Epitope-dependent mechanisms of CD27 neutralization revealed by X-ray crystallography.

CD27 is a T and B cell co-stimulatory protein of the TNF receptor superfamily dependent on the availability of the TNF-like ligand CD70. Two anti-CD27 neutralizing monoclonal antibodies were obtained from mouse hybridoma and subsequently humanized and optimized for binding the target. The two antibodies are similar in terms of their CD27-binding affinity and ability to block NF-κB signaling, however their clearance rates in monkeys are very different. The pharmacokinetics profiles could be epitope dependent. To identify the epitopes, we determined the crystal structure of the ternary complex between CD27 and the Fab fragments of these non-competing antibodies. The structure reveals the binding modes of the antibodies suggesting that their mechanisms of action are distinctly different and provides a possible explanation of the in vivo data.

Structural diversity in a human antibody germline library.

To support antibody therapeutic development, the crystal structures of a set of 16 germline variants composed of 4 different kappa light chains paired with 4 different heavy chains have been determined. All four heavy chains of the antigen-binding fragments (Fabs) have the same complementarity-determining region (CDR) H3 that was reported in an earlier Fab structure. The structure analyses include comparisons of the overall structures, canonical structures of the CDRs and the VH:VL packing interactions. The CDR conformations for the most part are tightly clustered, especially for the ones with shorter lengths. The longer CDRs with tandem glycines or serines have more conformational diversity than the others. CDR H3, despite having the same amino acid sequence, exhibits the largest conformational diversity. About half of the structures have CDR H3 conformations similar to that of the parent; the others diverge significantly. One conclusion is that the CDR H3 conformations are influenced by both their amino acid sequence and their structural environment determined by the heavy and light chain pairing. The stem regions of 14 of the variant pairs are in the 'kinked' conformation, and only 2 are in the extended conformation. The packing of the VH and VL domains is consistent with our knowledge of antibody structure, and the tilt angles between these domains cover a range of 11 degrees. Two of 16 structures showed particularly large variations in the tilt angles when compared with the other pairings. The structures and their analyses provide a rich foundation for future antibody modeling and engineering efforts.

Epitope mapping and structural basis for the recognition of phosphorylated tau by the anti-tau antibody AT8.

Microtubule-associated protein tau becomes abnormally phosphorylated in Alzheimer's disease and other tauopathies and forms aggregates of paired helical filaments (PHF-tau). AT8 is a PHF-tau-specific monoclonal antibody that is a commonly used marker of neuropathology because of its recognition of abnormally phosphorylated tau. Previous reports described the AT8 epitope to include pS202/pT205. Our studies support and extend previous findings by also identifying pS208 as part of the binding epitope. We characterized the phosphoepitope of AT8 through both peptide binding studies and costructures with phosphopeptides. From the cocrystal structure of AT8 Fab with the diphosphorylated (pS202/pT205) peptide, it appeared that an additional phosphorylation at S208 would also be accommodated by AT8. Phosphopeptide binding studies showed that AT8 bound to the triply phosphorylated tau peptide (pS202/pT205/pS208) 30-fold stronger than to the pS202/pT205 peptide, supporting the role of pS208 in AT8 recognition. We also show that the binding kinetics of the triply phosphorylated peptide pS202/pT205/pS208 was remarkably similar to that of PHF-tau. The costructure of AT8 Fab with a pS202/pT205/pS208 peptide shows that the interaction interface involves all six CDRs and tau residues 202-209. All three phosphorylation sites are recognized by AT8, with pT205 acting as the anchor. Crystallization of the Fab/peptide complex under acidic conditions shows that CDR-L2 is prone to unfolding and precludes peptide binding, and may suggest a general instability in the antibody.

Induced conformational change in human IL-4 upon binding of a signal-neutralizing DARPin.

The crystal structure of DARPin 44C12V5 that neutralizes IL-4 signaling has been determined alone and bound to human IL-4. A significant conformational change occurs in the IL-4 upon DARPin binding. The DARPin binds to the face of IL-4 formed by the A and C α-helices. The structure of the DARPin remains virtually unchanged. The conformational changes in IL-4 include a reorientation of the C-helix Trp91 side chain and repositioning of CD-loop residue Leu96. Both side chains move by >9 Å, becoming buried in the central hydrophobic region of the IL-4:DARPin interface. This hydrophobic region is surrounded by a ring of hydrophilic interactions comprised of hydrogen bonds and salt bridges and represents a classical "hotspot." The structures also reveal how the DARPin neutralizes IL-4 signaling. Comparing the IL-4:DARPin complex structure with the structures of IL-4 bound to its receptors (Hage et al., Cell 1999; 97, 271-281; La Porte et al., Cell 2008, 132, 259-272), it is found that the DARPin binds to the same IL-4 face that interacts with the junction of the D1 and D2 domains of the IL-4Rα receptors. Signaling is blocked since IL-4 cannot bind to this receptor, which it must do first before initiating a productive receptor complex with either the IL-13α1 or the γc receptor.

Neuronal injury in the motor cortex after chronic stroke and lower limb motor impairment: a voxel-based lesion symptom mapping study.

Many studies have examined motor impairments using voxel-based lesion symptom mapping, but few are reported regarding the corresponding relationship between cerebral cortex injury and lower limb motor impairment analyzed using this technique. This study correlated neuronal injury in the cerebral cortex of 16 patients with chronic stroke based on a voxel-based lesion symptom mapping analysis. Neuronal injury in the corona radiata, caudate nucleus and putamen of patients with chronic stroke could predict walking speed. The behavioral measure scores were consistent with motor deficits expected after damage to the cortical motor system due to stroke. These findings suggest that voxel-based lesion symptom mapping may provide a more accurate prognosis of motor recovery from chronic stroke according to neuronal injury in cerebral motor cortex.

Protein crystallization with microseed matrix screening: application to human germline antibody Fabs.

The crystallization of 16 human antibody Fab fragments constructed from all pairs of four different heavy chains and four different light chains was enabled by employing microseed matrix screening (MMS). In initial screening, diffraction-quality crystals were obtained for only three Fabs, while many Fabs produced hits that required optimization. Application of MMS, using the initial screens and/or refinement screens, resulted in diffraction-quality crystals of these Fabs. Five Fabs that failed to give hits in the initial screen were crystallized by cross-seeding MMS followed by MMS optimization. The crystallization protocols and strategies that resulted in structure determination of all 16 Fabs are presented. These results illustrate the power of MMS and provide a basis for developing future strategies for macromolecular crystallization.

Structure and specificity of an antibody targeting a proteolytically cleaved IgG hinge.

The functional role of human antihinge (HAH) autoantibodies in normal health and disease remains elusive, but recent evidence supports their role in the host response to IgG cleavage by proteases that are prevalent in certain disorders. Characterization and potential exploitation of these HAH antibodies has been hindered by the absence of monoclonal reagents. 2095-2 is a rabbit monoclonal antibody targeting the IdeS-cleaved hinge of human IgG1. We have determined the crystal structure of the Fab of 2095-2 and its complex with a hinge analog peptide. The antibody is selective for the C-terminally cleaved hinge ending in G236 and this interaction involves an uncommon disulfide in VL CDR3. We probed the importance of the disulfide in VL CDR3 through engineering variants. We identified one variant, QAA, which does not require the disulfide for biological activity or peptide binding. The structure of this variant offers a starting point for further engineering of 2095-2 with the same specificity, but lacking the potential manufacturing liability of an additional disulfide. Proteins 2014; 82:1656-1667. © 2014 Wiley Periodicals, Inc.

Antibody modeling assessment II. Structures and models.

To assess the state-of-the-art in antibody structure modeling, a blinded study was conducted. Eleven unpublished Fab crystal structures were used as a benchmark to compare Fv models generated by seven structure prediction methodologies. In the first round, each participant submitted three non-ranked complete Fv models for each target. In the second round, CDR-H3 modeling was performed in the context of the correct environment provided by the crystal structures with CDR-H3 removed. In this report we describe the reference structures and present our assessment of the models. Some of the essential sources of errors in the predictions were traced to the selection of the structure template, both in terms of the CDR canonical structures and VL/VH packing. On top of this, the errors present in the Protein Data Bank structures were sometimes propagated in the current models, which emphasized the need for the curated structural database devoid of errors. Modeling non-canonical structures, including CDR-H3, remains the biggest challenge for antibody structure prediction.

Second antibody modeling assessment (AMA-II).

To assess the state of the art in antibody 3D modeling, 11 unpublished high-resolution x-ray Fab crystal structures from diverse species and covering a wide range of antigen-binding site conformations were used as a benchmark to compare Fv models generated by seven structure prediction methodologies. The participants included: Accerlys Inc, Chemical Computer Group (CCG), Schrodinger, Jeff Gray's lab at John Hopkins University, Macromoltek, Astellas Pharma/Osaka University and Prediction of ImmunoGlobulin Structure (PIGS). The sequences of benchmark structures were submitted to the modelers and PIGS, and a set of models were generated for each structure. We provide here an overview of the organization, participants and main results of this second antibody modeling assessment (AMA-II). Also, we compare the results with the first antibody assessment published in this journal (Almagro et al., 2011;79:3050).

IgG variable region and VH CDR3 diversity in unimmunized mice analyzed by massively parallel sequencing.

Most antigen-specific mouse antibodies have been derived by hybridoma technology, predominantly through use of the Balb/c strain. Much of the Balb/c germline repertoire of variable genes (V regions) is known. However, there is little information about the background expressed repertoire of IgG antibodies in mice, which reflects the baseline against which antigen-specific antibodies are generated through immunization. To assess this baseline repertoire, RNA was isolated from splenic B-cells enriched for expression of IgG from three mice. The RNA was individually amplified with three distinct PCR primer sets for comprehensive recovery of the heavy and light chain variable regions. Each PCR product was independently subjected to deep sequencing using 454 pyro-sequencing technology and analysed for redundancy, open reading frame, germline representation, and CDR3 sequence of the heavy chain variable region (VH CDR3) within and across the primer sets and mice. A highly skewed abundance of heavy and light chain variable gene usage was observed for all three primers in all three mice. While showing considerable overlap, there were differences among these profiles indicative of primer bias and animal-to-animal variation. VH CDR3 sequences were likewise highly skewed indicating that the heavy chain genes profiles substantially reflected individual antibodies. This observation was confirmed through analysis of randomly selected complete heavy chain variable sequences. However, there was very little redundancy in VH CDR3 sequences across the different mice. We conclude that the background IgG repertoire in young, unimmunized mice is highly skewed within individual mice and is diverse among them, a pattern similar to that observed in highly immunized mice.

Lateral clustering of TLR3:dsRNA signaling units revealed by TLR3ecd:3Fabs quaternary structure.

Toll-like receptor 3 (TLR3) recognizes dsRNA and initiates an innate immune response through the formation of a signaling unit (SU) composed of one double-stranded RNA (dsRNA) and two TLR3 molecules. We report the crystal structure of human TLR3 ectodomain (TLR3ecd) in a quaternary complex with three neutralizing Fab fragments. Fab15 binds an epitope that overlaps the C-terminal dsRNA binding site and, in biochemical assays, blocks the interaction of TLR3ecd with dsRNA, thus directly antagonizing TLR3 signaling through inhibition of SU formation. In contrast, Fab12 and Fab1068 bind TLR3ecd at sites distinct from the N- and C-terminal regions that interact with dsRNA and do not inhibit minimal SU formation with short dsRNA. Molecular modeling based on the co-structure rationalizes these observations by showing that both Fab12 and Fab1068 prevent lateral clustering of SUs along the length of the dsRNA ligand. This model is further supported by cell-based assay results using dsRNA ligands of lengths that support single and multiple SUs. Thus, their antagonism of TLR3 signaling indicates that lateral clustering of SUs is required for TLR3 signal transduction.

Structural basis for high selectivity of anti-CCL2 neutralizing antibody CNTO 888.

Human CC chemokine ligand 2 (CCL2), also known as monocyte chemoattractant protein-1 (MCP-1), is a member of the ß chemokine family whose actions are mediated through the G-protein-coupled receptor CCR2. Binding of CCL2 to its receptor CCR2 triggers calcium mobilization and chemotaxis. CCL2 is implicated in the pathogenesis of certain inflammatory diseases and cancer. CNTO 888, a neutralizing human anti-CCL2 antibody, was derived by antibody phage display. The antibody binds human CCL2 with high affinity (K(D)=22 pM) and inhibits CCL2 binding to its receptor. The crystal structure of the CNTO 888 Fab alone and in complex with the monomeric form of CCL2 (P8A variant) was determined at 2.6 Å and 2.8 Å resolution, respectively. CNTO 888 recognizes a conformational epitope encompassing residues 18-24 and 45-51 that overlaps the mapped receptor binding site. The epitope of CNTO 888 does not overlap with the dimerization site of CCL2, and thus its inhibitory activity is not expected to result from interference with the oligomeric state of CCL2. Comparison of the X-ray-determined epitopes of CNTO 888 and another CCL2-neutralizing antibody, 11K2, provides insight into the molecular basis of antibody selectivity and functional inhibition.

Crystallization of a challenging antigen-antibody complex: TLR3 ECD with three noncompeting Fabs.

The mechanism of action of therapeutic antibodies can be elucidated from the three-dimensional crystal structures of their complexes with antigens, but crystallization remains the primary bottleneck to structure determination. Methods that resulted in the successful crystallization of TLR3 ECD in complex with Fab fragments from three noncompeting, neutralizing anti-TLR3 antibodies are presented. The crystallization of this 238 kDa complex was achieved through fine purification of the quaternary complex of TLR3 with the three Fab fragments combined with microseed matrix screening and additive screening. Fine purification entailed the application of a very shallow gradient in anion-exchange chromatography, resulting in the resolution of two separate complex peaks which had different crystallizabilities. Subsequent structure determination defined the epitopes of the respective antibodies and revealed a mechanistic hypothesis that is currently under investigation. The results also showed that cocrystallization with multiple noncompeting Fab fragments can be a viable path when an antigen complex with a single Fab proves to be recalcitrant to crystallization.

Antibody modeling assessment.

A blinded study to assess the state of the art in three-dimensional structure modeling of the variable region (Fv) of antibodies was conducted. Nine unpublished high-resolution x-ray Fab crystal structures covering a wide range of antigen-binding site conformations were used as benchmark to compare Fv models generated by four structure prediction methodologies. The methodologies included two homology modeling strategies independently developed by CCG (Chemical Computer Group) and Accerlys Inc, and two fully automated antibody modeling servers: PIGS (Prediction of ImmunoGlobulin Structure), based on the canonical structure model, and Rosetta Antibody Modeling, based on homology modeling and Rosetta structure prediction methodology. The benchmark structure sequences were submitted to Accelrys and CCG and a set of models for each of the nine antibody structures were generated. PIGS and Rosetta models were obtained using the default parameters of the servers. In most cases, we found good agreement between the models and x-ray structures. The average rmsd (root mean square deviation) values calculated over the backbone atoms between the models and structures were fairly consistent, around 1.2 Å. Average rmsd values of the framework and hypervariable loops with canonical structures (L1, L2, L3, H1, and H2) were close to 1.0 Å. H3 prediction yielded rmsd values around 3.0 Å for most of the models. Quality assessment of the models and the relative strengths and weaknesses of the methods are discussed. We hope this initiative will serve as a model of scientific partnership and look forward to future antibody modeling assessments.

Promoting crystallization of antibody-antigen complexes via microseed matrix screening.

The application of microseed matrix screening to the crystallization of antibody-antigen complexes is described for a set of antibodies that include mouse anti-IL-13 antibody C836, its humanized version H2L6 and an affinity-matured variant of H2L6, M1295. The Fab fragments of these antibodies were crystallized in complex with the antigen human IL-13. The initial crystallization screening for each of the three complexes included 192 conditions. Only one hit was observed for H2L6 and none were observed for the other two complexes. Matrix self-microseeding using these microcrystals yielded multiple hits under various conditions that were further optimized to grow diffraction-quality H2L6 crystals. The same H2L6 seeds were also successfully used to promote crystallization of the other two complexes. The M1295 crystals appeared to be isomorphous to those of H2L6, whereas the C836 crystals were in a different crystal form. These results are consistent with the concept that the conditions that are best for crystal growth may be different from those that favor nucleation. Microseed matrix screening using either a self-seeding or cross-seeding approach proved to be a fast, robust and reliable method not only for the refinement of crystallization conditions but also to promote crystal nucleation and increase the hit rate.

Antibody Fab display and selection through fusion to the pIX coat protein of filamentous phage.

Fab antibody display on filamentous phage is widely applied to de novo antibody discovery and engineering. Here we describe a phagemid system for the efficient display and affinity selection of Fabs through linkage to the minor coat protein pIX. Display was successful by fusion of either Fd or Lc through a short linker to the amino terminus of pIX and co-expression of the counter Lc or Fd as a secreted, soluble fragment. Assembly of functional Fab was confirmed by demonstration of antigen-specific binding using antibodies of known specificity. Phage displaying a Fab specific for RSV-F protein with Fd linked to pIX showed efficient, antigen-specific enrichment when mixed with phage displaying a different specificity. The functionality of this system for antibody engineering was evaluated in an optimization study. A RSV-F protein specific antibody with an affinity of about 2nM was randomized at 4 positions in light chain CDR1. Three rounds of selection with decreasing antigen concentration yielded Fabs with an affinity improvement up to 70-fold and showed a general correlation between enrichment frequency and affinity. We conclude that the pIX coat protein complements other display systems in filamentous phage as an efficient vehicle for low copy display and selection of Fab proteins.

De novo selection of high-affinity antibodies from synthetic fab libraries displayed on phage as pIX fusion proteins.

Filamentous phage was the first display platform employed to isolate antibodies in vitro and is still the most broadly used. The success of phage display is due to its robustness, ease of use, and comprehensive technology development, as well as a broad range of selection methods developed during the last two decades. We report here the first combinatorial synthetic Fab libraries displayed on pIX, a fusion partner different from the widely used pIII. The libraries were constructed on four V(L) and three V(H) domains encoded by IGV and IGJ germ-line genes frequently used in human antibodies, which were diversified to mirror the variability observed in the germ-line genes and antibodies isolated from natural sources. Two sets of libraries were built, one with diversity focused on V(H) by keeping V(L) in the germ-line gene configuration and the other with diversity in both V domains. After selection on a diverse panel of proteins, numerous specific Fabs with affinities ranging from 0.2 nM to 20 nM were isolated. V(H) diversity was sufficient for isolating Fabs to most antigens, whereas variability in V(L) was required for isolation of antibodies to some targets. After the application of an integrated maturation process consisting of reshuffling V(L) diversity, the affinity of selected antibodies was improved up to 100-fold to the low picomolar range, suitable for in vivo studies. The results demonstrate the feasibility of displaying complex Fab libraries as pIX fusion proteins for antibody discovery and optimization and lay the foundation for studies on the structure-function relationships of antibodies.