Structural Characterization of the Full-Length Anti-CD20 Antibody Rituximab

Abstract

Rituximab, a murine-human chimera, is the first monoclonal antibody (mAb) developed as a therapeutic agent to target CD20 protein. Its Fab domain and its interaction with CD20 have been extensively studied and high-resolution atomic models obtained by X-ray diffraction or cryo-electron microscopy are available. However, the structure of the full-length antibody is still missing as the inherent protein flexibility hampers the formation of well-diffracting crystals and the reconstruction of 3D microscope images. The global structure of rituximab from its dilute solution is here elucidated by small-angle X-ray scattering (SAXS). The limited data resolution achievable by this technique has been compensated by intensive computational modelling that led to develop a new and effective procedure to characterize the average mAb conformation as well as that of the single domains. SAXS data indicated that rituximab adopts an asymmetric average conformation in solution, with a radius of gyration and a maximum linear dimension of 52 Å and 197 Å, respectively. The asymmetry is mainly due to an uneven arrangement of the two Fab units with respect to the central stem (the Fc domain) and reflects in a different conformation of the individual units. As a result, the Fab elbow angle, which is a crucial determinant for antigen recognition and binding, was found to be larger (169°) in the more distant Fab unit than that in the less distant one (143°). The whole flexibility of the antibody has been found to strongly depend on the relative inter-domain orientations, with one of the Fab arms playing a major role. The average structure and the amount of flexibility has been studied in the presence of different buffers and additives, and monitored at increasing temperature, up to the complete unfolding of the antibody. Overall, the structural characterization of rituximab can help in designing next-generation anti-CD20 antibodies and finding more efficient routes for rituximab production at industrial level.

Keywords: anti-CD20 mab; full-length antibody; molecular dynamics flexible fitting; small-angle X-ray scattering; structural comparison.

FIGURE 1

(A) Pair distribution function calculated from the measured SAXS profiles for rituximab. (B) graph with suggested q _max (red vertical line) based on the q * I (q) vs q curve and its derivative. (C) Likelihood score of D _max values calculated for different alpha values by using the Moore function to model P(r).

FIGURE 2

Results of the unrestrained molecular dynamics. χ² value of the fit between calculated and observed SAXS profile (A) and radius of gyration (B) as a function of the simulation time. The best fitting model (on the left) and the final one (on the right) are shown. Fa, Fb and Fc domains are coloured in cyan, blue and red, respectively. Glycans are coloured in green and put in licorice representation.

FIGURE 3

Results of the restrained molecular dynamics carried out by using the MDFF protocol: (A) starting (red) and final (blue) structural models superposed to the experimental molecular envelope; (B) Pearson’s correlation factor between calculated and observed molecular envelopes; (C) radius of gyration as a function of the simulation time; (D) χ² value of the fit between calculated and observed SAXS profile as a function of the simulation time.

FIGURE 4

(A) Best fitting model obtained by the static modelling procedure based on the MDFF protocol. Fa, Fb and Fc domains are coloured in cyan, blue and red, respectively. Glycans are coloured in green and put in licorice representation. (B) Best fit of the observed SAXS profile (red line) along with experimental errors (grey bar) and that calculated scattering profile from the above model (green line).

FIGURE 5

Elbow angle, i.e., the angle between the variable and constant Fab domains. (A) comparison of values measured from the crystal structure with PDB code 4KAQ and from both Fab domains of the full-length model used for MD simulations: initial model (MD init), best-fitting model in unrestrained MD (MD best) and best-fitting model in MDFF (MDFF best); (B) Fa (black line) and Fb (red line) elbow angle values as a function of the simulated time for the unrestrained MD and the MDFF simulation (inset). The simulation time corresponding to MD best and MDFF best is highlighted with arrows.

FIGURE 6

Results of the modelling procedure based on the ensemble optimization method. (A) best fit of the calculated SAXS profile (green line) vs the observed one (red line, experimental error bars in grey); (B) distribution of the radius of gyration (R _g ) and the maximum interatomic distance (D _max) values for the selected models (green and cyan lines, respectively) and initial pool of random mAb structures (red and blue lines, respectively).

FIGURE 7

Comparison of the structural models generated by the ensemble optimization method based on principal component analysis (PCA) of related geometrical features. (A) PCA loadings values showing the individual geometrical parameters in discriminating the models; (B) PCA scores values showing the model discrimination. The models are drawn next to their representative points, with Fa, Fb and Fc coloured in cyan, blue and red, respectively, together with their relative weight in the EOM ensemble. The fraction of the total data variance explained by the first (PC1) and second (PC2) principal component is reported on relative the axes of the loadings (A) and scores (B) plots.

FIGURE 8

Results of the SAXS analysis of data taken at different temperatures. Radius of gyration R _g (A) and R _flex parameter determined by the ensemble optimization method for the selected structures (blue line) and the pool of generated structures (red line) (A).

FIGURE 9

(A) Radius of gyration (R _g ) and (B) maximum inter-particle distance (D _max) calculated from SAXS data as a function of the mAb concentration. Pearson’s correlation factor for R _g (C) and D _max (D) against concentration values. Samples refer to mAb in buffer A and in the presence of the additives listed in Supplementary Table S1.

FIGURE 10

PCA analysis of the raw SAXS data taken for rituximab in solutions containing different additives. Superposition of the individual SAXS profiles, representing the data matrix supplied to PCA (A); loadings of the first (PC1) and second (PC2) principal components (B); PC1 scores as a function of the dataset considered, with mAb concentration in mg/ml and type of additive reported (C). Plots B and C have been obtained after removing the profile of the sample A + TWE from the data matrix, as it produces large deviations in the PCA.

FIGURE 11

Effect of additives on structural parameters of rituximab derived from SAXS data. Radius of gyration (R _g ) and maximum inter-particle distance (D _max) (A) and R _flex figure of merit for the generated pool of structures and those selected (B).

FIGURE 12

Comparison based on PCA of geometrical features of rituximab models obtained from SAXS measurements in the presence of several additives. (A) PCA loadings values showing the role of individual geometrical parameters in discriminating the models; (B) PCA scores values showing the model discrimination. The models are drawn next to their representative points, with Fa, Fb and Fc coloured in cyan, blue and red, respectively. The fraction of the total data variance explained by the first (PC1) and second (PC2) principal component is reported on relative the axes of the loadings (A) and scores (B) plots.

FIGURE 13

PCA analysis of the raw SAXS data taken for rituximab in solutions containing different sugars. Superposition of the individual SAXS profiles, representing the data matrix supplied to PCA (A); loadings of the first (PC1) and second (PC2) principal components (B); PC1 scores as a function of the dataset considered, with mAb concentration in mg/ml and type of additive reported (C).

FIGURE 14

(A) Radius of gyration (R _g ) and (B) maximum inter-particle distance (D _max) measured for rituximab in buffer B at 8 mg/ml and in presence of mono-, di- and tri-saccharides.

FIGURE 15

Comparison of the structural models based on PCA of geometrical features of rituximab models obtained from SAXS measurements in the presence of several additives. (A) PCA loadings values showing the role of individual geometrical parameters in discriminating the models; (B) PCA scores values showing the model discrimination. The models are drawn next to their representative points, with Fa, Fb and Fc coloured in cyan, blue and red, respectively. The fraction of the total data variance explained by the first (PC1) and second (PC2) principal component is reported on relative the axes of the loadings (A) and scores (B) plots.