Prediction of polyspecificity from antibody sequence data by machine learning.

Antibodies are generated with great diversity in nature resulting in a set of molecules, each optimized to bind a specific target. Taking advantage of their diversity and specificity, antibodies make up for a large part of recently developed biologic drugs. For therapeutic use antibodies need to fulfill several criteria to be safe and efficient. Polyspecific antibodies can bind structurally unrelated molecules in addition to their main target, which can lead to side effects and decreased efficacy in a therapeutic setting, for example via reduction of effective drug levels. Therefore, we created a neural-network-based model to predict polyspecificity of antibodies using the heavy chain variable region sequence as input. We devised a strategy for enriching antibodies from an immunization campaign either for antigen-specific or polyspecific binding properties, followed by generation of a large sequencing data set for training and cross-validation of the model. We identified important physico-chemical features influencing polyspecificity by investigating the behaviour of this model. This work is a machine-learning-based approach to polyspecificity prediction and, besides increasing our understanding of polyspecificity, it might contribute to therapeutic antibody development.

CAR-J cells for antibody discovery and lead optimization of TCR-like immunoglobulins.

T-cell bispecific antibodies (TCBs) are a novel class of engineered immunoglobulins that unite monovalent binding to the T-cell receptor (TCR) CD3e chain and bivalent binding to tumor-associated antigens in order to recruit and activate T-cells for tumor cell killing. In vivo, T-cell activation is usually initiated via the interaction of the TCR with the peptide-HLA complex formed by the human leukocyte antigen (HLA) and peptides derived from intracellular proteins. TCR-like antibodies (TCRLs) that recognize pHLA-epitopes extend the target space of TCBs to peptides derived from intracellular proteins, such as those overexpressed during oncogenesis or created via mutations found in cancer. One challenge during lead identification of TCRL-TCBs is to identify TCRLs that specifically, and ideally exclusively, recognize the desired pHLA, but not unrelated pHLAs. In order to identify TCRLs suitable for TCRL-TCBs, large numbers of TCRLs have to be tested in the TCB format. Here, we propose a novel approach using chimeric antigen receptors (CARs) to facilitate the identification of highly selective TCRLs. In this new so-called TCRL-CAR-J approach, TCRL-candidates are transduced as CARs into Jurkat reporter-cells, and subsequently assessed for their specificity profile. This work demonstrates that the CAR-J reporter-cell assay can be applied to predict the profile of TCRL-TCBs without the need to produce each candidate in the final TCB format. It is therefore useful in streamlining the identification of TCRL-TCBs.

Rabbits transgenic for human IgG genes recapitulating rabbit B-cell biology to generate human antibodies of high specificity and affinity.

Transgenic animals incorporating human antibody genes are extremely attractive for drug development because they obviate subsequent antibody humanization procedures required for therapeutic translation. Transgenic platforms have previously been established using mice, but also more recently rats, chickens, and cows and are now in abundant use for drug development. However, rabbit-based antibody generation, with a strong track record for specificity and affinity, is able to include gene conversion mediated sequence diversification, thereby enhancing binder maturation and improving the variance/selection of output antibodies in a different way than in rodents. Since it additionally frequently permits good binder generation against antigens that are only weakly immunogenic in other organisms, it is a highly interesting species for therapeutic antibody generation. We report here on the generation, utilization, and analysis of the first transgenic rabbit strain for human antibody production. Through the knockout of endogenous IgM genes and the introduction of human immunoglobulin sequences, this rabbit strain has been engineered to generate a highly diverse human IgG antibody repertoire. We further incorporated human CD79a/b and Bcl2 (B-cell lymphoma 2) genes, which enhance B-cell receptor expression and B-cell survival. Following immunization against the angiogenic factor BMP9 (Bone Morphogenetic Proteins 9), we were able to isolate a set of exquisitely affine and specific neutralizing antibodies from these rabbits. Sequence analysis of these binders revealed that both somatic hypermutation and gene conversion are fully operational in this strain, without compromising the very high degree of humanness. This powerful new transgenic strategy will allow further expansion of the use of endogenous immune mechanisms in drug development.

Characterization and prediction of positional 4-hydroxyproline and sulfotyrosine, two post-translational modifications that can occur at substantial levels in CHO cells-expressed biotherapeutics.

Biotherapeutics may contain a multitude of different post-translational modifications (PTMs) that need to be assessed and possibly monitored and controlled to ensure reproducible product quality. During early development of biotherapeutics, unexpected PTMs might be prevented by in silico identification and characterization together with further molecular engineering. Mass determinations of a human IgG1 (mAb1) and a bispecific IgG-ligand fusion protein (BsAbA) demonstrated the presence of unusual PTMs resulting in major +80 Da, and +16/+32 Da chain variants, respectively. For mAb1, analytical cation exchange chromatography demonstrated the presence of an acidic peak accounting for 20%. A + 79.957 Da modification was localized within the light chain complementarity-determining region-2 and identified as a sulfation based on accurate mass, isotopic distribution, and a complete neutral loss reaction upon collision-induced dissociation. Top-down ultrahigh resolution MALDI-ISD FT-ICR MS of modified and unmodified Fabs allowed the allocation of the sulfation to a specific Tyr residue. An aspartate in amino-terminal position-3 relative to the affected Tyr was found to play a key role in determining the sulfation. For BsAbA, a + 15.995 Da modification was observed and localized to three specific Pro residues explaining the +16 Da chain A, and +16 Da and +32 Da chain B variants. The BsAbA modifications were verified as 4-hydroxyproline and not 3-hydroxyproline in a tryptic peptide map via co-chromatography with synthetic peptides containing the two isomeric forms. Finally, our approach for an alert system based on in-house in silico predictors is presented. This system is designed to prevent these PTMs by molecular design and engineering during early biotherapeutic development.

Sphinx: merging knowledge-based and ab initio approaches to improve protein loop prediction.

Motivation: Loops are often vital for protein function, however, their irregular structures make them difficult to model accurately. Current loop modelling algorithms can mostly be divided into two categories: knowledge-based, where databases of fragments are searched to find suitable conformations and ab initio, where conformations are generated computationally. Existing knowledge-based methods only use fragments that are the same length as the target, even though loops of slightly different lengths may adopt similar conformations. Here, we present a novel method, Sphinx, which combines ab initio techniques with the potential extra structural information contained within loops of a different length to improve structure prediction. Results: We show that Sphinx is able to generate high-accuracy predictions and decoy sets enriched with near-native loop conformations, performing better than the ab initio algorithm on which it is based. In addition, it is able to provide predictions for every target, unlike some knowledge-based methods. Sphinx can be used successfully for the difficult problem of antibody H3 prediction, outperforming RosettaAntibody, one of the leading H3-specific ab initio methods, both in accuracy and speed. Availability and Implementation: Sphinx is available at http://opig.stats.ox.ac.uk/webapps/sphinx. Contact: deane@stats.ox.ac.uk. Supplementary information: Supplementary data are available at Bioinformatics online.

Length-independent structural similarities enrich the antibody CDR canonical class model.

Complementarity-determining regions (CDRs) are antibody loops that make up the antigen binding site. Here, we show that all CDR types have structurally similar loops of different lengths. Based on these findings, we created length-independent canonical classes for the non-H3 CDRs. Our length variable structural clusters show strong sequence patterns suggesting either that they evolved from the same original structure or result from some form of convergence. We find that our length-independent method not only clusters a larger number of CDRs, but also predicts canonical class from sequence better than the standard length-dependent approach. To demonstrate the usefulness of our findings, we predicted cluster membership of CDR-L3 sequences from 3 next-generation sequencing datasets of the antibody repertoire (over 1,000,000 sequences). Using the length-independent clusters, we can structurally classify an additional 135,000 sequences, which represents a ∼20% improvement over the standard approach. This suggests that our length-independent canonical classes might be a highly prevalent feature of antibody space, and could substantially improve our ability to accurately predict the structure of novel CDRs identified by next-generation sequencing.

MoFvAb: Modeling the Fv region of antibodies.

Knowledge of the 3-dimensional structure of the antigen-binding region of antibodies enables numerous useful applications regarding the design and development of antibody-based drugs. We present a knowledge-based antibody structure prediction methodology that incorporates concepts that have arisen from an applied antibody engineering environment. The protocol exploits the rich and continuously growing supply of experimentally derived antibody structures available to predict CDR loop conformations and the packing of heavy and light chain quickly and without user intervention. The homology models are refined by a novel antibody-specific approach to adapt and rearrange sidechains based on their chemical environment. The method achieves very competitive all-atom root mean square deviation values in the order of 1.5 Å on different evaluation datasets consisting of both known and previously unpublished antibody crystal structures.

Structure-based prediction of asparagine and aspartate degradation sites in antibody variable regions.

Monoclonal antibodies (mAbs) and proteins containing antibody domains are the most prevalent class of biotherapeutics in diverse indication areas. Today, established techniques such as immunization or phage display allow for an efficient generation of new mAbs. Besides functional properties, the stability of future therapeutic mAbs is a key selection criterion which is essential for the development of a drug candidate into a marketed product. Therapeutic proteins may degrade via asparagine (Asn) deamidation and aspartate (Asp) isomerization, but the factors responsible for such degradation remain poorly understood. We studied the structural properties of a large, uniform dataset of Asn and Asp residues in the variable domains of antibodies. Their structural parameters were correlated with the degradation propensities measured by mass spectrometry. We show that degradation hotspots can be characterized by their conformational flexibility, the size of the C-terminally flanking amino acid residue, and secondary structural parameters. From these results we derive an accurate in silico prediction method for the degradation propensity of both Asn and Asp residues in the complementarity-determining regions (CDRs) of mAbs.

A robust high throughput platform to generate functional recombinant monoclonal antibodies using rabbit B cells from peripheral blood.

We have developed a robust platform to generate and functionally characterize rabbit-derived antibodies using B cells from peripheral blood. The rapid high throughput procedure generates a diverse set of antibodies, yet requires only few animals to be immunized without the need to sacrifice them. The workflow includes (i) the identification and isolation of single B cells from rabbit blood expressing IgG antibodies, (ii) an elaborate short term B-cell cultivation to produce sufficient monoclonal antigen specific IgG for comprehensive phenotype screens, (iii) the isolation of VH and VL coding regions via PCR from B-cell clones producing antigen specific and functional antibodies followed by the sequence determination, and (iv) the recombinant expression and purification of IgG antibodies. The fully integrated and to a large degree automated platform (demonstrated in this paper using IL1RL1 immunized rabbits) yielded clonal and very diverse IL1RL1-specific and functional IL1RL1-inhibiting rabbit antibodies. These functional IgGs from individual animals were obtained at a short time range after immunization and could be identified already during primary screening, thus substantially lowering the workload for the subsequent B-cell PCR workflow. Early availability of sequence information permits one to select early-on function- and sequence-diverse antibodies for further characterization. In summary, this powerful technology platform has proven to be an efficient and robust method for the rapid generation of antigen specific and functional monoclonal rabbit antibodies without sacrificing the immunized animal.

Differential splicing generates new transmembrane receptor and extracellular matrix-related targets for antibody-based therapy of cancer.

Alternative splicing has been shown to be deregulated in cancer and a link to growth stimulation has been established. Here we describe transmembrane and extracellular matrix-related targets generated by alternative splicing with a restricted pattern of expression in normal tissues and a deregulated pattern of expression in cancer as possible targets for therapeutic intervention with antibody-related agents. We focus on isoforms of transmembrane and extracellular matrix proteins, such as CD44, Claudin 18, L1 cell adhesion molecule and epithelial cellular adhesion molecule, fibronectin, tenascin, osteopontin and versican as well as transmembrane tyrosine kinases, such as fibroblast growth factor receptors, epidermal growth factor receptor and receptor d'origin nantais.

The Roche Cancer Genome Database 2.0.

BACKGROUND: Cancer is a disease of genome alterations that arise through the acquisition of multiple somatic DNA sequence mutations. Some of these mutations can be critical for the development of a tumor and can be useful to characterize tumor types or predict outcome. DESCRIPTION: We have constructed an integrated biological information system termed the Roche Cancer Genome Database (RCGDB) combining different human mutation databases already publicly available. This data is further extended by hand-curated information from publications.The current version of the RCGDB provides a user-friendly graphical interface that gives access to the data in different ways: (1) Single interactive search by genes, samples, cell lines, diseases, as well as pathways, (2) batch searches for genes and cell lines, (3) customized searches for regularly occurring requests, and (4) an advanced query interface enabling the user to query for samples and mutations by various filter criteria. CONCLUSION: The interfaces of the presented database enable the user to search and view mutations in an intuitive and straight-forward manner. The database is freely accessible at http://rcgdb.bioinf.uni-sb.de/MutomeWeb/.

Intracellular proteins displayed on the surface of tumor cells as targets for therapeutic intervention with antibody-related agents.

The identification of targets which are located intracellularly in normal cells and are exposed on the surface of malignant cells is an issue in the target selection process for the development of anticancer agents. Targets with these characteristics should increase the specificity of intervention and the corresponding therapeutic window. We discuss targets such as heat-shock protein 70 (HSP70) and heat-shock protein 90 (HSP90), glucose-regulated protein 78 (GRP78), actin, cytokeratins, vimentin, nucleolin, nucleosomes, estrogen receptor-alpha variant 36 (ER-α36) and feto-acinar pancreatic protein (FAPP). Involvement of these targets in cellular processes, tumor specificity and tractability with antibody-related agents, are discussed.

Interleukin 6/interleukin 6 receptor interaction and its role as a therapeutic target for treatment of cachexia and cancer.

Interleukin 6 (IL6) mediates pleiotropic physiological functions through its interaction with the IL6 receptor (IL6R). Signal transduction can occur via cis- and trans-signaling. The role of IL6/IL6R interaction via autocrine and paracrine loops in tumor proliferation and progression is discussed. The potential role of IL6/IL6R interaction in different experimental systems and tumor entities is summarized while the focus is on inhibition of IL6 signaling with monoclonal antibodies directed against IL6 or IL6R and their potential impact for treatment of tumor-associated cachexia and as antitumoral agents as monotherapy and in combination with small molecule compounds.

ALCAM/CD166: cancer-related issues.

Activated leucocyte adhesion molecule (ALCAM) was originally identified as a transmembrane receptor which is involved in T-cell activation and has other still unresolved functions in hematopoiesis, development, inflammation and transendothelial migration of neutrophils. ALCAM is a member of a subfamily of immunoglobulin receptors with five immunoglobulin-like domains (VVC2C2C2) in the extracellular domain and is expressed in many types of tumors. The tumor-type-dependent impact of its expression level with respect to prognosis points to a possible context-dependent function. Most functional investigations have focused on malignant melanoma, in which high ALCAM expression at the protein level correlates with a poor prognosis. ALCAM mediates low-affinity homophilic interactions and much stronger interactions with CD6. Modulation of ALCAM function with agents such as transfected dominant negative ALCAM and ligand-binding secreted ALCAM both lead to inhibition of matrix metalloproteinase-2 activation, but their impact with respect to invasion in vitro and metastasis in vivo are different. Single-chain Fv fragments directed against ALCAM are efficiently internalized, paving the way for exploration of immunoconjugates as therapeutic agents. Validation experiments of the target with modulatory agents for possible therapeutic application in oncology are discussed.

Human variation databases.

More than 100,000 human genetic variations have been described in various genes that are associated with a wide variety of diseases. Such data provides invaluable information for both clinical medicine and basic science. A number of locus-specific databases have been developed to exploit this huge amount of data. However, the scope, format and content of these databases differ strongly and as no standard for variation databases has yet been adopted, the way data is presented varies enormously. This review aims to give an overview of current resources for human variation data in public and commercial resources.

Cancer-related issues of CD147.

CD147 is involved in many physiological functions, such as lymphocyte responsiveness, spermatogenesis, implantation, fertilization and neurological functions at early stages of development. Here we specifically review the role of CD147 in cancer. We focus on the following aspects: expression of CD147 in malignant versus normal tissues and its possible impact on prognosis, interaction of tumor cell-expressed CD147 with stroma cells and induction of matrix metalloproteinases, as well as the role of CD147 in tumor angiogenesis. The function of CD147 in supercomplexes with monocarboxylate transporters (MCT) and amino acid transporters such as CD98hc and large neutral amino acid transporter 1 (LAT1), as well as the functional contribution of CD147 in complexes with caveolin-1 and integrins, is discussed. Target validation experiments making use of CD147-directed RNAi and monoclonal antibodies are summarized. Finally, the relevance of CD147 as a target for therapeutic intervention in cancer patients is discussed.

The Roche Cancer Genome Database (RCGDB).

Sequence variations are being studied for a better understanding of the mechanism and development of cancer as a mutation-driven disease. The systematic sequencing of genes in tumors and technological advances in high-throughput techniques combined with efficient data acquisition methods have resulted in an explosion of available cancer genome-related data. Despite the technological progress and increase of data, improvements in the application area, for example, drug target discovery, have failed to keep pace with increased research and development spending. One reason for this discrepancy is the ever increasing number of databases and the absence of a unified access to the mutation data. Currently, researchers typically have to browse several, often highly specialized databases to obtain the required information. A more complete understanding of relations and dependencies between mutations and cancer, however, requires the availability of an efficient integrative cancer genome information system. To facilitate this, we developed the Roche Cancer Genome Database (RCGDB), a freely available biological information system integrating different kinds of mutation data. The database is the first comprehensive integration of disparate cancer genome data like single nucleotide variants, single nucleotide polymorphisms, and chromosomal aberrations (CGH and FISH). RCGDB is freely accessible via a Google-like Web interface at http://rcgdb.bioinf.uni-sb.de/MutomeWeb/.

L1-CAM as a target for treatment of cancer with monoclonal antibodies.

L1 cell adhesion molecule (L1-CAM) is a neuronal adhesion molecule which is expressed in many tumor entities. L1-CAM was shown to be involved in proliferation, invasion and metastasis both in vitro and in vivo. L1-CAM is engaged in homophilic interactions and complexes with many other ligands in a context-dependent manner. Activation and modulation of the extracellular signal-related kinase pathway by L1-CAM has been documented. In normal tissues, L1-CAM expression is restricted to nerve bundles and kidney tubules; however, L1-CAM is expressed in many tumor entities and, with the exception of neuroblastoma, L1-CAM expression correlates with poor prognosis. L1-CAM occurs in two isoforms, full-length L1-CAM and a variant in which exons 2 and 27 have been deleted. Preclinical experiments with available monoclonal antibodies are summarized and L1-CAM is analysed as a target for treatment of cancer with monoclonal antibodies.