge regions varying in their location in domain IV, depending on the binding partner of the receptor (Additional files 7 and 8). Two different motions are detected in the apo state of Her2; the first one involves a shearing motion between the stalk portion of the receptor and the large rigid-body unit; whereas the second one corresponds to a hinge motion within domain IV. The overall change in conformation involves a large conformational space scanned by the dimerization arm with respect to the membrane surface (with an angle of approximately 50° for the parallel and approximately 78° for the motion perpendicular to the membrane), which represents periscope-like movements that this receptor might execute in search of binding partners. In the case of the ternary complex, the binding of trastuzumab causes a shift in the hinge region along domain IV. This is associated with small swivelling motions across an axis orthogonal to the membrane; if the receptor lies parallel to the membrane  then this would represent "out of the membrane" motions. The dimerization arm spans approximately 16 Å in comparison to the apo form which covers approximately 33 Å (measured from the Cα of the atom (Pro252) at the base of this motif). In contrast, in the binary states, the pattern of motion differs: domain IV behaves as a rigid body and only one hinge region is apparent. When trastuzumab is bound, a hinge motion is observed (18°) between the bulky domain I-to-III and domain IV (crystallographic data show approximately 7° in the angle subtended between domains III and IV); the Her2:pertuzumab complex undergoes more complex motions. The deletion in the truncated mutant results in attenuation of two-body rotations along the parallel axis (from 48° in the "full-length" receptor to approximately 28° in the truncated form). In contrast, the bound form of this truncated receptor undergoes a hinged motion spanning approximately 20°.
The overall comparison of the conformations that Her2 ectodomain adopts in the different conditions analysed here indicates that domains I, II and III adopt a relatively fixed interdomain relationship, but their relative orientations with respect to domain IV varies considerably (up to approximately 80°), and appear to be pivoted across the region 504 to 532 that lies between domains III and IV. This is indicative of the plasticity inherent in these receptors and suggests that the extracellular domain can swivel quite considerably across domain IV that itself is constrained by being embedded partly in the membrane.
In summary, MD simulations suggest that the clinical synergism observed for the two antibodies arises partly from an enhanced affinity that originates in cooperative interactions between these two antibodies when they are co-localized on Her2 and "clamp" it; this may inhibit dimerization and possibly higher oligomerizations with neighbouring receptors. In the presence of trastuzumab, the receptor becomes highly plastic, especially domains I-III, and this appears to promote increased association with pertuzumab. While combining antibodies is a feature that increasingly appears to have major effects in therapy (see, for example, the work of Chao and co-workers , which combines two different antibodies against two different cell surface molecules for treatment of non-Hodgkin's lymphoma), a model incorporating the idea that two antibodies bind to the same receptor, albeit at distinct epitopes, is relatively new; indeed there has been only one report of such a feature as highlighted for targeting the IGF-1 receptor . These findings could be tested, for example, in 2D and 3D preclinical cell culture models to explore trastuzumab efficacy [36, 37]. While we await experimental investigations of this novel mechanism in Her2, if validated, it offers routes towards the design of new antibodies whereby combinations of trastuzumab and pertuzumab (or parts thereof) may be constructed that have the potential to generate therapeutics with much higher affinities and hopefully efficacies. The technology of generating hybrid antibodies or bi-specific antibodies  is known; however, this technique has been used for linking two antibodies that target different molecules. In contrast our hypotheses open the avenue for linking two antibodies that target distinct epitopes on the same molecule.
CSV is adjunct at DBS (NUS) and SBS (NTU); Grant from European Research Council (AdG09 250244) is acknowledged.
- 4.Garrett TP, McKern NM, Lou M, Elleman TC, Adams TE, Lovrecz GO, Kofler M, Jorissen RN, Nice EC, Burgess AW, Ward CW: The crystal structure of a truncated ErbB2 ectodomain reveals an active conformation, poised to interact with other ErbB receptors. Mol Cell. 2003, 11: 495-505. 10.1016/S1097-2765(03)00048-0.CrossRefPubMedGoogle Scholar
- 12.Baselga J, Gelmon KA, Verma S, Wardley A, Conte P, Miles D, Bianchi G, Cortes J, McNally VA, Ross GA, Fumoleau P, Gianni L: Phase II trial of pertuzumab and trastuzumab in patients with human epidermal growth factor receptor 2-positive metastatic breast cancer that progressed during prior trastuzumab therapy. J Clin Oncol. 2010, 28: 1138-1144. 10.1200/JCO.2009.24.2024.CrossRefPubMedPubMedCentralGoogle Scholar
- 13.Dong J, Demarest SJ, Sereno A, Tamraz S, Langley E, Doern A, Snipas T, Perron K, Joseph I, Glaser SM, Ho SN, Reff ME, Hariharan K: Combination of two insulin-like growth factor-I receptor inhibitory antibodies targeting distinct epitopes leads to an enhanced antitumor response. Mol Cancer Ther. 2010, 9: 2593-2604. 10.1158/1535-7163.MCT-09-1018.CrossRefPubMedGoogle Scholar
- 15.Case DA, Darden TA, Cheatham TE, Simmerling CL, Wang J, Duke RE, Luo R, Merz KM, Pearlman DA, Crowley M, Walker RC, Zhang W, Wang B, Havik S, Riotberg A, Seabra G, Wong KF, Paesani F, Wu X, Brozell S, Tsui V, Gohlke H, Yang L, Tan C, Mongan J, Hornak V, Cui G, Beroza P, Mathews DH, Schafmeister C, et al: AMBER 9. 2006, University of California, San FranciscoGoogle Scholar
- 24.PyMOL. DeLano Scientific, Palo Alto, CA, [http://www.pymol.org]
- 25.Kollman PA, Massova I, Reyes C, Kuhn B, Huo S, Chong L, Lee M, Lee T, Duan Y, Wang W, Donini O, Cieplak P, Srinivasan J, Case DA, Cheatham TE: Calculating structures and free energies of complex molecules: combining molecular mechanics and continuum models. Acc Chem Res. 2000, 33: 889-897. 10.1021/ar000033j.CrossRefPubMedGoogle Scholar
- 35.Chao MP, Alizadeh AA, Tang C, Myklebust JH, Varghese B, Gill S, Jan M, Cha AC, Chan CK, Tan BT, Park CY, Zhao F, Kohrt HE, Malumbres R, Briones J, Gascoyne RD, Lossos IS, Levy R, Weissman IL, Majeti R: Anti-CD47 antibody synergizes with rituximab to promote phagocytosis and eradicate non-Hodgkin lymphoma. Cell. 2010, 142: 699-713. 10.1016/j.cell.2010.07.044.CrossRefPubMedPubMedCentralGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.