Advertisement

Analytical and Bioanalytical Chemistry

, Volume 410, Issue 8, pp 2141–2159 | Cite as

Characterization of the NISTmAb Reference Material using small-angle scattering and molecular simulation

Part I: Dilute protein solutions
  • Maria Monica Castellanos
  • Steven C. Howell
  • D. Travis Gallagher
  • Joseph E. Curtis
Research Paper

Abstract

Both conformational and colloidal stability of therapeutic proteins must be closely monitored and thoroughly characterized to assess the long-term viability of drug products. We characterized the IgG1 NISTmAb reference material in its histidine formulation buffer and report our findings on the higher order structure and interactions of NISTmAb under a range of conditions. In this paper we present the analysis of experimental small-angle scattering data with atomistic molecular simulations to characterize the monodisperse dilute solution of NISTmAb. In part II we describe the characterization of the NISTmAb at high protein concentration (Castellanos et al. 2018). The NISTmAb was found to be a flexible protein with a radius of gyration of 49.0 ± 1.2 Å in histidine formulation buffer using a variety of neutron and X-ray scattering measurements. Scattering data were then modeled using molecular simulation. After building and validating a starting NISTmAb structure from the Fc and Fab crystallographic coordinates, molecular dynamics and torsion-angle Monte Carlo simulations were performed to explore the configuration space sampled in the NISTmAb and obtain ensembles of structures with atomistic detail that are consistent with the experimental data. Our results indicate that the small-angle scattering profiles of the NISTmAb can be modeled using ensembles of flexible structures that explore a wide configuration space. The NISTmAb is flexible in solution with no single preferred orientation of Fc and Fab domains, but with some regions of configuration space that are more consistent with measured scattering profiles. Analysis of inter-domain atomistic contacts indicated that all ensembles contained configurations where residues between domains are ≤ 4 Å, although few contacts were observed for variable and C H 3 regions.

Graphical Abstract

Heavy atom self contact maps of the NISTmAb indicate a highly-flexible structure.

Keywords

Small-angle scattering NISTmAb reference material Antibody structure Protein conformation Antibody flexibility Higher order structure 

Introduction

Therapeutic proteins have been used to treat a set of challenging conditions such as cancer, immunological disorders and infectious diseases [46, 47]. In particular, monoclonal antibodies and their derivatives are the leading and fastest growing class of biotherapeutic products [17, 46, 47]. The development of a successful therapeutic candidate not only requires high efficacy and potency, but also thermodynamic and chemical stability of the native structure as it is often only marginally stable compared to unfolded conformations.

Small-angle scattering can provide information on the higher-order structure and colloidal stability of therapeutic proteins, because it is sensitive to the spatial relation of atoms in the sample. These measurements can be performed using either neutrons or X-rays, and thus these techniques are known as small-angle neutron scattering (SANS) and small-angle X-ray scattering (SAXS) respectively. Details on the technique applied to soft-materials and biotechnology [21, 27, 43, 55], and in particular to therapeutic proteins [8], can be found in the literature. Briefly, a neutron or x-ray beam interacts with the sample and the intensity of the scattered neutrons or photons is measured using a 2D detector. The 2D intensity pattern is generally isotropic and can be radially averaged to produce a characteristic 1D scattering profile of the scattered intensity as a function of scattering angle. The scattering angle correlates inversely with the length scale among scattering nuclei and the measured intensity correlates with the frequency of correlations at each scattering angle. Although small-angle scattering does not provide exact structural details at atomic resolution, changes in molecular conformation can produce differences in the resulting scattering profiles and combined with computational models or simulations can provide information not readily available by other techniques.

Both SANS and SAXS are suitable for the study of therapeutic proteins under a range of concentrations and conditions. However, each scattering technique has some advantages and drawbacks as shown in Table 1. One of the main advantages of SANS is the sensitivity to hydrogen and deuterium, which allows the use of contrast matching techniques to study different components of a multicomponent system [54, 63]. Therefore, by mixing different ratios of H2O and D2O, the subunits of a protein complex can be studied; in addition, the structure of specific domains can be investigated using selective labeling. Moreover, because neutrons have no charge, they can deeply penetrate a wide range of sample environments without interference; thus, samples can be studied under extreme conditions of temperature and pressure in situ. On the other hand, SAXS has a much higher throughput than SANS and can require less material. Depending on the protein and sample concentrations, SAXS measurements can take seconds to a few minutes, and sample loading can be automated with a robot. However, care should be taken to avoid radiation damage, especially when using synchrotron sources. Both SANS and SAXS access the length scales of interest for therapeutic proteins from ∼10 to ∼3000 Å and can be used to measure samples over a wide range of concentrations and formulation buffers. Additionally, both techniques can be combined with inline size-exclusion chromatography (SEC) to remove aggregates or unwanted species before the scattering measurements [28, 32].
Table 1

Comparison of SANS and SAXS measurements for therapeutic proteins

Small-angle neutron scattering (SANS)

Small-angle X-ray scattering (SAXS)

Sensitive to isotopes: contrast matching to study multicomponent systems

Not sensitive to isotopic forms of an element

No radiation damage

Radiation damage might occur

Volume ∼ 400-700 μ L. Sample can be recovered

Volume ∼5–100 μ L. Sample is difficult to recover

Experiments must be done in a neutron scattering facility

Experiments can be done using a commercial source or at a synchrotron

Measurement times from minutes to hours

Measurement times from seconds to minutes (high throughput)

Flexible sample environment, measurements under extreme temperatures or pressures

Limited for amorphous systems

Suitable to study protein structure, protein-protein interactions, protein complexes

Length scales: 15 Å to 3000 Å (10 Å = 1 nm)

Protein concentrations: ∼0.5 mg/mL up to the solubility limit (hundreds of mg/mL)

Measurements can be done inline with SEC at certain facilities

The scattered intensity I(Q) of molecules with volume V p and volume fraction ϕ can be expressed as:
$$ I (Q) = \phi V_{p} ({\Delta} \rho)^{2} P(Q) S^{\prime}(Q), $$
(1)
where P(Q) is the form factor and S(Q) is the effective structure factor [23]. Q is the momentum transfer and is proportional to the scattering angle 2𝜃. Moreover, Q can be related to the length scale d using Bragg’s law as d = 2π/Q. Δρ, known as the contrast, depends on the scattering length densities or atomic scattering factors of the components of the system. Specifically, the contrast corresponds to the difference in scattering length densities between the molecules and the solvent (buffer) and it is independent of Q. Therefore, only P(Q) and S(Q) have a Q-dependence and contain information on the protein conformation and protein-protein interactions respectively. The following paragraphs describe how structural information about the native state can be obtained from small-angle scattering.

Small-angle scattering can be used to obtain information on the shape of the molecule and conformation in solution. Under dilute conditions, molecules are far from each other so that no intermolecular correlations are observed in the scattering profile and S(Q) = 1. Therefore, I(Q) varies with Q only according to P(Q). For a monodisperse system, that is, when molecules have the same molecular weight, P(Q) contains information on the radius of gyration R g and the distribution of atoms in the molecule (i.e. protein structure and shape). Moreover, if the coordinates of the atoms for a structural model are known, for example from a crystal structure, the scattering profile of the model structure can be calculated using the Debye equation [16], and compared to the experimental scattering profile.

Guinier [22] showed that the radius of gyration R g of a macromolecule can be obtained by calculating the slope in the low-Q limit of a plot of ln(I(Q)) as a function of Q2 as follows:
$$ \ln(I(Q)) = \ln(I(0)) - \left( {{R^{2}_{g}}\over{3}}\right) Q^{2}. $$
(2)
I(0) represents the scattered intensity at zero angle, that is, Q = 0. I(0) depends on the molecular weight, the scattering length density or atomic scattering factor, the molecular volume and the concentration of protein. This expression is valid at the generally accepted limit of QR g ≤ 1.3, which corresponds to the low Q limit where the length scales are relevant to the molecular size. Moreover, Eq. 2 is only valid for dilute conditions, that is, when S(Q) = 1, and thus the system can be considered monodisperse without intermolecular interactions. The dilute condition not only depends on the protein, but also on the formulation buffer, temperature, and pressure. To determine if a system is monodisperse and without intermolecular interactions, one considers conditions where R g does not change with increasing protein concentration.
The pair distribution function P(r) describes the probabilities of distances between all atom pairs in the system. Thus, if the coordinates of the atoms are known, P(r) can be calculated as:
$$ P(r) = \sum\limits_{i,j = 1, i\neq j}^{N} \| \mathbf{r_{j}}-\mathbf{r_{i}} \|, $$
(3)
where N is the total number of atoms, r i and r j are the vectors describing the position of atoms i and j respectively. From a scattering curve of I(Q) vs. Q, obtained under dilute conditions such that I(Q) ∼ P(Q), P(r) can be calculated using an indirect Fourier transform as follows [43]:
$$ P(r) = {r\over{2\pi^{2}}} {{\int}_{0}^{\infty} I(Q)Q\sin{(Qr)}dQ}. $$
(4)
The resulting curve depicts the probability of finding a pair of atoms at a distance r. High values of a P(r) function indicate that a larger number of pairs of atoms are found at that particular distance. Contrary to I(Q), P(r) is a real-space distribution and thus provides interatomic distance information, such as characteristic distances within and the maximum length of a molecule. For example, characteristic distances could indicate domain-domain distances for a protein with several globular domains.

The compactness of the molecule can be qualitatively assessed using a Kratky plot, represented as I(Q)Q2 as a function of Q. For globular proteins, the Kratky plot results in a bell-shaped curve, whereas elongated and flexible proteins show deviations from the bell-shape and linearity with respect to Q at high Q. More details on these metrics and their usage can be found in the literature [43].

In addition, if atomic coordinates from experiment or a homology model of the protein are available, these coordinates can be used to calculate a scattering profile. Thus, molecular models from atomistic simulations can be a valuable complement to the experimental scattering data. Model structures or an ensembles of structures can be rejected if their scattering profiles do not match the experimental data. SASSIE [13, 14] is a program suite freely available to create atomistic models of molecular systems, calculate their scattering profiles and compare model profiles to experimental data. The calculation of scattering profiles from atomistic models can be performed using the golden ratio method [59] for both SANS or SAXS.

SANS and SAXS have been used to study the conformation of antibodies in solution. One of the earliest SAXS studies with antibodies analyzed the structure of an antigen-Fab complex to determine the relative position of the domains in the antigen and characterize the epitope and structure of the complex [60]. More recent studies using small-angle scattering focused on the conformation of antibodies in solution and their flexibility [1, 3, 12, 31, 34, 37, 38, 45, 56, 57]. The aggregation mechanism and loss of structure of an IgG1 was studied using SANS, which showed that the morphology of the resulting aggregate depended on the aggregation mechanisms [4, 5]. Importantly, some of these studies have proposed that SAXS can be a valuable technique for formulation development of monoclonal antibodies [26, 56]. The effect of excipients on the solution structure of therapeutic proteins can be studied with small-angle scattering.

In this manuscript, we present small-angle scattering data on the NISTmAb Primary Sample 8670 (NISTmAb) to characterize solution structure(s) in a histidine formulation buffer. The NISTmAb is an IgG1 that serves as a common framework to determine to assess performance of existing and new analytical technologies [53]. A collection of three volumes of the book “State-of-the-Art and Emerging Technologies for Therapeutic Monoclonal Antibody Characterization” have been published and contain detailed information on the analytical, technological and regulatory aspects of this molecule [50, 51, 52]. These volumes and the present publication series on the quality and availability of the NISTmAb are intended to provide extensive information on the NISTmAb as a reference material and a guide for emerging technologies to characterize biotherapeutics. Besides presenting new biophysical data on the NISTmAb, we anticipate that this manuscript and its accompanying study at high protein concentrations [7] will highlight the advantages and capabilities of small-angle scattering to characterize the conformation and colloidal stability of therapeutic proteins.

This first section of the manuscript describes the monodispersity of the samples used for the small-angle scattering study, followed by a comparison of the dilute scattering profiles, using SANS and SAXS, of NISTmAb solutions in buffers prepared in H2O, D2O and with 150 mM NaCl. We analyze the resulting scattering profiles with the Guinier approximation, Kratky plots and P(r). This analysis is followed by simulations on a structure of the NISTmAb built from crystallographic coordinates of the domains. The scattering profiles of structures from simulations are compared to the experimental data and ensembles of structures that describe the experimental data are further analyzed to gain information on the configuration space sampled by each ensemble, the distribution of Fab and Fc domains, and plausible regions on the surface of the molecule that can be in contact. Our study shows the importance of the backbone torsion-angle sampling to explore the wide configuration space sampled by antibodies and that ensembles of flexible NISTmAb structures describe the experimental scattering data in solution.

Methods

Sample preparation

The NISTmAb Primary Sample 8670 was obtained from the National Institute of Standards and Technology (NIST) in frozen vials at 10 and 100 mg/mL. After thawing the vials overnight at 4 C, NISTmAb solutions were buffer-exchanged using Amicon Ultra-15 centrifugal filters (Millipore, UFC903024)1 with a 30 kDa molecular weight cutoff in a swinging bucket centrifuge (Thermo Scientific, Sorvall ST 40R Centrifuge, 75004525) at 4000 relative centrifugal force. Six cycles of fresh buffer addition were performed to reach > 99.9% of the desired buffer. The formulation buffer contained 25 mM histidine with either 0 or 150 mM NaCl (Sigma Aldrich, S9888). The deuterated buffer consisted of 12.5 mM L-histidine monohydrochloride (JT Baker, JT2081-6) and 12.5 mM L-histidine (JT Baker, JT2080-5) in 99.9% D2O (Sigma-Aldrich, 151882) and adjusted to pD= 6.4 (pH= 6.0) with a 1M sodium deuteroxide solution (Sigma Aldrich, 372072) before filtering with a 0.2 μ m filter. Similar steps were taken for the histidine buffer in H2O, but using Millipore SuperQ water instead of D2O. Samples were measured at 25 C within three days of being prepared, unless noted otherwise, and stored at 4 C if not in use.

Small-angle scattering measurements

SANS measurements were performed at the NIST Center for Neutron Research using the NG7 and NGB 30 m SANS instruments. The neutron wavelength λ was 6 Å and 8.4 Å with a wavelength spread δλ/λ of 0.15 Å. Scattered neutrons were detected with a 64 cm x 64 cm two-dimensional position-sensitive detector with 128 x 128 pixels at a resolution of 0.5 cm/pixel. Data reduction was performed using SANS macro routines developed at the NCNR using Igor [30]. Raw counts were normalized to a common monitor count and corrected for empty cell counts, ambient room background counts and non-uniform detector response. Data were placed on absolute scale by normalizing the scattered intensity to the incident beam flux. Finally, all the data were found to be isotropic and thus data were radially averaged to produce a 1D scattered intensity, I(Q), versus Q profiles, where Q = 4π sin(𝜃)/λ and 2𝜃 is the scattering angle. Measurements were performed at 25 C using three configurations by varying the sample-to-detector distance between 1.3 and 13.4 m to cover a Q range of 0.003 to 0.45 Å− 1. We have shown that the scattering profile of the NISTmAb does not vary with temperature in the range from -5 C to 25 C [7]. Scattered intensities were corrected for buffer scattering and incoherent scattering from hydrogen.

Synchrotron SAXS measurements (SAXS-1) were collected at the European Synchrotron Radiation Facility on the BM29 BioSAXS beamline [40]. Samples were loaded in 1.8 mm diameter quartz capillaries for data collection at 20 C. The beam size was 500 μ m x 500 μ m at the detector plane. The sample to detector distance was up to 2.867 m and a tunable energy between 7 to 15 keV was used to cover the range 0.0074 Å− 1 < Q < 0.45 Å− 1. Scattered photons were detected by a two-dimensional CMOS hybrid pixel Pilatus 1M detector. Data acquisition was done with the Biosaxs Customized Beamline Environment software. The average profile is a merge of a sample at 1.25 mg/mL (low and intermediate Q) and 10 mg/ml (intermediate and high Q).

In-house SAXS measurements (SAXS-2) were performed using a Rigaku X-ray source and the SAXSLab Ganesha platform at the Institute for Bioscience and Biotechnology Research. Samples were loaded into a 96-well plate and sealed with tape to prevent solvent evaporation. 20 μ L of sample were suctioned, loaded into a 1.3 mm capillary by an automated robot and oscillated during data collection. Sample to detector distance was varied from 0.7 to 1.7 m and a wavelength of 1.5418 Å was used to cover the range 0.005 Å− 1<Q< 0.45 Å− 1. Scattered photons were detected by a two-dimensional Pilatus 300K detector. To clean the cell, three steps were carried out after each measurement: flowing water, flowing hellmanex at 5% and drying with air. Buffer measurements were collected between sample measurements to ensure the absence of protein carryover in the capillary.

The absence of radiation damage was confirmed by collecting short exposures and comparing the consistency of the radius of gyration and low-Q scattering profile of the initial and final exposures. Free-radical scavengers were not added to the protein solutions as radiation damage was not observed in any of the SAXS measurements in the histidine formulation buffer. Data reduction was performed using ScÅtter [44] and RAW software [35]. Absolute scaling was achieved by measuring a water sample at 20 C, and then subtracting the scattering of the empty capillary and calculating the scaling factor that produced a scattering intensity of 1.632 x 10− 2 cm− 1 [36]. Pair distribution functions and Guinier fits were calculated using the RAW software [35].

Monomer purity characterization

Size-exclusion chromatography was performed using a UV-detection based high performance size exclusion chromatography (HP-SEC Thermo Scientific/Dionex U3000), consisting of a HPG 3400 binary pump (Thermo Fisher Scientific, 5040.0046), a thermostatted WPS-3000TRS autosampler (Thermo Fisher Scientific, 5840.0020), a thermostatted column compartment TCC-3000RS (Thermo Fisher Scientific, 5730.0000), and a four channel variable wavelength detector (Thermo Fisher Scientific, 5074.0010). Phosphate buffer saline solution at pH 7.4 was used as the mobile phase. Samples were not diluted prior to injection and thus the injected volume depended upon the concentration of protein in each sample. Samples were injected onto a TSKgel G3000SWxl column (Tosoh Bioscience, 08541) and absorbance was measured at 280 nm. The flow rate was set to 0.45 mL/min. The limits of detection and quantification for this method have been estimated as 0.026% and 0.086% respectively.

Dynamic Light Scattering (DLS) experiments were performed with a DynaPro NanoStar instrument (Wyatt Technology Corporation). Samples were allowed to equilibrate to 25 C before collecting 5-10 measurements for the same sample. Samples were then measured in duplicates before and after the small-angle scattering measurements with no statistical differences observed. A laser wavelength λ of 663 nm with a 90 o scattering angle 𝜃 were used. The scattering vector was calculated as q = 4πn sin(𝜃/2)/λ, where n is the refractive index. Hydrodynamic radius R H was obtained using the Stokes-Einstein relation R H = k B T/(6πηD), where k B is the Boltzmann constant, T is the temperature, η is the viscosity, and D is the diffusion coefficient for each relaxation mode obtained in the autocorrelation function using the regularization analysis.

Simulations and model structure of the intact NISTmAb

To perform MD and torsion-angle MC simulations of the NISTmAb, a starting model of the intact antibody was built by combining separate crystallographic structures for the three components: the Fab fragment (used twice), the Fc fragment, and the hinge region. Fab coordinates were taken from the Fab in the 2.0 Å resolution NIST Fab crystal structure PDB 5K8A. This Fab model has a complete light chain and its heavy chain is complete through residue Cys 223.

Fc coordinates were taken from PDB 5VGP, the 2.1 Å crystal structure of the NIST Fc. The Fc structure has complete chains from Gly239 through Ser447 and includes G1F/G0F glycans, consistent with mass spectrometry data and glycan analysis [18, 42]. The fragments were produced using papain cleavage and purification of the Fc was carried out using protein A (Thermo Scientific Pierce Fab Preparation Kit, Model 44985) and gel filtration.

The starting model for the hinge was made from PDB crystal structure 1HZH [49], which is an intact human IgG1 mAb. The hinge in this structure had the same protein sequence as the NISTmAb, but the coordinates lacked three residues and one disulfide bond. The hinge was combined with the Fab and Fc domains to build a starting model of the complete NISTmAb for molecular simulations. The domains in the starting model were arranged in a similar conformation as in the 1HZH structure resulting in a distinctly asymmetric structure with a close contact between the Fc and one of the Fabs. However, the close Fc-Fab contact did not persist in the molecular simulations.

Any residues not observed in the crystallographic structures, but reported in the primary sequence [18], were added using psfgen 1.6 from NAMD [41]. Disulfide bonds and hydrogens were also added using psfgen. Glycan structures were created using appropriate residue definitions and patches [58] to provide intra-chain bonding in the glycans and creating the bond between glycans and protein via Asn300 of each heavy chain. The R g of the starting structure was 50.2 Å. The molecule was solvated with a 15 Å padding of TIP3P water [29] in a rectangular box with 12 chlorine counterions using VMD [25]. The starting structure was energy minimized using the conjugate gradient algorithm in NAMD with the CHARMM36 force field [58]. Energy minimization was done in steps by restraining first all protein atoms, the heavy atoms (non-hydrogen), the protein backbone, residues with alpha or beta sheets, and finally without restraints, where all atoms were allowed to relax. A 10 ns of MD simulation in explicit water was performed, of which the first 100 picosecond were in the NVT ensemble followed by a 9.9 ns NPT simulation. The time step was set to 1 femtosecond, temperature to 300 K, and pressure to 101.325 kPa. The Particle Mesh Ewald algorithm [15] was used using a 1 Å grid to account for long-range electrostatic interactions. Bond-length constraints were applied using the SHAKE algorithm [48].

After 2 ns of NPT simulation, water and ions were removed from the system and torsion-angle MC simulations were performed using the web-version of SASSIE [6, 14, 39]. The backbone torsion-angles of residues Asp224 to Thr228 were sampled while the other regions of the protein remained as rigid bodies. This region encompasses the residues between the disulfide bond linking the two heavy chains and the disulfide bond between the light and heavy chains (see Fig. 3). Although the region in the hinge from Pro233 to Leu238, encompassed by one of the disulfide bonds between the heavy chains and the first residue of the Fc observed in the crystal structure, is also believed to be flexible, this region was not explicitly sampled in the torsion-angle MC simulations, because the required concerted move was not available. Although this limitation might affect the overall configuration space sampled, the resulting ensemble of structures sampled an extensive range of configurations and Fab-Fc Fab-Fab distributions. The resulting 136568 accepted structures (∼50% acceptance rate) were energy minimized for 5000 steps in vacuum, before calculating scattering profiles using the golden ratio method [59]. The scattering profiles of the model structures were compared to the experimental scattering profile using the χ2 parameter (see equation 5). All images of NISTmAb structures were generated using VMD [25].

When evaluating the relative inter-domain angles and distances, shown in Figs. 9 and 10, we used the center of mass and principal axes to define a reference frame for each of the three domains. For each domain, the principal axes defined three orthogonal lines that intersected at the domain’s center of mass. We defined three principal vectors aligned along these principal axes. The first, second, and third principal vectors, I1, I2, and I3, were aligned to the axis that corresponds to the longest, second longest, and shortest dimension of the domain, respectively. Originating at the center of mass, I1 pointed away from the center of mass of the complete NISTmAb. I2 was aligned along the second longest dimension of the domain and was chosen to point toward heavy-chain-1, light-chain-1, and light-chain-2, for Fc, Fab1, and Fab2 respectively. I3 was defined as I1 × I2 with its orientation set using the right-hand rule.

Results

In order to confirm the monodispersity and stability of samples, size-exclusion chromatography (SEC) and dynamic light scattering (DLS) were performed on the same samples used during the small-angle scattering measurements. The monomeric purity of samples was studied using SEC. Figure 1a shows the chromatograms obtained for the NISTmAb solutions after storage at 4 C for up to two weeks. All samples had a monomer purity higher than 99.4% and low content of high-molecular-weight (HMW) species. The inset of Fig. 1 displays the peaks for the HMW aggregates and domains (Fab and Fc) found during the measurement; note the two orders of magnitude of change in the y-axis scale, denoting a low amount of these species. Figure 1b displays the size distribution obtained from the DLS measurements. Only measurements at low concentrations, for which intermolecular interactions are negligible, were used to calculate the hydrodynamic size. Contrary to the Stokes radius, the hydrodynamic size from DLS depends on protein concentration. DLS showed that the NISTmAb appears to be slightly larger in the presence of 150 mM NaCl, but the hydrodynamic sizes were comparable to those previously reported for the NISTmAb [20]. Moreover, the DLS data of Fig. 1b for the 0 mM NaCl sample showed a small population at ∼10–15 μ m that contributed to less than 2% of the intensity (data not shown).
Fig. 1

a Size-exclusion chromatography data for solutions of NISTmAb. Inset displays a closer view of the smallest peaks from the SEC chromatogram. b Percentage of intensity as a function of hydrodynamic size for NISTmAb solutions using dynamic light scattering

Figure 2 shows SAXS and SANS data of the NISTmAb at dilute conditions. At low concentrations, the scattering profiles are dominated by the form factor P(Q), which contains information on the intramolecular structure. SAXS scattering profiles collected at two different facilities were compared with SANS data and evaluated for data reproducibility and consistency. Scattering profiles were also measured in D2O buffer, because most SANS measurements require deuterated solvents to minimize the contribution of incoherent scattering. On the other hand, X-rays do not distinguish between hydrogen and deuterium and thus can provide a control for changes resulting from the exchange to a deuterated solvent. Figure 2a depicts the scattering intensity I(Q) obtained with two SAXS instruments from different sources: the European Synchrotron Radiation Facility (ESRF) in Grenoble, France (referred to as SAXS-1), and the in-house source at the Institute for Bioscience and Biotechnology Research (IBBR) in Rockville, MD, United States (SAXS-2). No significant differences were observed among the SAXS profiles at similar concentrations and the results were independent of where the data were collected. Moreover, the SAXS profile in D2O buffer compared well to the SAXS profile in H2O buffer. Figure 2b shows the Guinier and Kratky plots for the NISTmAb data from Fig. 2a. The Kratky plots have the expected deviation from a bell-shaped curve as antibodies are flexible in solution. The Guinier plots (inset of Fig. 2b) were used to estimate the R g (size) of the protein as described previously. The results of this analysis are summarized in Table 2 for the NISTmAb in histidine buffer with H2O and D2O, after adding NaCl, and obtained with SANS and SAXS in different facilities. The resulting R g for the NISTmAb in H2O histidine buffer was 49.0 ± 1.2 Å (uncertainty corresponds to one standard deviation).
Fig. 2

SAXS and SANS profiles of the NISTmAb measured at 5 mg/mL, except the profile for SAXS-1 at 1.25 mg/mL (see experimental section for details). a SAXS profiles from instruments SAXS-1 (synchrotron facility) and SAXS-2 (in-house source). b SAXS derived Kratky and Guinier plots, confirming the non-globular shape of the NISTmAb and enabling the estimation of R g respectively. c SANS data in H2O and D2O at different ionic strengths. D. Pair distribution functions representing correlations for pairs of all atoms. Profiles have been arbitrarily scaled for comparison purposes. Error bars correspond to ± 1 the standard deviation

Note that for the 0 mM NaCl samples, the dilute condition applied up to concentrations of ∼6 mg/mL and deviations in R g , although small, were already observed at ∼8 mg/mL as shown in Table 2.
Table 2

Guinier fit results for scattering curves of NISTmAb obtained with various instruments using different solvent conditions and protein concentrations

Protein concentration (mg/mL)

Buffer solvent

Facility

Source

R g (Å)

Range

I(0) (cm− 1)

1.25

H2O

ESRF

SAXS-1

49.6 ± 0.2

0.49 ≤ QR g ≤ 1.28

5.5

D2O

IBBR

SAXS-2

46.1 ± 1.5

0.54 ≤ QR g ≤ 1.28

0.44

7.6

H2O

IBBR

SAXS-2

43.2 ± 2.8

0.50 ≤ QR g ≤ 1.28

0.64

5.6

H2O

IBBR

SAXS-2

48.2 ± 4.6

0.50 ≤ QR g ≤ 1.30

0.48

3.7

H2O

IBBR

SAXS-2

47.4 ± 5.2

0.47 ≤ QR g ≤ 1.30

0.34

1.9

H2O

IBBR

SAXS-2

49.9 ± 17.5

0.88 ≤ QR g ≤ 1.23

0.16

5.5

D2O

NCNR

SANS

43.3 ± 2.0

0.45 ≤ QR g ≤ 1.24

0.71

5.4

D2O + 150 mM NaCl

NCNR

SANS

52.1 ± 2.3

0.60 ≤ QR g ≤ 1.25

0.84

5.5

H2O

NCNR

SANS

43.8 ± 7.3

0.40 ≤ QR g ≤ 1.25

0.34

5.4

H2O + 150 mM NaCl

NCNR

SANS

49.6 ± 6.1

0.33 ≤ QR g ≤ 1.23

0.41

Uncertainty corresponds to the error in the linear fitting. Uncertainties in I(0) are smaller than the last significant figure and thus not reported

Figure 2c depicts the SANS scattering profile of the NISTmAb collected at the NIST Center for Neutron Research in Gaithersburg, MD, United States. In the case of SANS, the profiles were measured in buffers with H2O and D2O, and in buffer containing 150 mM (1 mM = 10− 3 mol/L) NaCl. At Q> 0.1 Å− 1 the incoherent background in H2O contributed to the scattering more than the protein; thus, in that Q-range, there was limited structural information to be extracted from the SANS profile in H2O buffer. The addition of NaCl did not change the scattering profile, with the exception of a minor increase in the low-Q intensity. Two characteristic shoulders in the Q region 0.05 Å− 1<Q< 0.2 Å− 1, which have also been observed for other antibodies [9, 12, 19, 61], were present in SAXS and SANS profiles. Figure 2d shows the pair distribution function P(r) for different NISTmAb samples using SAXS and SANS. P(r) describes the distances between all atom pairs in the molecule. Two peaks were observed in P(r) at ∼45 and ∼80 Å, that correspond to Q of ∼0.14 and ∼0.08 Å− 1 respectively, where the shoulders in the I(Q) profiles were identified. Since Fig. 2d includes all types of correlations, one can only infer about the origin of the peaks based on their position. However, using all atom structures, we showed that these distances correspond to intra- and inter-domain correlations (see sections below). Therefore, these peaks contain information on the relative positions of the Fabs and Fc, i.e., the overall conformation of the antibody. The P(r) curves also showed that the maximum pair distance between atoms, where P(r) = 0 at large values of r, was slightly smaller in the deuterated buffer and increased in the presence of 150 mM NaCl.

Measured scattering data were compared to molecular simulations that explore the flexibility of the NISTmAb. Figure 3 shows a structure of the NISTmAb that is consistent with the experimental data as explained below. See methods for details on building the starting model and the simulations. Briefly, 10 ns molecular dynamics (MD) simulations in explicit water were performed to produce a representative equilibrated structure. However, given the limitations of computational resources, all atom MD simulations alone could not efficiently generate the wide range of conformations that antibodies can explore. Therefore, MD simulations were followed by Monte Carlo (MC) simulations that sample backbone torsion-angles in the hinge residues Asp224 through Thr228 (DKTHT) of each heavy chain; this region is depicted in the inset of Fig. 3.
Fig. 3

a Surface representation of a representative NISTmAb structure with its corresponding regions. The structure was colored as follows: Fab is represented in red (light chain) and gray (heavy chain); hinge in light blue; and Fc in orange (heavy chains) and green (glycans). In the inset, the yellow highlights the hinge segments sampled using the torsion-angle MC simulations. b Illustration of the principal moments used to characterize the inter-domain angles. For each domain, principal moment 1, 2, and 3 (annotated using I1, I2, and I3) were respectively colored blue, purple, and gray

The MD simulations reached steady state where temperature, pressure and energy fluctuated around mean values within the first nanosecond of the simulation (data not shown). Figure 4 shows the root-mean-squared deviation (RMSD) from the starting Fab and Fc crystallographic domains during the 10 ns MD simulation aligned to backbone (N, C α , C, and O) atoms to evaluate internal fluctuations of the domains. RMSD compares the positions of atoms at two different time steps, where one of them is typically the starting structure (time zero). Therefore, the larger the RMSD the larger the changes in the positions of atoms, which usually results from motions of flexible regions. If the MD structures are not aligned to a reference, then RMSD contains additional information from overall diffusive motions. The RMSD of the Fc and Fab domains remained within 1–2.5 Å throughout the simulation, indicating fluctuations about an equilibrium structure. The plot in Fig. 4a shows the average for all residues as a function of time. Figure 4b represents the RMSD for each residue of the domains, averaged over the course of the MD simulation. These values were qualitatively compared to the temperature factors reported for the crystal structures of the individual Fc and Fab domains. Overall, the domains were stable during the simulations and the same regions in the domains were flexible in both the experimentally determined crystal structures and the simulation models.
Fig. 4

a Time evolution of the root-mean-square deviations (RMSD) in the MD simulation for Fab and Fc using the crystal structures as the reference. b Comparisons of temperature factors (crystal) and RMSD by residue referenced to the crystal (simulation) for each Fab and Fc domain

After the MD simulation, torsion-angle MC simulations were performed to obtain structures that account for the flexibility in the hinge region. By sampling backbone torsion-angles in the hinge an ensemble of 136568 structures was obtained from which scattering profiles were calculated and compared to the experimental data. Figure 5 compares the scattering profiles for all the structures obtained from the torsion-angle MC simulations to the experimental data. Although the profiles of some structures significantly differed from the data, the average profile of the ensemble agreed with the experimental data. Figure 5a represents the scattered intensity and Fig. 5b shows the pair distribution function for the ensemble of structures calculated from the coordinates using (3). Figure 5b highlights the pair distributions for the most compact and extended NISTmAb structures of the ensemble, showing the effect of conformation on the peaks in the distribution functions.
Fig. 5

a Theoretical SAXS profiles from 136568 torsion-angle MC structures (gray) compared to the experimental SAXS profile (brown). The best (lowest χ2), worst (highest χ2), and average profiles are also represented. b Pair distribution functions for the torsion-angle MC structures (gray). The R g of the structures with low and high R g were 36.3 and 56.8 Å respectively. cχ2 as a function of R g for the torsion-angle MC structures (gray dots). Starting structures for further analysis with MD are shown as red dots and indicated with arrows. Points in blue show a subset of the ensemble with closest match to the experimental SAXS profile (low χ2). d Representation of the four MD structures denoted in C

Comparison of theoretical scattering profiles to experimental small-angle scattering data was assessed by calculating the χ2 parameter as follows:
$$ \chi^{2} = {1\over{(N-1)}}{\sum\limits_{Q_{i}}{{(I_{exp}(Q_{i})-I_{model}(Q_{i}))^{2}}\over{\sigma_{exp}(Q_{i})^{2}}}}, $$
(5)
where N is the number of points, I e x p (Q i ) is the interpolated experimental intensity at Q = Q i , I m o d e l (Q i ) is the theoretically calculated intensity at Q i for the model structure, and σ e x p (Q i ) is the standard error of the mean for the experimental intensity at Q i . A lower χ2 represents a better match between the theoretical and experimental scattering intensity. Figure 5c shows χ2 as a function of R g for all structures from the torsion-angle MC simulations. From these results, a subset of low χ2 (best) structures was selected to investigate the configurations that best describe the experimental data. In addition, four structures in the extremes of the plot were highlighted: the structure with the lowest R g (MD-1), the structure with the lowest χ2 and same R g as in the experimental data (MD-2), the structure with the same R g as in the experimental data but with the highest χ2 (MD-3), and the structure with the highest R g of the ensemble (MD-4). Note that the structure with the lowest R g was also the structure with the highest χ2. A snapshot of these four structures of the ensemble is presented in Fig. 5d.

Scattering profiles and pair distribution functions of a subensemble of the best 861 structures are presented in Fig. 6a and b respectively (more information on the torsion-angle MC subensembles is presented in Electronic Supplementary Material (??)). For these structures, no major differences were observed in the scattering profiles and the average scattering profile agreed well with the best match and the experimental data. These results demonstrate that a variety of structures, and even an ensemble, can describe the experimental data and thus there was not a single structure that uniquely describes the scattering profile of the NISTmAb. Moreover, the pair distribution functions showed two peaks at ∼45 and ∼80 Å, and the position of these peaks did not differ for the most compact and extended structures of the subensemble. In this case, the main difference in the distribution functions was observed in the maximum distance of 145 Å for the compact structure compared to 170 Å for the most extended structure of the best ensemble.

The origin of the peaks in the pair distribution function can be explained by calculating the distribution functions for each domain and pairs of domains from the coordinates of structures from the simulation (3). Figure 6c presents the pair distribution functions for each Fab, Fc, Fab-Fc pairs, Fab-Fab pair, and the NISTmAb. The first peaks in the distributions at ∼30-50 Å can be attributed to intra-domain correlations, because the distributions matched those of the individual Fab and Fc domains. For the Fab-Fab and Fab-Fc pairs, the distributions showed peaks at distances comparable to those of the NISTmAb. However, since the first peak is from intra-domain distances, the second peak can be attributed to inter-domain correlations. Note that the distributions were similar regardless of the specific domain or pair, because both Fab domains were comparable in size and molecular weight to the Fc. The pair distribution function for the NISTmAb was consistent to the sum of distributions of all pairs minus the contribution of each domain, with only a minor contribution from the hinge region, which did not affect the position of the peaks (not shown).
Fig. 6

a Measured SAXS profile compared to a subensemble of 861 low χ2 structures from torsion-angle MC (gray), best (lowest χ2), worst (highest χ2 in subensemble) and average of the profiles. b Pair distribution functions for 861 simulated structures from torsion-angle MC in the ensemble of best structures (gray). The R g of the structures with low and high R g were 47.4 and 52.8 Å respectively. c Contributions from the Fab and Fc domains and the corresponding pairs to the pair distribution function. Toroids represent domains of the NISTmAb, either Fab or Fc. The first peak in P(r) represents intra-domain correlations (single toroid), whereas the second peak corresponds to inter-domain correlations (two toroids)

To explore the configuration space sampled by the NISTmAb with torsion-angle MC and MD simulations, further 10 ns MD simulations were carried out using the structures of Fig. 5d as the starting configuration. Figure 7a shows the RMSD as a function of time for the full-length NISTmAb during the four MD simulations using the structures of Fig. 5d as reference structures. For these calculations, the structures were aligned to the Fc to account for the motion of Fabs. These simulations indicated a change in RMSD up to ∼ 25 Å, with the lowest change of ∼5 Å observed for the most compact structure (MD-1). Some fluctuations in the RMSD were accompanied by changes in R g as shown in Fig. 7b, where a change of ∼5 Å occured for the most extended structure (MD-4). Although the scattering profile for the most compact structure (MD-1) significantly differed from the experimental scattering data, no significant changes in configuration were observed within a 10 ns MD simulation of the compact structure. Figure 7c compares the scattering profiles of structures from MD by calculating χ2 as a function of the size R g . Over the time scales of the MD simulations, none of the configurations converged to the same structure and only small regions of the configuration space were sampled. Moreover, most of the changes occurred in χ2 and not R g .
Fig. 7

a Time evolution of the root-mean-square displacement in four MD simulations of the full-length NISTmAb using the starting structure as the reference. b Time evolution of R g in four MD simulations of the NISTmAb. cχ2 as a function of R g for the structures from the four MD simulations compared to all the torsion-angle MC structures (gray dots). For the four MD simulations, the time evolution was represented from darker to lighter colors. The colors for each simulation are as follows: MD-1 in blue, MD-2 in green, MD-3 in purple, and MD-4 in red. See Fig. 5d for the configurations used as starting structure in each MD simulation

The configuration space sampled during the simulations can be displayed as density plots, which is a representation of the volumetric space occupied by the atoms in an ensemble of structures. Figure 8 displays density plots for each of the MD and torsion-angle MC simulations. For the torsion-angle MC structures, we defined subensembles comprising 861 structures with the following criteria: low R g , low χ2, high χ2 and high R g (see ESM for details). The torsion-angle MC simulations yielded structures that explore the widest configuration space. In contrast, during a 10 ns MD simulation, only a fraction of the configuration space was explored. The subset of best structures covered most of the configuration space sampled by all structures, even though the subset represents less than 1% of the ensemble with all torsion-angle MC structures. For the MD simulations, fluctuations within the initial configuration were observed in agreement with the results of Fig. 7.
Fig. 8

Spatial range sampled during the torsion-angle MC simulations (top), the MD simulations (left), and the torsion-angle MC simulations for regions of interest (right). In these structures, the Fc domain was used as a frame of reference and thus only the spatial range of the Fab domains is represented

Figure 9 compiles a subset of the metrics used to evaluate the relative angles and distances between Fc and Fab domains. Figure 9a presents the distribution of the center of mass distance between the Fabs and shows that the most common center of mass distance was between 80 to 90 Å, but can vary from roughly 40 to 110 Å. These results were similar for the distances calculated between the Fc and Fab center of mass, shown in Fig. 9b, c and d show the angles between the principal moments of the different domains. These distributions in particular showed a distinct difference between the high and low χ2 subensembles compared to the entire ensemble. For instance, in Fig. 9d, the distribution of the angles between the Fab and Fc principal moments for the high χ2 subensemble showed a bimodal distribution peaked around 25 and 120. In contrast, the most common angle for the low χ2 subensemble and the entire ensemble was around 70–100 and 100, respectively; this suggests that the domains were more likely to be uniformly separated compared to the structures in the high χ2 subensemble.

Figure 9e displays the distribution of angles between the vectors connecting the center of mass of the Fc to the two Fab domains, Fab-Fc-Fab. For this angle, the distribution for the entire ensemble showed a maximum at ∼60 with a range between 25 and 110, while the high and low χ2 distributions were peaked around 90. For all conditions, the distributions for the low and high χ2 subensembles did not vary as continuously as the distribution for the entire ensemble, a consequence of much fewer structures in the subensembles.
Fig. 9

Frequency distributions of Fab and Fc domains from molecular simulations. The gray, purple, and green represent the entire torsion-angle MC ensembles, the high χ2, and the low χ2 subensemble of interest, respectively. a Distance between the centers of mass of each Fab. b Distance between the center of mass of the Fc and the center of mass of each Fab. c Angle between the first principal moments of the Fabs. d Angle between the first principal moments of the each Fab and the Fc. e Angle formed by the vectors describing the centers of mass in the Fc and each Fab. f Representative structures shown from the high χ2 and low χ2 subensembles of interest. The vertical green and purple lines indicate the distance or angle values for the representative structures shown in F. Insets illustrate the distances or angles measured by each metric

Further analysis of the inter-domain spatial correlations were carried out by considering the distribution of angles between the inter domain principal moments of inertia (PMI) as defined in Fig. 3b. A plot of the angles between each 27 pairs of inter-domain PMI is shown in ??. Most angles between PMI of different domains were symmetrically distributed about a mean angle between 70 and 80 with a representative distribution shown in Fig. 10a. Interestingly, a few distributions were asymmetric including skewed (Fig. 10b) and a single bi-modal (Fig. 10c) distributions.
Fig. 10

The relative angle between select pairs of principal moments from the different domains, chosen to illustrate the different distributions (all 27 combinations of inter-domain principal moment angles are shown in ??). For each of these three distributions, we selected the 100 structures with an inter-domain angle closest to the angle represented by the vertical blue and pink line. a An example of the most common distribution of angles, resembling a Gaussian distribution. b A skewed distribution. c A uniquely bimodal distribution; this was the only one of the 27 inter-domain principal moment angle combinations with this type of distribution. For each distribution we examined 100 structures with angles indicated by the vertical blue and pink lines located at 82 (A), 134 (B), 44 (blue C), and 142 (pink C). These values were chosen to provide insight into a selection of structures at the peak or shoulder of the different distributions. The illustrations to the right of the plots depict the principal moment vectors of interest and the spatial ranges of the Fab domain considered

Analysis of the asymmetric distributions indicated that a larger cone-like envelope is sampled compared to the typical symmetric distributions (not shown). The crystal structure of an intact antibody (1HZH) chosen to build the complete model of the NISTmAb had an asymmetric orientation of the residues between the Fc and the disulfide bonds in the hinge region. The torsion-angle MC algorithm is currently unable to sample concerted moves involving parallel chains only and thus it was unclear whether including extra degrees of freedom would lead to different distributions between the PMI as shown in Fig. 10b and c. Regardless, a majority of inter-domain orientations showed a similarly symmetric set of configurations of Fc and Fab domains indicating robust sampling and a large degree of inherent flexibility.

Finally, the contacts between residues in different domains are presented in Fig. 11. If any pair of heavy atoms (excludes hydrogen atoms) from two residues on different domains were within a distance of 4 Å or less, they were considered to be in contact. The total numbers of contacts for each domain are presented in the ??. For the contact maps, the number of contacts was normalized to total number of contacts in each domain as presented in the bar scale of Fig. 11. Further details on the scale bar are described in the ESM. These contact maps indicate that residues in different domains are at distances of 4 Å or less, depending on the size R g and how well the structures described the experimental scattering data. For instance, the structures in the low R g subensemble showed significant contacts in the constant regions of the NISTmAb, whereas the contacts in the structures from the high R g subensemble were limited to residues closer to the hinge. For the low χ2 subensemble, the contacts were primarily located at the C H 1 and C H 2 regions of the NISTmAb. Note also that no contacts were observed in the variable region of the NISTmAb for the best structures. Therefore, the results suggest that, because of the flexibility in the hinge region, it is possible for antibodies to have residues from nearly all regions in contact. Nevertheless, the results suggest that in solution, minimal contacts occurred in the variable and the C H 3 regions of the antibody. Moreover, the residues close to the hinge were always in close proximity.
Fig. 11

Contact maps for the ensemble of all the MC structures and each of the 4 subensembles of interest: low χ2, high χ2, low R g , and high R g

Discussion

To obtain information about the solution structure of the NISTmAb, small-angle scattering measurements were performed at different conditions including buffers in H2O and D2O and with 0 and 150 mM NaCl. As a reference material, it is important to investigate the reproducibility of the experimental scattering profile of the NISTmAb collected with different instruments and facilities. Although the main goal of this work was not an inter-laboratory study of small-angle scattering data, the results of Fig. 2 confirmed that the data are reproducible in different facilities. No major differences were detected in the dilute scattering profiles of the NISTmAb in deuterated buffer or after adding 150 mM NaCl, compared to the 25 mM histidine buffer in H2O. Nevertheless, Guinier analysis and P(r) distributions indicated subtle differences in the size of the molecule, such as a minor decrease (∼3 Å) of the protein size in the deuterated buffer. Moreover, the addition of 150 mM NaCl increased the R g of the NISTmAb as a result of either surface adsorption of the chlorine and sodium ions, slight reversible aggregation resulting from charge screening, or both. The SEC data showed that irreversible aggregates did not form in the solution after adding NaCl and their population was less than 0.5% in all conditions, in agreement with the absence of a low-Q upturn in the scattering profiles typically seen in aggregated protein solutions [10]. Data from DLS experiments showed that the size distribution was slightly larger in the solutions with 150 mM NaCl than in the 0 mM NaCl solution. Note that DLS does not have the resolution to distinguish between monomers and oligomers such as dimers or trimers.

The analysis of SEC chromatograms confirmed that all samples were mainly monomeric during the small-angle scattering measurements. Irreversible aggregation did not occur as a consequence of adding 150 mM NaCl to the formulation buffer or exchanging from H2O to D2O buffer. Although DLS is sensitive to reversible aggregation due to changes in concentration, deviations from the Stokes-Einstein equation at high concentrations, when intermolecular interactions become important, limit the applicability to isolate reversible aggregation using DLS. The smaller hydrodynamic sizes of NISTmAb from DLS for the 0 mM NaCl compared to the 150 mM NaCl solutions are consistent with the results from the Guinier analysis and the surface adsorption of ions that screen charges on the protein surface.

By combining experimental scattering data with molecular simulations, ensembles of structures with theoretical scattering profiles that agree with experimental data were obtained. A starting model was built using crystallographic coordinates of the domains for MD and torsion-angle MC simulations. Comparison of the structures from simulation with scattering data in solution can provide quantitative insights on the flexibility in the hinge region of the NISTmAb. However, small-angle scattering alone, in general, does not have enough information content to quantitatively determine the probability of certain structures or subensembles being present in solution. Our study combined experiments and simulations of the NISTmAb, where the experimental data served to validate the models obtained from the simulations. For example, the R g obtained during the MD and torsion-angle MC simulations compared well to the experimentally determined values. The torsion-angle MC simulations explore the wide configuration space sampled by the Fabs, which cannot be achieved from 10 ns MD simulations as demonstrated by Figs. 7 and 8. Moreover, previous theoretical and experimental studies suggested that hinge bending motion of domains in an antibody molecule occur in the 100 ns to 1 μ s time scales [24, 33, 62]. Therefore, the MD simulations performed here cannot fully capture domain motions in the NISTmAb. Nevertheless, longer MD simulations are unlikely to cover the wide configuration space sampled by antibodies, as multiple microsecond MD simulations would be required. Although microsecond simulations are possible for some peptides and small biomolecules, a single NISTmAb molecule has 20669 atoms itself and more than 370000 atoms exist in solvated model systems. Therefore, combining MD with the the torsion-angle MC approach was suitable to explore the known flexibility of antibody molecules in the hinge region [2]. In addition, the wall-clock time for the torsion-angle MC simulations using a single processor was more than ten times shorter than the time for the MD simulations, that used 32 processors in parallel.

The distributions of distances and angles showed that a wide range of values are likely to occur. These results were in agreement with a previous study on an IgG2 antibody, both of which found that a range of structural configurations are plausible in solution [12]. Moreover, the resulting distributions are comparable to those obtained by AFM and individual particle electron tomography [11, 64]. Although subtle differences were observed in the quantitative values for distances and angles, both the electron tomography study and ours found a variation up to ∼60 Å in the distance between the domains and an angle variation from ∼20 to ∼100. Overall, these results are consistent with a broad conformational flexibility between the domains, which explains the difficulty to obtain crystallographic structures of intact antibodies.

Conclusions

Small-angle scattering is a powerful technique to study the conformational flexibility, structure and protein-protein interactions of therapeutic proteins in a wide range of conditions. The scattering profiles of the NISTmAb using SANS and SAXS were reproduced in different instruments and facilities. No major changes in the conformation were observed after adding 150 mM NaCl, or changing the histidine buffer from H2O to D2O, with the exception of a few Å change in the size as determined by R g . No significant irreversible aggregation is detected and the scattering profiles showed no evidence of aggregates smaller than 0.2 μ m, which corresponds to the maximum length scale probed in the SANS experiments.

Small-angle scattering experiments combined with molecular simulations provided important insights on the conformation of the NISTmAb in solution. MD simulations in explicit water were combined with torsion-angle MC simulations to obtain ensembles of structures that explore the large configuration space sampled by antibodies. From these structures, scattering profiles were calculated and compared with measured scattering profiles. Obtaining an ensemble of structures that explore the flexibility of antibodies was only possible with a combined MD and torsion-angle MC simulation approach. With current computational resources, MD simulations of full antibody structures were not suitable to explore the flexibility of the NISTmAb domains mediated by the hinge region.

From the comparisons with experimental scattering data, a unique structure that described the experimental scattering profile was not found. On the contrary, ensembles of structures provided an average scattering profile that agreed with the experimental data in solution. The structures that best described the experimental data sampled a wide range of configuration space and distributions of angles and distances between the Fab and Fc domains. Further experimental constraints or free energy calculations could be performed in the future to further refine the ensemble of flexible NISTmAb structures.

Footnotes

  1. 1.

    Certain commercial equipment, instruments, materials, suppliers, or software are identified in this paper to foster understanding. Such identification does not imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply that the materials or equipment identified are necessarily the best available for the purpose.

Notes

Acknowledgements

The authors acknowledge the following scientists for supporting this work: Emilie Poudevigne (previously affiliated with ESRF) for collecting SAXS data at the ESRF, Zhiyuan Wang (Tsinghua University) and Yun Liu (NIST, University of Delaware) for their help with SANS data collection, John Schiel (NIST, IBBR) for making the material available for this study and for helping with the SEC measurements, James Snyder (NIST) for performing the energy minimization of the resulting structures from the torsion-angle MC simulation. MMC acknowledges financial support from the NIST biomanufacturing initiative. SCH acknowledges financial support from the NIST NRC Postdoctoral Research Associateship Program. This work used CCP-SAS software developed through a joint EPSRC (EP/K039121/1) and NSF (CHE-1265821) grant.

Compliance with Ethical Standards

Conflict of interests

All authors of this article declare no conflict of interest.

Supplementary material

216_2018_868_MOESM1_ESM.pdf (1.3 mb)
(PDF 1.25 MB )

References

  1. 1.
    Abe Y, Gor J, Bracewell DG, Perkins SJ, Dalby PA. Masking of the fc region in human igg4 by constrained x-ray scattering modelling: implications for antibody function and therapy. Biochem J 2010;432(1): 101–114.CrossRefGoogle Scholar
  2. 2.
    Arata Y, Honzawa M, Shimizu A. Proton nuclear magnetic resonance studies of human immunoglobulins: conformation of the hinge region of the igg1 immunoglobulin. Biochemistry 1980;19(22):5130–5135. PMID: 6257277.CrossRefGoogle Scholar
  3. 3.
    Ashish SAK, Boone CD, Krueger JK. Global structure of hiv-1 neutralizing antibody igg1 b12 is asymmetric. Biochem Biophys Res Commun 2010;391(1):947–951.CrossRefGoogle Scholar
  4. 4.
    Barnett GV, Qi W, Amin S, Lewis EN, Razinkov V, Kerwin B, Liu Y, Roberts CJ. Structural changes and aggregation mechanisms for anti-streptavidin igg1 at elevated concentration. J Phys Chem B 2015;119(49):15,150–15,163. PMID: 26563591.CrossRefGoogle Scholar
  5. 5.
    Barnett GV, Razinkov V, Kerwin B, Laue T, Woodka AH, Butler PD, Perevozchikova T, Roberts CJ. Specific-ion effects on the aggregation mechanisms and protein–protein interactions for anti-streptavidin immunoglobulin gamma-1. J Phys Chem B 2015;119(18):5793–5804. PMID: 25885209.CrossRefGoogle Scholar
  6. 6.
    Brookes EH, Anjum N, Curtis JE, Marru S, Singh R, Pierce M. Genapp module execution and airavata integration. 2014 9th gateway computing environments workshop; 2014. p. 9–12.Google Scholar
  7. 7.
    Castellanos MM, Mattison K, Krueger S, Curtis JE. Characterization of the NISTmAb Reference Material using small-angle scattering and molecular simulation. Part II: Concentrated protein solutions. Anal Bioanal Chem. (2018);  https://doi.org/10.1007/s00216-018-0869-1.
  8. 8.
    Castellanos MM, McAuley A, Curtis JE. Investigating structure and dynamics of proteins in amorphous phases using neutron scattering. Comput Struct Biotechnol J 2016;15:117–130.CrossRefGoogle Scholar
  9. 9.
    Castellanos MM, Pathak JA, Colby RH. Both protein adsorption and aggregation contribute to shear yielding and viscosity increase in protein solutions. Soft Matter 2014;10(1):122–131.CrossRefGoogle Scholar
  10. 10.
    Castellanos MM, Pathak JA, Leach W, Bishop SM, Colby RH. Explaining the non-newtonian character of aggregating monoclonal antibody solutions using small-angle neutron scattering. Biophys J 2014;107(2): 469–476.CrossRefGoogle Scholar
  11. 11.
    Chaves RC, Teulon JM, Odorico M, Parot P, Chen SWW, Pellequer JL. Conformational dynamics of individual antibodies using computational docking and afm. J Mol Recognit 2013;26 (11):596–604.CrossRefGoogle Scholar
  12. 12.
    Clark NJ, Zhang H, Krueger S, Lee HJ, Ketchem RR, Kerwin B, Kanapuram SR, Treuheit MJ, McAuley A, Curtis JE. Small-angle neutron scattering study of a monoclonal antibody using free-energy constraints. J Phys Chem B 2013;117(45):14,029–14,038.CrossRefGoogle Scholar
  13. 13.
  14. 14.
    Curtis JE, Raghunandan S, Nanda H, Krueger S. Sassie: a program to study intrinsically disordered biological molecules and macromolecular ensembles using experimental scattering restraints. Comput Phys Commun 2012;183(2):382–389.CrossRefGoogle Scholar
  15. 15.
    Darden T, York D, Pedersen L. Particle mesh ewald: an n.log(n) method for ewald sums in large systems. J Chem Phys 1993;98 (12):10,089–10,092.CrossRefGoogle Scholar
  16. 16.
    Debye P. Zerstreuung von röntgenstrahlen. Ann Phys 1915;351(6):809–823.CrossRefGoogle Scholar
  17. 17.
    Dimitrov DS. Therapeutic proteins. Therapeutic proteins, methods in molecular biology, vol 899, book section 1. In: Voynov V and Caravella JA, editors. New York: Humana Press; 2010. p. 1–26.Google Scholar
  18. 18.
    Formolo T, Ly M, Levy M, Kilpatrick L, Lute S, Phinney K, Marzilli L, Brorson K, Boyne M, Davis D, Schiel J. Determination of the nistmab primary structure. State-of-the-Art and Emerging Technologies for Therapeutic Monoclonal Antibody Characterization Volume 2. Biopharmaceutical Characterization: the NISTmAb Case Study, chap. 1, pp. 1–62. ACS Symposium Series. In: Schiel JE, Davis DL, and Borisov OV, editors; 2015.Google Scholar
  19. 19.
    Godfrin PD, Zarraga IE, Zarzar J, Porcar L, Falus P, Wagner NJ, Liu Y. Effect of hierarchical cluster formation on the viscosity of concentrated monoclonal antibody formulations studied by neutron scattering. J Phys Chem B 2016;120(2):278–291.CrossRefGoogle Scholar
  20. 20.
    Gokarn Y, Agarwal S, Arthur K, Bepperling A, Day ES, Filoti D, Greene DG, Hayes D, Kroe-Barrett R, Laue T, Lin J, McGarry B, Razinkov V, Singh S, Taing R, Venkataramani S, Weiss W, Yang D, Zarraga IE. 2015. Biophysical Techniques for Characterizing the Higher Order Structure and Interactions of Monoclonal Antibodies, chap. 6, pp. 285–327 ACS Symposium Series.Google Scholar
  21. 21.
    Grillo I. Small-angle neutron scattering and applications in soft condensed matter. Soft matter characterization, chap. 13. In: Borsali R and Pecora R, editors. Dordrecht: Springer; 2008. p. 723–782.Google Scholar
  22. 22.
    Guinier A. Théorie et technique de la radiocristallographie. Paris: Dunod; 1956.Google Scholar
  23. 23.
    Hammouda B. 2016. Probing nanoscale structures-the sans toolbox. Natl. Institute Standards Technology Center for Neutron Research, Gaithersburg, MD p. https://www.ncnr.nist.gov/staff/hammouda/theSANStoolbox.pdf.
  24. 24.
    Hanson DC, Yguerabide J, Schumaker VN. Segmental flexibility of immunoglobulin g antibody molecules in solution: a new interpretation. Biochemistry 1981;20(24):6842–6852. PMID: 7317358.CrossRefGoogle Scholar
  25. 25.
    Humphrey W, Dalke A, Schulten K. Vmd: visual molecular dynamics. J Mol Graphics Modell 1996;14 (1):33–38.CrossRefGoogle Scholar
  26. 26.
    Inouye H, Houde D, Temel DB, Makowski L. Utility of solution x-ray scattering for the development of antibody biopharmaceuticals. J Pharm Sci 2016;105(11):3278–3289.CrossRefGoogle Scholar
  27. 27.
    Jacrot B. The study of biological structures by neutron scattering from solution. Rep Prog Phys 1976;39(10): 911.CrossRefGoogle Scholar
  28. 28.
    Jordan A, Jacques M, Merrick C, Devos J, Forsyth VT, Porcar L, Martel A. SEC-SANS: Size exclusion chromatography combined in situ with small-angle neutron scattering. J Appl Crystallogr 2016;49(6): 2015–2020.CrossRefGoogle Scholar
  29. 29.
    Jorgensen WL, Chandrasekhar J, Madura JD, Impey RW, Klein ML. Comparison of simple potential functions for simulating liquid water. J Chem Phys 1983;79(2):926–935.CrossRefGoogle Scholar
  30. 30.
    Kline S. Reduction and analysis of sans and usans data using igor pro. J Appl Crystallogr 2006;39(6):895–900.CrossRefGoogle Scholar
  31. 31.
    Lilyestrom WG, Shire SJ, Scherer TM. Influence of the cosolute environment on igg solution structure analyzed by small-angle x-ray scattering. J Phys Chem B 2012;116(32):9611–9618.CrossRefGoogle Scholar
  32. 32.
    Mathew E, Mirza A, Menhart N. Liquid-chromatography-coupled SAXS for accurate sizing of aggregating proteins. J Synchrotron Rad 2004;11(4):314–318.CrossRefGoogle Scholar
  33. 33.
    McCammon JA, Karplus M. Internal motions of antibody molecules. Nature 1977;268(5622):765–766.CrossRefGoogle Scholar
  34. 34.
    Mosbæk CR, Konarev PV, Svergun DI, Rischel C, Vestergaard B. High concentration formulation studies of an igg2 antibody using small angle x-ray scattering. Pharm Res 2012;29(8):2225–2235.CrossRefGoogle Scholar
  35. 35.
    Nielsen SS, Toft KN, Snakenborg D, Jeppesen MG, Jacobsen JK, Vestergaard B, Kutter JP, Arleth L. BioXTAS RAW, a software program for high-throughput automated small-angle X-ray scattering data reduction and preliminary analysis. J Appl Crystallogr 2009;42(5):959–964.CrossRefGoogle Scholar
  36. 36.
    Orthaber D, Bergmann A, Glatter O. SAXS Experiments on absolute scale with Kratky systems using water as a secondary standard. J Appl Crystallogr 2000;33(2):218–225.CrossRefGoogle Scholar
  37. 37.
    Perkins SJ, Bonner A. Structure determinations of human and chimaeric antibodies by solution scattering and constrained molecular modelling. Biochem Soc Trans 2008;36(1):37–42.CrossRefGoogle Scholar
  38. 38.
    Perkins SJ, Okemefuna AI, Nan R, Li K, Bonner A. Constrained solution scattering modelling of human antibodies and complement proteins reveals novel biological insights. J Royal Soc Inter 2009;6(Suppl 5): S679–S696.CrossRefGoogle Scholar
  39. 39.
    Perkins SJ, Wright DW, Zhang H, Brookes EH, Chen J, Irving TC, Krueger S, Barlow DJ, Edler KJ, Scott DJ, Terrill NJ, King SM, Butler PD, Curtis JE. Atomistic modelling of scattering data in the Collaborative Computational Project for Small Angle Scattering (CCP-SAS). J Appl Crystallogr 2016;49 (6):1861–1875.CrossRefGoogle Scholar
  40. 40.
    Pernot P, Round A, Barrett R, De Maria Antolinos A, Gobbo A, Gordon E, Huet J, Kieffer J, Lentini M, Mattenet M, Morawe C, Mueller-Dieckmann C, Ohlsson S, Schmid W, Surr J, Theveneau P, Zerrad L, McSweeney S. Upgraded ESRF BM29 beamline for SAXS on macromolecules in solution. J Synchrotron Rad 2013;20(4):660–664.CrossRefGoogle Scholar
  41. 41.
    Phillips JC, Braun R, Wang W, Gumbart J, Tajkhorshid E, Villa E, Chipot C, Skeel RD, Kalé L, Schulten K. Scalable molecular dynamics with namd. J Comput Chem 2005;26(16):1781–1802.CrossRefGoogle Scholar
  42. 42.
    Prien JM, Stöckmann H, Albrecht S, Martin SM, Varatta M, Furtado M, Hosselet S, Wang M, Formolo T, Rudd PM, Schiel JE. Orthogonal technologies for nistmab n-glycan structure elucidation and quantitation. State-of-the-Art and Emerging Technologies for Therapeutic Monoclonal Antibody Characterization Volume 2. Biopharmaceutical Characterization: The NISTmAb Case Study, chap. 4, pp. 185–235. ACS Symposium Series. In: Schiel JE, Davis DL, and Borisov OV, editors; 2015.Google Scholar
  43. 43.
    Putnam CD, Hammel M, Hura GL, Tainer JA. X-ray solution scattering (saxs) combined with crystallography and computation: defining accurate macromolecular structures, conformations and assemblies in solution. Q Rev Biophys 2007;40(3):191– 285.CrossRefGoogle Scholar
  44. 44.
  45. 45.
    Rayner LE, Hui GK, Gor J, Heenan RK, Dalby PA, Perkins SJ. The solution structures of two human igg1 antibodies show conformational stability and accommodate their c1q and fcγ r ligands. J Biol Chem 2015;290(13):8420–8438.CrossRefGoogle Scholar
  46. 46.
    Reichert JM. Marketed therapeutic antibodies compendium. MAbs 2012;4(3):413–415.CrossRefGoogle Scholar
  47. 47.
    Reichert JM. Antibodies to watch in 2017. MAbs 2017;9(2):167–181. PMID: 27960628.CrossRefGoogle Scholar
  48. 48.
    Ryckaert JP, Ciccotti G, Berendsen HJC. Numerical integration of the cartesian equations of motion of a system with constraints: molecular dynamics of n-alkanes. J Comput Phys 1977;23(3):327–341.CrossRefGoogle Scholar
  49. 49.
    Saphire EO, Parren PWHI, Barbas Carlos FI, Burton DR, Wilson IA. Crystallization and preliminary structure determination of an intact human immunoglobulin, b12: an antibody that broadly neutralizes primary isolates of hiv-1. Acta Crystallogr, Sect D 2001;57(1):168–171.CrossRefGoogle Scholar
  50. 50.
    Schiel JE, Davis D, Borisov OV. State-of-the-art and emerging technologies for therapeutic monoclonal antibody characterization volume 1. monoclonal antibody therapeutics: structure, function, and regulatory space. State-of-the-Art and Emerging Technologies for Therapeutic Monoclonal Antibody Characterization Volume 1. Monoclonal Antibody Therapeutics: Structure, Function, and Regulatory Space, vol. 1, pp. ix–xi. ACS Symposium Series. In: Schiel JE, Davis DL, and Borisov OV, editors; 2014.Google Scholar
  51. 51.
    Schiel JE, Davis D, Borisov OV. State-of-the-art and emerging technologies for therapeutic monoclonal antibody characterization volume 2. biopharmaceutical characterization: The nistmab case study. State-of-the-art and emerging technologies for therapeutic monoclonal antibody characterization volume 2. Biopharmaceutical Characterization: The NISTmAb Case Study, pp. ix–xii. ACS Symposium Series. In: Schiel JE, Davis DL, and Borisov OV, editors; 2015.Google Scholar
  52. 52.
    Schiel JE, Davis D, Borisov OV. State-of-the-art and emerging technologies for therapeutic monoclonal antibody characterization volume 3. defining the next generation of analytical and biophysical techniques. State-of-the-Art and Emerging Technologies for Therapeutic Monoclonal Antibody Characterization Volume 3. Defining the Next Generation of Analytical and Biophysical Techniques, pp. ix–x. ACS Symposium Series. In: Schiel JE, Davis DL, and Borisov OV, editors; 2015.Google Scholar
  53. 53.
    Schiel JE, Mire-Sluis A, Davis D. Monoclonal antibody therapeutics: The need for biopharmaceutical reference materials. State-of-the-Art and Emerging Technologies for Therapeutic Monoclonal Antibody Characterization Volume 1. Monoclonal Antibody Therapeutics: Structure, Function, and Regulatory Space, vol. 1, chap. 1, pp. 1–34. ACS Symposium Series. In: Schiel JE, Davis DL, and Borisov OV, editors; 2014.Google Scholar
  54. 54.
    Stuhrmann HB. Small-angle scattering and its interplay with crystallography, contrast variation in SAXS and SANS. Acta Crystallographica Sect A 2008;64(1):181–191.CrossRefGoogle Scholar
  55. 55.
    Svergun DI, Koch MHJ. Small-angle scattering studies of biological macromolecules in solution. Rep Prog Phys 2003;66(10):1735–1782.CrossRefGoogle Scholar
  56. 56.
    Tian X, Langkilde AE, Thorolfsson M, Rasmussen HB, Vestergaard B. Small-angle x-ray scattering screening complements conventional biophysical analysis: Comparative structural and biophysical analysis of monoclonal antibodies igg1, igg2, and igg4. J Pharm Sci 2014;103(6):1701–1710.CrossRefGoogle Scholar
  57. 57.
    Tian X, Vestergaard B, Thorolfsson M, Yang Z, Rasmussen HB, Langkilde AE. In-depth analysis of subclass-specific conformational preferences of igg antibodies. IUCrJ 2015;2(Pt 1):9–18.CrossRefGoogle Scholar
  58. 58.
    Vanommeslaeghe K, Hatcher E, Acharya C, Kundu S, Zhong S, Shim J, Darian E, Guvench O, Lopes P, Vorobyov I, Mackerell AD. Charmm general force field: a force field for drug-like molecules compatible with the charmm all-atom additive biological force fields. J Comput Chem 2009;31(4):671–690.Google Scholar
  59. 59.
    Watson MC, Curtis JE. Rapid and accurate calculation of small-angle scattering profiles using the golden ratio. J Appl Crystallogr 2013;46(4):1171–1177.CrossRefGoogle Scholar
  60. 60.
    Wilhelm P, Friguet B, Djavadi-Ohaniance L, Pilz I, Goldberg ME. Epitope localization in antigen-monoclonal-antibody complexes by small-angle x-ray scattering. Eur J Biochem 1987;164 (1):103–109.CrossRefGoogle Scholar
  61. 61.
    Yearley EJ, Zarraga IE, Shire SJ, Scherer TM, Gokarn Y, Wagner NJ, Liu Y. Small-angle neutron scattering characterization of monoclonal antibody conformations and interactions at high concentrations. Biophys J 2013;105(3):720–731.CrossRefGoogle Scholar
  62. 62.
    Yguerabide J, Epstein HF, Stryer L. Segmental flexibility in an antibody molecule. J Mol Biol 1970;51 (3):573–590.CrossRefGoogle Scholar
  63. 63.
    Zaccai NR, Sandlin CW, Hoopes JT, Curtis JE, Fleming PJ, Fleming KG, Krueger S. Chapter six - deuterium labeling together with contrast variation small-angle neutron scattering suggests how skp captures and releases unfolded outer membrane proteins. Isotope labeling of biomolecules - applications, methods in enzymology. In: Kelman Z, editors. Academic Press; 2016. p. 159–210.Google Scholar
  64. 64.
    Zhang X, Zhang L, Tong H, Peng B, Rames MJ, Zhang S, Ren G. 2015. 3d structural fluctuation of igg1 antibody revealed by individual particle electron tomography. Sci. Rep. 5 9803 EP.Google Scholar

Copyright information

© This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply 2018

Authors and Affiliations

  1. 1.NIST Center for Neutron ResearchNational Institute of Standards and TechnologyGaithersburgUSA
  2. 2.Institute for Bioscience and Biotechnology ResearchUniversity of MarylandRockvilleUSA
  3. 3.Material Measurement LaboratoryNational Institute of Standards and TechnologyGaithersburgUSA
  4. 4.Institute for Bioscience and Biotechnology ResearchRockvilleUSA
  5. 5.Drug Product Development US, GSK VaccinesRockvilleUSA

Personalised recommendations