Solution structure of deglycosylated human IgG1 shows the role of C_H2 glycans in its conformation

Abstract

The human immunoglobulin G (IgG) class is the most prevalent antibody in serum, with the IgG1 subclass being the most abundant. IgG1 is composed of two Fab regions connected to a Fc region through a 15-residue hinge peptide. Two glycan chains are conserved in the Fc region in IgG; however, their importance for the structure of intact IgG1 has remained unclear. Here, we subjected glycosylated and deglycosylated monoclonal human IgG1 (designated as A33) to a comparative multidisciplinary structural study of both forms. After deglycosylation using peptide:N-glycosidase F, analytical ultracentrifugation showed that IgG1 remained monomeric and the sedimentation coefficients s⁰_20,w of IgG1 decreased from 6.45 S by 0.16–0.27 S. This change was attributed to the reduction in mass after glycan removal. X-ray and neutron scattering revealed changes in the Guinier structural parameters after deglycosylation. Although the radius of gyration (R_G) was unchanged, the cross-sectional radius of gyration (R_XS-1) increased by 0.1 nm, and the commonly occurring distance peak M2 of the distance distribution curve P(r) increased by 0.4 nm. These changes revealed that the Fab-Fc separation in IgG1 was perturbed after deglycosylation. To explain these changes, atomistic scattering modeling based on Monte Carlo simulations resulted in 123,284 and 119,191 trial structures for glycosylated and deglycosylated IgG1 respectively. From these, 100 x-ray and neutron best-fit models were determined. For these, principal component analyses identified five groups of structural conformations that were different for glycosylated and deglycosylated IgG1. The Fc region in glycosylated IgG1 showed a restricted range of conformations relative to the Fab regions, whereas the Fc region in deglycosylated IgG1 showed a broader conformational spectrum. These more variable Fc conformations account for the loss of binding to the Fcγ receptor in deglycosylated IgG1.

Significance

Human immunoglobulin G subclass 1 (IgG1) antibody possesses two conserved glycans in its Fc region with unknown structural significance. Here, we established the role of the glycans in the overall structure of human IgG1. First, analytical ultracentrifugation revealed monomeric structures after enzymatic glycan removal, showing these were unaffected. Next, x-ray and neutron scattering revealed observable conformational changes in IgG1 after glycan removal. Atomistic Monte Carlo modeling fits of the IgG1 scattering curves showed that the best-fit structures after deglycosylation were different from the glycosylated best-fit structures. The Fc region occupied more conformational space. This greater flexibility after deglycosylation reveals the importance of the glycans in stabilizing the Fc regions and affects the way in which the Fc region interacts with its Fc receptors.

Introduction

Immunoglobulins are an important class of humoral (adaptive) glycoproteins, comprising 82–96% protein and 4–18% carbohydrate (1). The most abundant immunoglobulin class in human serum is immunoglobulin G (IgG), a Y-shaped molecule that is found as four subclasses, namely IgG1, IgG2, IgG3, and IgG4, which occur at 8.0, 4.0, 0.8, and 0.4 mg/mL in serum (2). The structure of IgG1 is formed as two Fab regions, which bind with high specificity and affinity to a specific antigen, and an Fc region that binds to Fcγ receptors (FcγRs) on the surface of immune cells, and to complement C1q to initiate the classical pathway of activation (Fig. 1 A). In IgG1, the Fabs and the Fc regions are connected by a 15-residue hinge held together by two disulphide bonds at Cys²²⁶ and Cys²²⁹ in the hinge (1). As well as being the most abundant class in serum, IgG1 is the predominant class used in therapeutic antibodies, where 54 IgG1 monoclonal antibodies are commercially available out of a total of 83 antibody-based products in a multibillion-dollar industry (3).

Figure 1.

The human IgG1 domains and its glycosylation. (A) The heavy chains are composed of V_H, C_H1, C_H2, and C_H3 domains, and the light chains are comprised of V_L and C_L domains. The heavy chains are connected by two Cys-Cys disulphide bridges at Cys²²⁶ and Cys²²⁹. An N-linked oligosaccharide at Asn²⁹⁷ is present on each of the C_H2 domains. The hinge region connecting the Fab and Fc regions was constructed from 23 residues ²¹⁶EPKSCDKTHTCPPCPAPELLGGP²³⁸. (B) At the left, the glycosylation of IgG1 Fc at two Asn²⁹⁷ residues in the Fc region is shown as stick models. The three hinge tripeptides that were conformationally varied in the TAMC searches are in red and circled in red. The central schematic shows the glycosylation pattern used in this study. At the right, the detailed view of a single C_H2 domain with its glycan chain is shown, with the glycan colors coordinated with those in the schematic. Gal, galactose; GlcNAc, N-acetyl glucosamine; Man, mannose; NeuNAc, N-acetyl neuraminic acid. To see this figure in color, go online.

The conserved N-linked glycosylation in the Fc region plays a key functional role in all four IgG subclasses (Fig. 1 B). A complex-type biantennary glycan with a Man₃GlcNAc₂ core and two NeuNAc.Gal.GlcNAc antennae is attached at Asn²⁹⁷ on each C_H2 domain (Fig. 2; (5)). However, the glycan structure is chemically heterogenous (6). The Fc glycans modulate several IgG-Fc effector functions (7). Glycoengineering is becoming increasingly important to elicit desired responses. For example, afucosylated IgG1 is able to activate a natural killer antibody-dependent cellular cytotoxicity response more effectively for the reason of its increased affinity for FcγRIIIa (8,9). Deglycosylated antibodies may be good candidates for therapeutics if a lower propensity to activate inflammatory cascades is desirable because the removal of glycan reduces the IgG interactions with Fc receptors (10). Thus, aglycosylated and deglycosylated IgG1 have an abrogated or reduced propensity for binding to the FcγRs and C1q but does not affect Fab antigen binding (11, 12, 13, 14). Deglycosylated antibodies have been of interest to treat autoimmune disorders (15,16) and to suppress immune complex-mediated inflammation in a mouse arthritis model by disrupting Fc-Fc interactions while maintaining intact antigen-antibody binding and complement binding (17). The deglycosylation of pathogen neuromyelitis optica antiaquaporin-4 IgG in patients reduced its complement-dependent and antibody-dependent cell-mediated cytotoxicity with a reduction in antigen binding, giving it therapeutic potential (18).

Figure 2.

Sequence alignment of human IgG1. (A–G) The IgG1 A33 sequence was kindly provided by UCB Pharma. The IgG1 6a and 19a sequences were taken from (4). The IgG1 b12 sequence was taken from the crystal structure (PDB: 1HZH). The Fc sequence was taken from the crystal structure 4W4N (PDB: 4W4N). (A and B) the V_L and C_L domains; (C–E) the V_H and C_H1 domains and the hinge, with the TAMC-varied tripeptides identified in green; (F and G) the C_H2 and C_H3 domains, with Asn²⁹⁷ in blue. The commonly-used EU sequence numbering for the constant domains was used and the CDR (complementarity-determining region) sequences were identified in red. Beneath the alignments, consensus symbols indicated the degree of conservation, where the asterisk indicates full conservation, the colon indicates conservation between groups of strongly similar properties based on the Gonnet PAM 250 matrix, the period indicates conservation between groups of weakly similar properties, and a space indicates no conservation. To see this figure in color, go online.

Structural studies of IgG antibodies are crucial to understand their function. Although many crystal structures are known for Fab regions, only one crystal structure is known for the full-length human IgG1 (Protein Data Bank, PDB: 1HZH) (19), together with other full-length structures for two murine IgG subclasses (20). These crystal structures only provide a static view of full-length IgG1, and do not take into proper account a mobile and flexible hinge region, which allows for the independent movement of the Fab and Fc regions in solution (21). Crystal structures for glycosylated human Fc regions revealed the two glycans to be found at the center of the Fc region in contact with each other (PDB: 4W4N, 4KU1, 4BM7, 3AVE, 4Q74, 4BYH, 1H3X) (22, 23, 24, 25, 26, 27, 28). Crystal structures of the Fc region in complexes with the receptors FcγRI and FcγRIII showed similar modes of receptor binding to the upper part of the Fc region with many conserved contacts, despite their varying affinities to the Fc region (22,29,30). In these structures, very few contacts were found between the glycans and the FcγR, and there is no information on the full IgG1 structure after deglycosylation, making it unclear what role the glycans have. Nonetheless, crystal structures for deglycosylated Fc regions showed a more compact conformation of the C_H2 domains compared with the glycosylated Fc region, indicating that the glycans stabilize the Fc regions (28). In the deglycosylated Fc structures, the C′E loop (Gln²⁹³-Phe³⁰³) of the C_H2 domain that is involved in FcγR binding is more disordered (31). Previous solution studies of glycosylated and deglycosylated human Fc gave a larger radius of gyration (R_G) for deglycosylated Fc than that of glycosylated Fc (31). These studies suggest that Asn²⁹⁷ glycosylation is important to stabilize the open conformation of the C_H2 domains.

The effect of the Fc glycan chains on the full IgG structure is not well understood. To address this question, small-angle x-ray scattering (SAXS), small-angle neutron scattering (SANS), and analytical ultracentrifugation (AUC) were jointly applied to intact IgG1 as powerful solution structural techniques for studying biological macromolecules. SAXS provides data sets measured in high-positive solute-solvent contrast in which the contribution of the hydrophilic surface regions of the glycoprotein are accentuated, whereas SANS measured with heavy water buffers provides data sets measured in high-negative solute-solvent contrast in which the contribution of the buried hydrophobic core of the glycoprotein is accentuated (4,32,33). The tightly-bound hydration layer is detected by SAXS because its electron density is similar to that of the protein and not to bulk water, whereas this same hydration layer is almost invisible by SANS measured in heavy water, because its nuclear density is almost the same as that of bulk water. The reproducibility of the two data sets corroborates the individual SAXS and SANS data sets because radiation effects in SAXS and aggregation in heavy water by SANS may perturb the output of either method. Their utility is much enhanced by the development of atomistic modeling of the SAXS and SANS data sets using molecular dynamics and Monte Carlo (MC) methods (34). Previous atomistic scattering modeling with glycosylated IgG1 revealed that IgG1 is conformationally stable, even in different buffer conditions, and exhibited an asymmetric IgG1 structure in which the arrangement of the Fab regions permitted the Fc region to bind to its FcγR and C1q ligands with no steric clashes (35,36). Here, we apply this joint SAXS-SANS-AUC approach together with atomistic modeling to show that deglycosylation does in fact result in a more flexible Fc structure within IgG1, in turn affecting the receptor-binding function of IgG1.

Materials and methods

Purification and composition of IgG1

IgG1 A33 (148 kDa) was generously supplied by Dr. John O’Hara and Dr. Berni Sweeney (Union Chimique Belge (UCB) Pharma Ltd., Slough, Berkshire, UK). Its enzymatic deglycosylation was performed using peptide:N-glycosidase F (PNGase F) (35.5 kDa; New England Biolabs, Ipswich, MA) for reason of its ability to remove glycans completely from glycosylated Asn residues (37). To digest the glycans, 3.7 μL PNGase F (1850 activity units) was used to deglycosylate 150 μL IgG1 A33 (16.3 mg/mL). Native IgG1 was incubated at 37°C for time points of 1 h (TP1), 6 h (TP6) and 10 h (TP10). Each deglycosylated IgG1 sample was filtered through three successive dilutions using Amicon Ultra 0.5-mL centrifugal filters (100-kDa cutoff), which simultaneously allowed the PNGase F to pass through the membrane, while concentrating the deglycosylated IgG1 sample. Immediately before SAXS, SANS, or AUC measurements, glycosylated and deglycosylated IgG1 were purified by gel filtration to remove any nonspecific aggregates using a Superose 6 Increase 10/300 GL column (Cytiva, Amersham, UK), then concentrated using Amicon Ultra 15-mL spin concentrators (100-kDa cutoff) and dialyzed at 4°C into 20 mM L-histidine, 138 mM NaCl, and 2.6 mM KCl buffer (pH 6.0). This histidine buffer was found to increase the stability of IgG1. The sequence of IgG1 A33 was aligned against those for IgG1 19a, 6a, and b12 (PDB: 1HZH), and an IgG1 Fc structure (PDB: 4W4N) (19,22,35; (Fig. 2)). The N-linked glycans at Asn²⁹⁷ on the C_H2 domains were approximated as complex-type biantennary structures with an Man₃GlcNAc₂ core and two NeuNAc.Gal.GlcNAc antennae (5). From this sequence, the molecular mass of glycosylated IgG1 A33 was calculated to be 148.4 kDa, its unhydrated volume was 191.4 nm³, its hydrated volume was 252.0 nm³ (based on 0.3 g of water/g of glycoprotein and an electrostrictive volume of 0.0245 nm³ per bound water molecule), its partial specific volume $\overline{v}$ was 0.731 mL/g, and its absorption coefficient was 14.0 (1%, 1-cm pathlength, 280 nm) (32). The molecular mass of deglycosylated IgG1 A33 was 144.0 kDa, its unhydrated volume was 186.7 nm³, its hydrated volume was 245.4 nm³, its partial specific volume $\overline{v}$ was 0.733 mL/g, and its absorption coefficient was 14.4 (32). The x-ray- and neutron-scattering densities of glycan residues are similar to those for hydrophilic (polar) amino acid residues, these being slightly higher than those for hydrophobic (nonpolar) amino acid residues (32). The buffer density was measured on an Anton Paar DMA 5000 density meter (Anton Paar, Graz, Austria) at 20°C to be 1.00578 g/mL in light water. In heavy water, the density was 1.11106 g/mL. Buffer viscosities were measured on an Anton Paar AMVn Automated microviscometer at 20°C. The viscosity in light water (pH 6.0) was 0.010190 P.

The completeness of deglycosylation was verified by Superose 6 gel filtration, SDS-PAGE, and mass spectrometry. In the Mass Spectrometry Facility at the Chemistry Department University College London (London, UK), the antibodies were analyzed on an Agilent 6510 Quadrupole time-of-flight liquid chromatography mass spectrometry system (Agilent Technologies, Manchester, UK). 10 μL of each sample was injected onto a PLRP-S, 1000 A, 8 μM, 150 mm × 2.1 mm column, which was maintained at 60°C at a flow of 0.3 mL/min. The separation was achieved using mobile phase A (water with 0.1% formic acid) and B (acetonitrile, with 0.1% formic acid) using a gradient elution. The column effluent was continuously electrosprayed into the capillary electrospray ionization source of the Agilent 6510 QTOF mass spectrometer, and electrospray ionization mass spectra were acquired in positive electrospray ionization mode using the m/z range 1000−3200 in profile mode. The raw data were converted to zero-charge mass spectra using the maximal entropy deconvolution algorithm in the MassHunter software version B.07.00. The glycan masses were calculated by subtracting the mass of the full glycosylated IgG1 from the partially deglycosylated glycoform giving the mass of a single glycan chain. The single glycan mass was also found by subtracting the mass of fully glycosylated IgG1 from that for deglycosylated IgG1 and halving this mass.

Sedimentation velocity data and analysis for IgG1

AUC data for native and deglycosylated IgG1 in light water at time points TP1, TP6, and TP10 were obtained on two Beckman XL-I instruments equipped with AnTi 50 rotors. Data were collected at 20°C at a rotor speed of 40,000 rpm in two sector cells with column heights of 12 mm for ∼6 h. Sedimentation analyses were performed using direct boundary Lamm fits of up to 896 scans using SEDFIT (version 15.01b) (38,39). SEDFIT resulted in size-distribution analyses c(s), for which the algorithm assumes that all species have the same frictional ratio f/f₀. The final SEDFIT analyses (Table 1) used a fixed resolution of 200 and optimized the c(s) fits by floating f/f₀ and the baseline until the overall root mean-square deviations and visual appearance of the fits were satisfactory. The percentage of oligomers in the total loading concentration was derived using the c(s) integration function. The observed s-values were corrected to s_20,w by:

$$s_{20,w}=s_{T,B}\left(\frac{{\eta }_{T,B}}{{\eta }_{20,w}}\right)\frac{{\left(1-\overline{v}\rho \right)}_{20,w}}{{\left(1-\overline{v}\rho \right)}_{T,B}},$$

where s is the sedimentation coefficient and the subscript _T,B refers to the temperature of the buffer. _20,w refers to water at 20°C. ρ is the solvent density, η is the solvent viscosity. $\overline{v}$ is the protein partial specific volume.

Table 1.

Experimental Data by X-Ray and Neutron Scattering and AUC for Glycosylated and Deglycosylated IgG1

	Concentration (mg/mL)	RG (nm)	RXS-1 (nm)	RXS-2 (nm)	L (nm)
X-ray data
IgG1 glycosylated	3.60	5.13 ± 0.27	2.50 ± 0.17	1.42 ± 0.12	17
	3.19	5.10 ± 0.28	2.50 ± 0.18	1.41 ± 0.14	17
	2.05	5.07 ± 0.34	2.49 ± 0.20	1.39 ± 0.14	17
	1.36	5.02 ± 0.37	2.46 ± 0.22	1.35 ± 0.16	17
IgG1 TP1	4.89	5.16 ± 0.30	2.50 ± 0.18	1.43 ± 0.13	17
	3.25	5.13 ± 0.28	2.50 ± 0.19	1.42 ± 0.13	17
	2.01	5.08 ± 0.31	2.50 ± 0.19	1.41 ± 0.14	17
	1.23	5.06 ± 0.41	2.49 ± 0.20	1.41 ± 0.16	17
IgG1 TP6	3.85	5.15 ± 0.29	2.51 ± 0.19	1.43 ± 0.13	17
	2.94	5.12 ± 0.32	2.51 ± 0.18	1.41 ± 0.13	17
	1.61	5.08 ± 0.35	2.51 ± 0.20	1.39 ± 0.14	17
	0.98	5.03 ± 0.41	2.49 ± 0.25	1.38 ± 0.17	17
IgG1 TP10	4.29	5.19 ± 0.28	2.52 ± 0.18	1.43 ± 0.13	17
	3.19	5.15 ± 0.31	2.51 ± 0.18	1.43 ± 0.13	17
	1.34	5.08 ± 0.34	2.50 ± 0.21	1.42 ± 0.16	17
	1.02	5.06 ± 0.41	2.50 ± 0.23	1.39 ± 0.18	17
Neutron data
IgG1 glycosylated	2.60	5.27 ± 0.28	2.35 ± 0.18	1.18 ± 0.14	16
	1.38	5.26 ± 0.34	2.36 ± 0.20	1.10 ± 0.14	16
IgG1 TP1	4.78	5.34 ± 0.25	2.44 ± 0.20	1.19 ± 0.14	16
	2.32	5.22 ± 0.63	2.42 ± 0.35	1.18 ± 0.25	16
IgG1 TP6	3.71	5.28 ± 0.74	2.42 ± 0.34	1.17 ± 0.19	16
	1.78	5.27 ± 0.74	2.43 ± 0.36	1.16 ± 0.26	16
IgG1 TP10	2.73	5.31 ± 0.77	2.39 ± 0.32	1.12 ± 0.21	16
	0.90	5.19 ± 0.89	2.38 ± 0.45	1.17 ± 0.35	16
AUC data		s20,w (S)	–	–	–
IgG1 glycosylated	3.00	6.16	–	–	–
	2.50	6.20	–	–	–
	2.00	6.32	–	–	–
	1.00	6.30	–	–	–
	0.50	6.43	–	–	–
IgG1 TP1	2.32	6.09	–	–	–
	1.78	6.13	–	–	–
	0.59	6.15	–	–	–
IgG1 TP6	6.01	6.04	–	–	–
	3.07	6.18	–	–	–
	1.54	6.16	–	–	–
	0.96	6.20	–	–	–
IgG1 TP10	2.51	6.09	–	–	–
	1.92	6.11	–	–	–
	1.09	6.14	–	–	–
	0.64	6.15	–	–	–

X-ray- and neutron-scattering data and analyses for IgG1

X-ray-scattering data were obtained during one beam session (October 2017) on Instrument B21 at the Diamond Light Source at the Rutherford Appleton Laboratory (Didcot, UK), operating with a ring energy of 3 GeV, and an operational energy of 12.4 keV. A PILATUS 2M detector (Dectris, Baden, Sweden) with a resolution of 1475 × 1679 pixels (pixel size of 172 × 172 μm) was used with a sample-to-detector distance of 4.01 m, giving a Q range from 0.04 to 4 nm⁻¹ (where Q = 4πsinθ/λ; 2θ = scattering angle; λ = wavelength). The glycosylated IgG1 (1.4–5.4 mg/mL) and the TP1 (0.7–4.9 mg/mL), TP6 (1.0–3.9 mg/mL), and TP10 (1.0–4.3 mg/mL) samples in light water were loaded onto a 96-well plate and placed into an EMBL Arinax sample holder (40,41). This condition showed the antibody molecule as a hydrated structure in a high-positive solute-solvent contrast (32). An automatic sampler injected 30 μL of sample from the well plate into a temperature-controlled quartz cell capillary with a diameter of 1.5 mm. Data sets of 30 frames with a frame exposure time of 1 s each were acquired in duplicate as a control of reproducibility. Checks during data acquisition confirmed the absence of radiation damage. ScÅtter (version 3.0) was used for buffer subtraction and data reduction, in which the 30 frames were averaged (42).

Neutron-scattering data on glycosylated IgG1 (2.60–1.38 mg/mL) and the TP1 (4.78–2.32 mg/mL), TP6 (3.71–1.78 mg/mL), and TP10 (2.73–0.90 mg/mL) samples in heavy water were obtained in two sessions (March 2017 and October 2017) on instrument SANS2D at the ISIS-pulsed neutron source at the Rutherford Appleton Laboratory (43). This condition showed the antibody structure in a high-negative solute-solvent contrast (32). No conformational differences in the antibody between light and heavy water were detected in this study or previously (36). A pulsed neutron beam was derived from proton beam currents of ∼40 μA. SANS2D data were recorded with 4 m of collimation, a 4-m sample-to-detector distance, a 12-mm sample aperture, and a wavelength range of 0.175–1.65 nm made available by time of flight. This gave a Q range from 0.05 to 4 nm⁻¹. The data were acquired using a two-dimensional ³He detector with 512 × 512 pixels of 7.5 × 7.5 mm² in size. Samples of volume 1 mL were measured in 2-mm path length circular banjo cells for 1–7 h in a thermostated sample rack at 20°C. Data were reduced using MANTID software (44). The MANTID data reduction steps include corrections for the Q resolution, i.e., beam divergence effects and smearing from the shape and size of the slits, as well as the wavelength overlap in each pulse (44). Using SASview software, the Guinier analyses (below) were found to be almost unaffected if the smearing was turned on or off.

Guinier analyses of the scattering data gave information of the R_G, cross-sectional radius (R_XS), and molecular mass. The scattering curve I(Q) intensities at low Q are defined by the R_G-value, which is the averaged distance of each scattering point from the center of scattering. In a given solute-solvent contrast, the R_G is a measure of structural elongation if the internal inhomogeneity of scattering densities within the protein has no effect. Guinier analyses at low Q gave the R_G-value and the forward scattering at zero-angle I(0) (45):

$$\ln I\left(Q\right)=\ln I\left(0\right)-\frac{R_G^2{Q}^2}{3}.$$

For antibodies, this expression is valid in a Q.R_G range up to 1.5, and was used in our previous studies (35,36), although the usual upper range reported in the literature is 1.0–1.3. If the structure is elongated, the mean R_G of the cross-sectional structure R_XS and the mean cross-sectional intensity at zero-angle [I (Q)Q]_Q→0 is obtained from (46):

$$\ln \left[I\left(Q\right)Q\right]={\left[I\left(Q\right)Q\right]}_{Q\to 0}-\frac{R_{XS}^2{Q}^2}{2}.$$

For immunoglobulins, it has been long recognized that the cross-sectional plot exhibits two regions, a steeper innermost one and a flatter outermost one (46), and the two analyses are denoted by R_XS-1 and R_XS-2, respectively. The R_XS-1 parameter represents the averaged overall spatial separation of the Fab and Fc regions, whereas the R_XS-2 parameter represents the averaged spatial cross section of the two Fab and one Fc region. The R_G and R_XS analyses were performed using SCT (Table 1) (47). The Q ranges for the R_G-, R_XS-1-, and R_XS-2-values were 0.10–0.22, 0.29–0.52, and 0.66–1.05 nm⁻¹, respectively, as previously described (35,36). Indirect transformation of the scattering data I(Q) in reciprocal space into real space to give the distance distribution function P(r) was carried out using GNOM (version 4.6) (48,49),

$$P\left(r\right)=\frac{1}{2{\pi }^2}\underset{0}{\overset{\infty }{\int }}I\left(Q\right)Qr\sin \left(Qr\right)dQ.$$

P(r) corresponds to the distribution of distances r between the volume elements in the macromolecule. This yields the maximal dimension of the macromolecule L and its most commonly occurring distance vector, M, in real space. For this P(r) analysis, the x-ray I(Q) curve utilized up to 755 data points in the Q range between 0.032 and 1.70 nm⁻¹ for both glycosylated and deglycosylated IgG1. The neutron P(r) curve utilized up to 155 I(Q) data points in the Q range between 0.055 and 1.60 nm⁻¹ for both glycosylated and deglycosylated IgG1.

Atomistic modeling of IgG1

Starting structures were created for each of glycosylated and deglycosylated IgG1 A33 based on the A33 sequence provided by UCB Pharma. The latter was aligned with the sequences of IgG1 6a, IgG1 19a, and IgG1 b12 (19,35; (Fig. 2)). This multiple sequence alignment was generated using Clustal Omega software (European Molecular Biology Laboratory-European Bioinformatics Institute) (50). The Fab structure (Fig. 1) was based on that found in the human IgG1 b12 crystal structure (PDB: 1HZH) (19) and the Fc structure was based on that for the rituximab IgG1 antibody Fc crystal structure (PDB: 4W4N) (22), which is unchanged from that of human IgG1 but is structurally complete compared with the IgG1 b12 structure where its Fc region showed gaps. Modeler (Version 9.19) (51) was used to generate the human IgG1 structure. In this, the IgG1 hinge was built using a PyMOL script build_seq (PyMOL Script Repository; Queen’s University, Ontario, Canada), based on the sequence of ²¹⁶EPKSCDKTHTCPPCPAPELLGGP²³⁸. The two N-linked oligosaccharides at Asn²⁹⁷ on the C_H2 domains were approximated as complex-type biantennary oligosaccharide structures with an Man₃GlcNAc₂ core and two NeuNAc.Gal.GlcNAc antennae (5). The glycan template was taken from the GitHub repository (https://github.com/dww100), which was energy minimized using NAMD (52) for 1 ns to achieve a relaxed structure. This glycan was added to the Fc region by bringing the C1 atom in the first GlcNAc residue to within 0.14 nm to the N sidechain atom of Asn²⁹⁷ in the C_H2 domain of IgG1 while ensuring no steric clashes with the Fc residues and the glycan chain. The PDB file was then opened on Discovery Studio (Dassault Systèmes BIOVIA, San Diego) where “CONECT” records were created for these glycosidic bonds. The CHARMM force field parameters and protein structure file, including those for the disulphide bridges and glycans, were generated using the CHARMM-Gui GlycanReader tool (53, 54, 55, 56, 57) to be compatible with the CHARMM36 force field (54, 55, 56, 57, 58, 59, 60). To relax this structure, the full IgG1 structure with and without glycans were then energy minimized for 10,000 steps using the simulation engine NAMD version 2.9 with the CHARMM36 force field.

For the MC simulations to generate trial structures, the starting IgG1 structure was renumbered and its naming nomenclature was adjusted to match the required format for the Torsion Angle MC (TAMC) module on SASSIE-web (61). For TAMC to work, the IgG1 residue numbering was changed to be continuous for two segments, one segment corresponding to the first Fab region, its hinge and the Fc region, and the other segment to only the second Fab region and the hinge connected to this. A library of physically realistic glycosylated and deglycosylated structural conformations was generated by subjecting the starting structures to the TAMC module in SASSIE-web (61). The flexible regions were assigned within the hinge, namely ²¹⁶EPK²¹⁸ and ²³¹APE²³³ on one side of IgG1 and just ²¹⁶EPK²¹⁸ on the other side (Figure 1, Figure 2 E). These tripeptides corresponded to surface-accessible structures outside the structurally-defined Fab and Fc regions and the disulphide-linked hinge core. These tripeptides could be structurally varied to create the required IgG1 conformers for testing against the scattering curve. The rest of the IgG1 structure was held rigid. Making ²¹⁶EPK²¹⁸ flexible on both sides of IgG1 rendered both Fab regions to be conformationally mobile, and making ²³¹APE²³³ flexible made the Fc region mobile. For each of these nine linker residues, the backbone φ and ψ torsion angles were varied in steps of up to either 30 or 180°. In the MC simulation, many attempted moves will be physically unrealistic and were therefore discarded. For the glycosylated IgG1 simulations, 2,500,000 moves were attempted, of which 123,284 were accepted. For the deglycosylated simulations, in which the glycan chains were omitted, 2,600,000 moves were attempted, of which 119,191 models were accepted.

For each of the 123,284 and 119,191 models, a scattering curve was generated using the SasCalc module in SASSIE-web. SasCalc calculates the scattering curve I(Q) using an all-atom expression for the scattering intensity in which the orientations of the Q vectors are taken from a quasi-uniform spherical grid generated by the golden ratio (62). For x-ray modeling, consideration of the hydration shell would require the explicit addition of a monolayer of water molecules to the protein surface before calculating I(Q), and would require much computational effort (62). Thus, the hydration shell was not considered here for x-rays, and was not required for neutrons. These scattering curves were compared with the x-ray- and neutron-experimental-scattering curves, using the R-factor function in SASSIE-web. This function calculates the difference between the modeled curve I_Model(Q_i) and the experimental curves I_Expt(Q_i), this function being analogous to that used in protein crystallography:

$$R=\frac{\sum \Vert \Vert I_{Expt}\left(Qi\right)\Vert -\eta I_{Model}\left(Qi\right)\Vert }{\sum \Vert I_{Expt}\left(Qi\right)\Vert }\times 100,$$

where Q_i is the Q-value of the i^th data point, I_Expt(Q_i) is the experimental scattering intensity, I_Model (Q_i) is the theoretical modeled scattering intensity, and $\eta$ is a scaling factor used to match the theoretical curve to the experimental I(0) (47). Lower R-factor values represent better fits. An iterative search to minimize the R-factor was used to determine $\eta$ (47). In the extrapolated experimental scattering curves, the lowest Q-values in the range before the fitted Guinier R_G region were interpolated to zero Q using MATLAB (The MathWorks, Natick, MA) to satisfy the input requirement for the SasCalc module in SASSIE-web. Interpolation makes the Q spacing uniform between the data points, and extrapolation extends the full I(Q) curve to zero Q. The resulting 680 and 72 I(Q)-values in the Q range of 0.0–1.5 nm⁻¹ were utilized for the x-ray and neutron curve fits respectively, and defined the Q spacing for SasCalc and the R-factor values. The use of χ² analyses to evaluate the fits was not possible because this requires the experimental data points to have errors associated with them, which were not available when interpolating the curve. For the neutron curve fits, no correction was required for a flat incoherent background because the IgG1 concentrations were relatively low and dialyzes had sufficiently reduced the proton content in the buffers. The 123,284 glycosylated and 119,191 deglycosylated models gave an R-factor versus R_G distribution that encompassed the experimental extrapolated R_G-value. This R-factor analysis was repeated for four experimental x-ray-scattering curves at different concentrations for each of glycosylated and deglycosylated IgG1 (Table 2). The same analysis was repeated for two neutron-scattering curves at different concentrations for each glycosylated and deglycosylated IgG1 (Table 3). For each concentration, the 100 best-fit models with the smallest R-factors were accepted.

Table 2.

Modeling Fits for the X-Ray Scattering and AUC Data in Light Water

Filter	Model	RG before minimization (nm)	RG after minimization (nm)	RXS-1 (nm)	RXS-2 (nm)	L (nm)	R-factor before minimization (%)	R-factor after minimization (%)	s20,w (S)
Library of glycosylated models	123284	4.55–5.59	NA	NA	NA	NA	0.70–8.43	NA	NA
Top 100 at 3.60 mg/mL	100	5.09–5.21	5.09–5.21	2.48–2.63	1.31–1.61	NA	0.70–1.13	0.71–1.21	6.57–6.73
Best fit at 3.60 mg/mL	1	5.14	5.14	2.53	1.50	17	0.70	0.71	6.64
Top 100 at 3.19 mg/mL	100	5.08–5.19	5.07–5.19	2.48–2.63	1.24–1.55	NA	0.75–1.18	0.75–1.23	6.59–6.73
Best fit at 3.19 mg/mL	1	5.14	5.14	2.53	1.49	17	0.75	0.75	6.70
Top 100 at 2.05 mg/mL	100	5.02–5.18	5.02–5.17	2.44–2.63	1.20–1.54	NA	1.05–1.54	1.04–1.55	6.59–6.73
Best fit at 2.05 mg/mL	1	5.15	5.15	2.48	1.37	17	1.05	1.04	6.71
Top 100 at 1.36 mg/mL	100	5.00–5.18	5.00–5.18	2.32–2.67	1.20–1.54	NA	0.94–2.18	0.96–2.29	6.62–6.77
Best fit at 1.36 mg/mL	1	5.08	5.08	2.48	1.32	17	0.94	0.96	6.68
Library of deglycosylated models	119191	4.63–5.67	NA	NA	NA	NA	0.87–8.60	NA	NA
Top 100 at 4.29 mg/mL	100	5.10–5.28	5.10–5.28	2.49–2.68	1.30–1.62	NA	0.80–1.67	0.81–1.67	6.24–6.45
Best fit at 4.29 mg/mL	1	5.21	5.21	2.54	1.52	17	0.80	0.81	6.27
Top 100 at 3.19 mg/mL	100	5.05–5.27	5.05–5.27	2.49–2.69	1.26–1.62	NA	0.87–1.65	0.85–1.65	6.28–6.45
Best fit at 3.19 mg/mL	1	5.17	5.17	2.59	1.38	17	0.87	0.85	6.36
Top 100 at 1.34 mg/mL	100	4.98–5.25	4.98–5.25	2.42–2.71	1.25–1.56	NA	0.93–1.77	0.94–1.79	6.28–6.50
Best fit at 1.34 mg/mL	1	5.11	5.10	2.57	1.31	17	0.93	0.94	6.45
Top 100 at 1.02 mg/mL	100	4.94–5.19	4.93–5.19	2.42–2.73	1.25–1.56	NA	1.06–1.93	1.03–1.95	6.28–6.50
Best fit at 1.02 mg/mL	1	5.11	5.11	2.57	1.31	17	1.06	1.03	6.45
PCA Group 1	217	5.00–5.28	5.00–5.28	2.32–2.63	1.25–1.61	NA	0.87–2.17	0.85–2.29	6.24–6.75
Glycosylated	175	5.00–5.18	5.00–5.18	2.32–2.63	1.25–1.61	NA	0.91–2.17	0.89–2.29	6.61–6.75
Deglycosylated	42	5.08–5.28	5.08–5.28	2.50–2.63	1.30–1.61	NA	0.87–1.84	0.85–1.82	6.24–6.33
Centroid	1	5.07	5.06	2.32	1.43	17	2.15	2.29	6.74
PCA Group 2	50	5.06–5.28	5.06–5.28	2.42–2.58	1.50–1.62	NA	0.94–2.18	0.96–2.21	6.25–6.72
Glycosylated	46	5.06–5.18	5.06–5.18	2.42–2.58	1.20–1.54	NA	0.94–2.18	0.96–2.21	6.62–6.72
Deglycosylated	4	5.12–5.28	5.12–5.28	2.50–2.56	1.34–1.62	NA	1.27–1.62	1.24–1.67	6.25–6.38
Centroid	1	5.10	5.10	2.51	1.20	17	1.48	1.48	6.64
PCA Group 3	234	5.02–5.28	5.02–5.28	2.46–2.61	1.20–1.59	NA	0.70–2.18	0.71–2.12	6.24–6.77
Glycosylated	176	5.02–5.21	5.02–5.21	2.46–2.59	1.20–1.59	NA	0.70–2.18	0.71–2.12	6.57–6.77
Deglycosylated	58	5.09–5.28	5.09–5.28	2.52–2.61	1.30–1.58	NA	0.87–1.92	0.87–1.91	6.24–6.38
Centroid	1	5.19	5.19	2.50	1.53	17	1.10	1.09	6.65
PCA Group 4	153	5.11–5.26	5.10–5.26	2.42–2.61	1.33–1.56	NA	0.80–2.11	0.81–2.10	6.29–6.71
Glycosylated	2	5.15	5.15	2.53	1.38	NA	2.11	2.10	6.71
Deglycosylated	151	5.11–5.26	5.10–5.26	2.42–2.61	1.33–1.56	NA	0.80–1.93	0.81–1.93	6.29–6.50
Centroid	1	5.21	5.21	2.55	1.52	17	1.67	1.67	6.40
PCA Group 5	146	4.94–5.22	4.93–5.22	2.58–2.73	1.25–1.61	NA	0.92–1.93	0.93–2.18	6.28–6.67
Glycosylated	1	5.02	5.02	2.67	1.31	NA	2.17	2.18	6.67
Deglycosylated	145	4.94–5.22	4.93–5.22	2.58–2.73	1.25–1.61	NA	0.92–1.93	0.93–1.95	6.28–6.46
Centroid	1	5.16	5.16	2.61	1.59	17	1.28	1.29	6.45

NA, not available.

Table 3.

Modeling Fits for the Neutron Scattering and AUC Data in Heavy Water

Filter	Model	RG before minimization (nm)	RG after minimization (nm)	Rxs-1 (nm)	Rxs-2 (nm)	L (nm)	R-factor before minimization (%)	R-factor after minimization (%)	s20,w (S)
Library of glycosylated models	123284	4.51–5.55	NA	NA	NA	NA	1.68–11.46	NA	NA
Top 100 at 2.60 mg/mL	100	5.16–5.33	5.15–5.32	2.32–2.53	1.03–1.32	NA	1.68–2.19	1.74–2.32	6.53–6.71
Best fit at 2.60 mg/mL	1	5.24	5.23	2.47	1.19	16	1.68	1.74	6.64
Top 100 at 1.38 mg/mL	100	5.19–5.35	5.17–5.34	2.32–2.49	1.10–1.25	NA	1.98–2.38	1.99–2.43	6.57–6.71
Best fit at 1.38 mg/mL	1	5.24	5.23	2.36	1.22	16	1.98	2.04	6.70
Library of deglycosylated models	119191	4.57–5.61	NA	NA	NA	NA	2.22–12.04	NA	NA
Top 100 at 2.73 mg/mL	100	5.20–5.34	5.20–5.34	2.46–2.57	1.11–1.27	NA	2.28–2.54	2.24–2.50	6.26–6.35
Best fit at 2.73 mg/mL	1	5.31	5.31	2.47	1.19	16	2.28	2.24	6.28
Top 100 at 0.90 mg/mL	100	5.08–5.22	5.07–5.21	2.42–2.57	1.15–1.39	NA	2.23–2.66	2.22–2.62	6.29–6.44
Best fit at 0.90 mg/mL	1	5.15	5.14	2.50	1.26	16	2.22	2.22	6.32
PCA Group 1	185	5.14–5.33	5.13–5.32	2.32–2.54	1.03–1.35	NA	1.87–2.63	1.95–2.60	6.28–6.71
Glycosylated	167	5.16–5.33	5.15–5.32	2.32–2.53	1.03–1.27	NA	1.87–2.19	1.95–2.32	6.60–6.71
Deglycosylated	18	5.14–5.29	5.13–5.29	2.48–2.54	1.16–1.35	NA	2.23–2.63	2.22–2.60	6.28–6.34
Centroid	1	5.14	5.13	2.51	1.34	16	2.52	2.51	6.32
PCA Group 2	18	5.09–5.28	5.08–5.27	3.37–2.57	1.16–1.39	NA	1.84–2.65	1.87–2.62	6.24–6.66
Glycosylated	9	5.22–5.28	5.21–5.27	2.37–2.50	1.18–1.29	NA	1.84–2.09	1.87–2.13	6.53–6.66
Deglycosylated	9	5.09–5.21	5.08–5.20	2.42–2.57	1.16–1.39	NA	2.26–2.65	2.25–2.62	6.32–6.44
Centroid	1	5.15	5.14	2.5	1.35	16	2.38	2.38	6.36
PCA Group 3	16	5.08–5.25	5.07–5.24	2.45–2.50	1.20–1.26	NA	1.86–2.48	2.00–2.56	6.25–6.62
Glycosylated	15	5.18–5.25	5.17–5.24	2.45–2.50	1.20–1.26	NA	1.86–2.14	2.00–2.28	6.54–6.62
Deglycosylated	1	5.08	5.07	2.46	1.26	NA	2.48	2.56	6.38
Centroid	1	5.25	5.24	2.49	1.21	16	2.13	2.16	6.32
PCA Group 4	143	5.12–5.35	5.11–5.34	2.41–2.53	1.11–1.39	NA	1.68–2.66	1.74–2.61	6.26–6.67
Glycosylated	9	5.19–5.35	5.18–5.30	2.41–2.50	1.19–1.32	NA	1.68–2.14	1.74–2.24	6.63–6.67
Deglycosylated	134	5.12–5.34	5.11–5.34	2.42–2.53	1.11–1.39	NA	2.28–2.66	2.23–2.61	6.26–6.37
Centroid	1	5.22	5.22	2.47	1.23	16	2.43	2.4	6.54
PCA Group 5	1	5.1	5.09	2.54	1.24	NA	2.61	2.61	6.29
Glycosylated	0	NA	NA	NA	NA	NA	NA	NA	NA
Deglycosylated	1	5.1	5.09	2.54	1.24	NA	2.61	2.61	6.29
Centroid	NA	NA	NA	NA	NA	NA	NA	NA	NA

NA, not available.

Principal component analysis (PCA) provided by the Bio3d package in R (63) was used to identify the main classes of best-fit IgG1 conformations found in the 800 best-fit glycosylated and deglycosylated models from x-ray scattering (Table 2). A separate analysis of the 400 best-fit models from neutron scattering was performed. To remove any bias in the PCA clustering of coordinate sets caused by the presence or absence of the glycans, the glycan coordinates were removed from the best-fit glycosylated models before generating the PCA. The x-ray and neutron models were assessed through two separate PCA. The average structure for each PCA group was identified using a centroid model computed using R. The 100 best-fit glycosylated and deglycosylated IgG1 structures at 3.60 and 4.29 mg/mL, respectively, are available to download in Data S1. The two single best-fit glycosylated and deglycosylated IgG1 structures were deposited in the SASBDB database (https://www.sasbdb.org/) with reference codes SAS2937 and SAS2938.

To model AUC parameters, the theoretical s_20,w-values were generated for the 800 glycosylated and 400 deglycosylated best-fit IgG1 models using HullRad Version 7 (64). Hullrad includes glycan residues for glycosylation; however, there are inconsistencies in the Protein Database and CHARMM-GUI nomenclature for glycans. The nomenclature in the Hullrad script was thus modified to ensure that the IgG1 glycosylation was correctly incorporated in the s_20,w calculation.

Results

Purification and characterization of glycosylated and deglycosylated IgG1

A protocol for the deglycosylation of the monoclonal human IgG1 A33 antibody was set up using PNGase F digests according to the manufacturer’s protocol (Materials and methods). The completeness of deglycosylation was verified by a combination of routine gel filtration, SDS-PAGE, mass spectrometry, and AUC (see below):

• 1)At the time points of TP1, TP6, and TP10 after the start of the digests, the elution of the IgG1-digested products from a gel filtration column preceded that of native glycosylated IgG1 (Fig. 3 A). Both glycosylated and deglycosylated IgG1 eluted as a main symmetrical peak at 17.88, 17.84, 17.82, and 17.86 mL for glycosylated IgG1, and deglycosylated IgG1 at time points of TP1, TP6, and TP10, respectively (Fig. 3 A). This process ensured that the IgG1 sample was monodisperse with no aggregates present immediately before AUC or scattering experiments.
• 2)When the IgG1 samples were submitted to nonreducing and reducing SDS-PAGE analyses at equimolar concentrations, purified glycosylated and deglycosylated IgG1 showed a single band between 200 and 116 kDa on 4–12% Bis Tris NuPage gel under nonreducing conditions, which is consistent with the expected masses of ∼147 kDa for IgG1 (Fig. 3 B). Under reducing conditions, two bands were present corresponding to the heavy chain (with an apparent mass of ∼50 kDa) and the light chain (with an apparent molecular mass of ∼25 kDa) (Fig. 3 B). These apparent molecular masses were, as expected, from the known sequence.
• 3)Liquid chromatography mass spectrometry measurements showed multiple peaks for glycosylated IgG1 (G) that were assigned to the presence of at least four glycoforms, separated by masses of 160–230 Da that corresponded to single sugar residues (Fig. 3 C; (65)). The most intense IgG1 glycosylated population had an observed deconvoluted mass of 147,010 Da. After an hour of digest, a partially deglycosylated IgG1 (P) was observed in which the number of glycoforms was diminished, and additional peaks were observed at ∼145.4 and 143,958 Da. After 6 or 10 h, only the single dominant deglycosylated peak (D) was seen at 143,958 Da. The peak at 145.4 kDa was attributed to IgG1 in which one of the two glycans at Asn297 was not present. The mass of each glycan chain was calculated by subtracting the glycosylated and deglycosylated masses and halving the outcome to give 1526 Da. This glycan mass was also calculated by subtracting the glycosylated and half-deglycosylated masses to give 1613 Da. These values agree well with an assumed glycan composition of Gal₂Man₃GlcNAc₄ that gives a mass of 1622 Da.

Figure 3.

Purification, SDS-PAGE, and mass spectrometry of human glycosylated and deglycosylated IgG1. (A) Elution peaks from a Superose 6 Increase 10/300 gel filtration column for four IgG1 samples, these being glycosylated (black) and from the TP1 (blue), TP6 (red), and TP10 (magenta) time points. The dashed vertical lines indicate the peak positions. (B) Lane 1 and 6, molecular mass markers are denoted in kDa. Lanes 2–5, nonreduced SDS-PAGE of glycosylated IgG1, TP1, TP6, and TP10 after gel filtration. Lanes 7–10 show reduced SDS-PAGE of glycosylated IgG1, TP1, TP6, and TP10 after gel filtration. (C) Mass spectra of glycosylated and deglycosylated IgG1, using the same color scheme as in (A). Peaks labeled G represents glycosylated species, P represents partially glycosylated species, and D represents fully deglycosylated species. To see this figure in color, go online.

AUC of glycosylated and deglycosylated IgG1

Sedimentation velocity experiments investigated the masses and solution structures of glycosylated and deglycosylated IgG1 at the TP1, TP6, and TP10 time points. The SEDFIT analyses of the boundaries involved fits of as many as 896 scans, and the good agreement between the experimental boundary scans and fitted lines is clear (left, Fig. 4 A). In the resulting size-distribution analyses c(s), a monomer peak that monitored the overall IgG1 solution structure was observed at average s_20,w-values of 6.25 S for glycosylated IgG1, 6.12 S for IgG1 TP1, 6.15 S for IgG1 TP6, and 6.12 S for IgG1 TP10 (right, Fig. 4 A). For the glycosylated forms, these values agreed well with those of 6.42 and 6.34 S for IgG1 6a and 19a, respectively (35), and with earlier studies (66, 67, 68). From the c(s) analyses, the molecular masses of the IgG1 monomer peak were 151 kDa (glycosylated), 147 kDa (TP1), 156 kDa (TP6), and 148 kDa (TP10). These values agree well with the composition-calculated masses of 148.4 and 144.4 kDa for the glycosylated and deglycosylated IgG1 monomers, respectively. These also agree well with the values from mass spectrometry of 147,010 and 143,958 Da for glycosylated and deglycosylated IgG1 (Fig. 3 C). The 0.2-S reduction (3%) in the s_20,w-values on deglycosylation is attributable to the 4-kDa reduction (3%) in the IgG1 mass, according to the Svedberg equation in which s_20,w is proportional to the mass divided by the frictional coefficient. This calculation assumes that the IgG1 conformation (i.e., the frictional coefficient) is unchanged after deglycosylation.

Figure 4.

Sedimentation velocity analyses of glycosylated and deglycosylated IgG1. (A) The experimentally observed sedimentation boundaries for a concentration series of glycosylated IgG1 and likewise for deglycosylated IgG1 at the TP1 (blue), TP6 (red), and TP10 (magenta) time points in histidine buffer. Scans were recorded at 30,000 rpm and 20°C, from which 34–46 boundaries (black outlines) are shown from totals of up to 896 scans. The SEDFIT fits are shown as white lines. The peaks in the corresponding size-distribution analyses c(s) revealed a monomer peak (M) at s⁰_20,w-values of 6.18–6.45 S for glycosylated IgG1 and the three deglycosylation time points TP1, TP6, and TP10. (B) The s⁰_20,w-values for the monomer peaks are shown as a function of concentration for glycosylated IgG1 (black circles), and the TP1 (blue circles), TP6 (red circles), and TP10 (magenta circles) time points. To see this figure in color, go online.

A slight concentration dependence was observed for the monomer s_20,w-values for glycosylated and deglycosylated IgG1 (Fig. 4 B), which increased with decreased concentration. In the 2015 study, minor peaks for IgG1 dimers were visible at ∼9 S for IgG1 6a and 19a (35). In this work, no dimer peaks were visible for IgG1 A33 (right panel, Fig. 4 A). This difference may result from the use of histidine buffer in this current study, in distinction to the phosphate buffer saline used before. If IgG1 A33 forms dimers, the histidine buffer may have increased the exchange rate between monomer and dimer such that separate monomer and dimer peaks were no longer seen. Interestingly, the peak width for glycosylated IgG1 is greater than that of deglycosylated IgG1 (Fig. 4 A). The increased width may indicate a mix of monomer and dimer in fast exchange in glycosylated IgG1, which is reduced to monomer-only upon deglycosylation.

X-ray and neutron scattering of glycosylated and deglycosylated IgG1

The overall solution structures of glycosylated and the three deglycosylated IgG1 samples at the TP1, TP6, and TP10 time points were analyzed by x-ray and neutron scattering. The two methods provided slightly different perspectives of the same solution structure. X-rays in light water buffers detect the hydration shell surrounding the protein structure, whereas the effect of the hydration shell is reduced by neutrons in heavy water buffers for reason of the different solute-solvent contrast in use (4,32,33). The IgG1 x-ray data collection at concentrations between 0.5 and 5.5 mg/mL used time frame analyses to ensure the absence of radiation damage effects. The resulting R_G and R_XS-1/R_XS-2-values monitor the elongation of the overall IgG1 structure and its approximate cross-sectional structures, respectively.

Guinier analyses resulted in high-quality linear plots for all four samples and revealed three distinctive regions of the I(Q) curves, as expected for antibodies (35,69,70). From these, the R_G-, R_XS-1-, and R_XS-2-values from the individual scattering curves were obtained within satisfactory Q.R_G and Q.R_XS limits of 0.5–1.4, 0.7–1.3, and 0.9–1.5, respectively (Fig. 5 A). A slight concentration dependence was observed in the x-ray I(0)/c-values that suggested a small amount of oligomer formation in the concentration series (Fig. 6 A; Table 1). This agreed with the AUC data (Fig. 4 B). After extrapolation to zero concentration, the x-ray R_G-values that monitor the overall structure for glycosylated IgG1, and the deglycosylated TP1, TP6, and TP10 IgG1 samples were almost unchanged at 5.10 ± 0.13, 5.10 ± 0.20, 5.11 ± 0.13, and 5.13 ± 0.13 nm, respectively. These x-ray R_G-values for glycosylated IgG1 A33 here agree well with previous R_G-values of 5.28–5.32 nm for two other human monoclonal IgG1 6a and 19a antibodies (35). The R_XS-1-values from the individual curves (Fig. 6 A) is an approximate monitor of the cross-sectional structure for glycosylated IgG1 and deglycosylated IgG1. These were extrapolated to zero concentration to show that these were slightly increased from 2.47 ± 0.01 to 2.51 ± 0.01 nm, respectively, showing some rearrangement between the Fab and Fc regions. The R_XS-2-values from the individual curves for glycosylated IgG1 and deglycosylated IgG1 at the TP1, TP6, and TP10 time points were extrapolated to zero concentration to show that these were almost unchanged at 1.40 ± 0.07, 1.41 ± 0.05, 1.41 ± 0.05, and 1.42 ± 0.04 nm. This showed that the individual Fab and Fc regions were unchanged in structure. In summary, the small increase of 0.04 nm in the extrapolated R_XS-1-values with increasing deglycosylation suggested that there were small increases of elongation in the cross-sectional IgG1 structure upon removal of the glycan chains.

Figure 5.

X-ray and neutron Guinier R_G and R_XS analyses for glycosylated and deglycosylated IgG1. (A) The SAXS curves for glycosylated and deglycosylated (TP1, TP6, and TP10) IgG1 at concentrations of 5.38–0.74 mg/mL. The filled circles between the arrows represent the Q.R_G and Q.R_XS fit ranges used to determine the R_G- and R_XS-values. The Q range used for the R_G-values was 0.01–0.027 nm⁻¹; those for the R_XS-1- and R_XS-2-values were 0.029–0.052 and 0.066–0.105 nm⁻¹, respectively. (B) The SANS curves for glycosylated and deglycosylated (TP1, TP6, and TP10) IgG1 at concentrations of 4.78–0.90 mg/mL. The Q range used for the R_G-values was 0.007–0.027 nm⁻¹ and those for the R_XS-1- and R_XS-2-values were 0.028–0.052 and 0.066–0.105 nm⁻¹, respectively.

Figure 6.

Concentration dependence of the SAXS and SANS Guinier analyses. The filled symbols show the values determined from the Guinier analyses and the open symbols in the R_G panels indicate those determined from the P(r) analyses. The colors denote the glycosylated IgG1 (black), TP1 (blue), TP6 (red), and TP10 (magenta) time points. (A) The SAXS R_G-, I(0)/c-, R_XS-1-, and R_XS-2-values for glycosylated (solid and open black circles) and deglycosylated TP1 (solid and open blue circles), TP6 (solid and open red circles), and TP10 (solid and open magenta circles). The solid lines correspond to linear regression fits of glycosylated IgG1, and the dashed lines correspond to the fits for deglycosylated IgG1. (B) The SANS R_G-, I(0)/c-, R_XS-1-, and R_XS-2-values for glycosylated and deglycosylated (TP1, TP6, and TP10) IgG1, each corresponding to a single measurement in histidine buffer in ²H₂O. The solid and dashed lines correspond to the mean values for glycosylated and deglycosylated IgG1. To see this figure in color, go online.

The corresponding neutron-scattering data sets for glycosylated and deglycosylated IgG1 (TP1, TP6, TP10) in 100% ²H₂O buffer were analyzed at concentrations of 0.71–2.73 mg/mL, this concentration range being similar to that used above for SAXS. Again, the Guinier analyses revealed high-quality linear fits for the R_G, R_XS-1, and R_XS-2 parameters (Fig. 5 B). A concentration dependence was not observed for IgG1, this being seen from the I(0)/c-values, which remained unchanged within error (Fig. 6 B). This difference between the neutron and x-ray data sets is attributable to the fewer data points obtained with neutrons, leading to reduced precision in the data sets. The mean neutron R_G-values for glycosylated IgG1 and deglycosylated IgG1 (TP1, TP6, and TP10) were unchanged at 5.27 ± 0.01, 5.28 ± 0.06, 5.28 ± 0.01, and 5.25 ± 0.06 nm, respectively (Fig. 6 B). The neutron R_G-values for glycosylated IgG1 A33 agree well with the R_G-values of 5.16–5.18 nm for human IgG1 6a and 19a (35). The mean neutron R_XS-1-values for glycosylated IgG1 and deglycosylated IgG1 (TP1, TP6, and TP10) were 2.35 ± 0.01, 2.43 ± 0.01, 2.42 ± 0.01, and 2.38 ± 0.01 nm, respectively, suggesting a small increase in the cross-sectional structure after deglycosylation. The mean neutron R_XS-2-values for glycosylated IgG1 and deglycosylated IgG1 (TP1, TP6, and TP10) were unchanged at 1.14 ± 0.04, 1.19 ± 0.01, 1.17 ± 0.01, and 1.15 ± 0.03 nm, respectively. The neutron values confirmed the x-ray analyses.

The distance distribution function P(r) is derived from Fourier transformation of the scattering curve I(Q) and provides structural information in real space on glycosylated and deglycosylated IgG1. The x-ray and neutron P(r) analyses gave R_G-values that were similar to those from the x-ray Guinier analyses, showing that the two analyses were self-consistent (open symbols, Fig. 6). The maximal lengths of glycosylated and deglycosylated IgG1 were determined from the value of r when the P(r) curve intersects zero on the r axis and was 17 nm for all four IgG1 samples. The neutron maximal lengths of glycosylated and deglycosylated IgG1, were 16 nm for all four samples. These were 1 nm smaller when compared with the x-ray value of 17 nm, this reduction in size being attributed to the reduced contribution of the hydration shell seen by neutron scattering (Fig. 7 B). These reductions in the neutron R_G- and L-values have been previously seen in our earlier joint SAXS and SANS studies of antibodies (35).

Figure 7.

SAXS and SANS distance distribution analyses P(r) for each of glycosylated and deglycosylated IgG1. The colors denote the glycosylated IgG1 (black), TP1 (blue), TP6 (red), and TP10 (magenta) time points. (A) The peak maxima at M1 and M2 and the maximal length L are indicated by arrows. The SAXS and SANS P(r) curves for glycosylated and deglycosylated (TP1, TP6, and TP10) IgG1 are shown at concentrations of 5.38–0.74 mg/mL. (B) The corresponding P(r) curves for the SANS curves for IgG1 4.78–0.90 mg/mL. (C and D) The concentration dependence of the peak maxima M1 and M2 for glycosylated and deglycosylated IgG1 are shown. The fitted lines are the mean values for glycosylated IgG1 (solid line), and for TP1, TP6, and TP10 averaged together (dashed lines). To see this figure in color, go online.

The maxima in the P(r) curves corresponded to the most frequently occurring distances between scattering elements within the structures, these being a monitor of the IgG1 structure. For the four IgG1 samples, two peaks, M1 and M2, were visible that are characteristic of antibody-shaped proteins. M1 corresponds primarily to the shorter distances within each Fab and Fc region, and is expected to be almost invariant for this reason. M2 corresponds primarily to the longer distances between pairs of Fab and Fc regions and monitors changes in the separation of the Fab and Fc regions (Figs. 1 and 7). No concentration dependence was observed in the positions of the M1 and M2 peaks, which were measured directly from their maximal values (Fig. 7, B and C). However, the x-ray M2 peak shifted significantly from a mean value of 7.44 ± 0.03 nm for glycosylated IgG1 to a mean value of 7.83 ± 0.02 nm for deglycosylated IgG1 (Fig. 7 C). The x-ray M1 peak shifted much less from 4.42 ± 0.02 nm for glycosylated IgG1 to 4.49 ± 0.01 nm for deglycosylated IgG1. The same change was seen in the neutron P(r) curves, when M2 increased from 7.27 ± 0.22 nm for glycosylated IgG1 to 7.77 ± 0.04 nm for deglycosylated IgG1 (Fig. 7 D). The neutron M1 peak was almost unchanged, with a shift from 4.36 ± 0.07 nm for glycosylated IgG1 to 4.21 ± 0.048 nm for deglycosylated IgG1. Both the x-ray and neutron analyses were consistent with each other.

Atomistic modeling of glycosylated and deglycosylated IgG1

To account for the changes seen in the R_XS-1- and M2-scattering parameters in IgG1 after deglycosylation, atomistic modeling simulations of the glycosylated and deglycosylated structures were performed, starting from two high-resolution crystal structures for the human Fab and Fc regions (Materials and methods). The sequence in the Fab structure was converted into that for IgG1 A33 using Modeler (Fig. 2, A–D). The Fab and Fc regions were joined by a peptide ²¹⁶EPKSCDKTHTCPPCPAPELLGGP²³⁸ that included the 15-residue hinge sequence (Fig. 2 E), built using PyMOL. The native glycosylated IgG1 models were created by adding complex-type biantennary glycans to the two Asn²⁹⁷ side chains in the Fc region (Fig. 1 B). This starting structure was subjected to energy minimization.

Physically realistic IgG1 models without steric overlaps or clashes were created for comparison with the experimental x-ray curves. By varying the torsion angles at three flexible regions at the start and end of the two IgG1 hinges (Materials and methods; (Fig. 1)), trial IgG1 structures were created that involved movements of the two Fab and one Fc regions relative to each other. For glycosylated IgG1, 2,500,000 models were generated in 16 MC simulations, from which 123,284 models were accepted because these showed no steric clashes between separate residues in the model. For deglycosylated IgG1, 2,600,000 models were generated in 20 MC simulations, of which 119,191 models were likewise acceptable. To ensure that no systematic trends were overlooked in the modeling outcome, four x-ray- and two neutron-scattering curves from up to four concentrations were fitted for each of the four samples in question (Fig. 8). For both glycosylated and deglycosylated IgG1, comparison of the four experimental x-ray-scattering curves at 1.36–3.60 mg/mL with the 123,284 and 119,191 theoretical curves gave a goodness-of-fit R-factor versus R_G distribution with clear minima in all eight cases (Fig. 8, A and B). The minima agreed with the experimental R_G-values (Fig. 6 A). The minima showed that enough trial x-ray models had been generated to result in good fits in each case. Filtering of the models to select these with the lowest R -factors gave the 100 best-fit models for each concentration (red, Fig. 8). The range of the 100 R-factors for each of the four concentrations was low at between 0.80 and 1.93% for the best-fit glycosylated models and 0.70–2.18% for the best-fit deglycosylated models (Table 2). This indicated good quality x-ray curve fits between the experimental and modeled curves.

Figure 8.

Atomistic modeling analyses for glycosylated and deglycosylated IgG1. The 123,284 models that were accepted for glycosylated IgG1 and the 119,191 models that were accepted for deglycosylated IgG1 are shown as circles. (A and B) Four experimental x-ray-scattering curves for four concentrations of glycosylated and deglycosylated IgG1 are shown. (C and D) Two experimental neutron-scattering curves for two concentrations of glycosylated and deglycosylated IgG1 are shown. These experimental curves were fitted to the 123,284 and 119,191 modeled curves. Those models with R_G-values closest to the experimental R_G-values showed the lowest goodness-of-fit R-factor, as expected. The top 100 best-fit models (red circles) showed the lowest goodness-of-fit R-factors. The experimental R_G is represented by a solid vertical line and the dashed vertical lines represent the ±2% upper and lower boundaries of these R_G-values. To see this figure in color, go online.

The eight sets of 100 best-fit models (Fig. 8) were examined to identify the resulting best-fit IgG1 conformations from the curve fits. For this, PCAs were performed (71). The PCA determines the correlated motions of protein residues as linearly uncorrelated variables, each being termed a principal component (71). These “essential motions” were extracted from a covariance matrix of the atomic coordinates of the frames in the selected structure set. The eigenvectors of this matrix each have an associated eigenvalue that characterizes the clustering of the models based on structural coordinates (or variance). To eliminate bias in the PCA, the glycan chains were removed from the glycosylated IgG1 models before comparison with the deglycosylated models. The PCA confirmed a clear difference between the glycosylated and deglycosylated x-ray IgG1 models (black and magenta, respectively, Fig. 9, A–D; Table 2). Thus, the distributions of the 400 best-fit glycosylated and 400 best-fit deglycosylated x-ray models were each clustered into five distinct groups with little overlap between glycosylated and deglycosylated groups. The glycosylated models mostly occurred in the PCA Groups 1, 2, and 3, and the deglycosylated models mostly occurred in the PCA Groups 4 and 5. This outcome verified the experimentally observed changes in the R_XS-1 and M2 parameters before and after deglycosylation (Figure 6, Figure 7 C). The visually excellent x-ray curve fits confirmed the validity of the modeling fits (Fig. 10, A and B). Of particular note was the agreement of the experimental and theoretical double peaks in the P(r) curves shown as insets.

Figure 9.

PCA of the best-fit glycosylated and deglycosylated IgG1 models. Glycosylated models are represented in black and deglycosylated models are represented in magenta. In this, Groups 1, 2, 3, 4, and 5 are represented by an open circle, a triangle, a plus sign, a multiplication sign, and a house, in that order, and the centroid model for each group is represented by large numbers (blue) and an solid star. (A–D) The eight sets of 100 best-fit models from the experimental x-ray-scattering curves for glycosylated and deglycosylated IgG1 were grouped by PCA into five groups as shown in three (A–C) of PC2 versus PC1, PC3 versus PC2, and PC3 versus PC1. PC1, PC2, and PC3 are the first three principal components of the analysis. (D) The first three eigenvalue rankings (PC1, PC2, and PC3) captured 81.9% of the variance in the 800 models. (E–H) The four sets of 100 best-fit models from the experimental neutron-scattering curves for glycosylated and deglycosylated IgG1 were grouped by PCA into five groups as shown in (E–G) of PC2 versus PC1, PC3 versus PC2, and PC3 versus PC1. (H) The first three eigenvalue rankings (PC1, PC2, and PC3) captured 70.4% of the variance in the 400 models. To see this figure in color, go online.

Figure 10.

Scattering curve fits to the experimental data for the best-fit model for each of the glycosylated and deglycosylated IgG1 samples. The experimental curve is denoted by black circles and the best-fit modeled curves are denoted by solid red lines. The distance distribution curves P(r) are shown in the top right of each panel. (A) Glycosylated and (B) deglycosylated IgG1 x-ray-scattering curve fits for four concentrations each. For the four x-ray fits in (A), the glycosylated IgG1 models were taken from PCA Group 3 (3.60, 3.19, and 2.06 mg/mL) and Group 2 (1.36 mg/mL) in that order (Table 2). In (B) the deglycosylated IgG1 models used for the fits as shown corresponded to PCA Group 4 (4.29 mg/mL) and Group 1 (3.19, 1.34 and 1.02 mg/mL) in that order. (C) Glycosylated and (D) deglycosylated IgG1 neutron curve fits shown for two concentrations each. In the four neutron fits, the glycosylated IgG1 models corresponded to PCA Group 1 (2.60 mg/mL) and PCA Group 2 (1.38 mg/mL) in that order (Table 2). Likewise, the deglycosylated IgG1 models corresponded to PCA Group 3 (2.73 mg/mL) and PCA Group 4 (0.90 mg/mL) in that order. The experimental curves at 3.60 and 4.29 mg/mL and their modeled curve fits are in Data S1. To see this figure in color, go online.

The same outcome was found with the theoretical neutron modeling and PCA, thus confirming the reproducibility of the curve fits, although the precision of the neutron-scattering curves was reduced. The same 123,284 and 119,191 theoretical curves were compared with the neutron-scattering curves at 0.90–2.73 mg/mL to show again that 100 best-fit structures could be identified at clear minima in each of the R-factor versus R_G neutron distributions (Fig. 8, C and D). The neutron PCA also indicated clear differences between the glycosylated and deglycosylated neutron IgG1 models (black and magenta, respectively, Fig. 9, E–H; Table 3). The distributions of the 200 best-fit glycosylated and 200 best-fit deglycosylated neutron models were each again clustered into two major groups, 1 and 4, and three more less-populated groups, with little overlap between glycosylated and deglycosylated groups. The glycosylated models mostly occurred in the PCA Group 1, whereas the deglycosylated models mostly occurred in the PCA Group 4. Visually excellent neutron curve fits were obtained (Fig. 10, C and D).

Further insights into the x-ray and neutron data and their modeling were obtained from the dimensionless Kratky analyses of (Q.R_G)².I(Q)/I(0) versus Q.R_G for the experimental scattering curves at the highest concentrations in use and the scattering curves from the modeled best-fit structures. These plots indicate whether the macromolecule in question is globular in its structure or possesses intrinsically disordered regions (72). The Kratky plots all demonstrated two clear peaks in both the experimental and theoretical modeled curves. For the x-ray Kratky curves (Fig. 11 A), the Q.R_G-values for the experimental glycosylated peaks were 1.96 and 4.05, which were in good accord with the modeled values of 1.93 and 3.98. The Q.R_G-values for the experimental deglycosylated peaks of 2.01 and 4.15 were also in good accord with the modeled deglycosylated peaks of 1.97 and 4.08. It was interesting to note that the second peak showed higher intensities for deglycosylated IgG1 (magenta) than for glycosylated IgG1 (black), suggesting that there was a small increase in antibody disorder after deglycosylation. For the SANS Kratky curves (Fig. 11 B), the Q.R_G-values for the experimental peaks for glycosylated IgG1 were 2.45 and 4.68, which were similar to the modeled peaks at 1.99 and 4.43. The Q.R_G-values for the experimental peaks for deglycosylated IgG1 were 1.91 and 5.16, but showed less agreement for the modeled peaks at 1.91 and 4.37. Again, the second neutron peak showed higher intensities for deglycosylated IgG1 (magenta) when compared with glycosylated IgG1 (black), suggesting that a greater disorder was present after deglycosylation.

Figure 11.

Normalized Kratky plots for the experimental and best-fit glycosylated and deglycosylated IgG1 scattering curves. (A) X-ray experimental data (solid lines) and model fits (dashed lines) were shown in black for glycosylated IgG1 at 3.60 mg/mL and in magenta for deglycosylated IgG1 at 4.29 mg/mL. (B) Neutron experimental data (solid lines) and model fits (dashed lines) were shown in black for glycosylated IgG1 at 2.60 mg/mL and in magenta for deglycosylated IgG1 at 2.73 mg/mL. To see this figure in color, go online.

As another test of the scattering modeling, the s⁰_20,w-values for the eight sets of 100 best-fit glycosylated and deglycosylated models from each x-ray concentration (Figs. 8 and 10) were calculated using HullRad (64). This gave an s⁰_20,w range of 6.57–6.77 S for the four x-ray concentrations for glycosylated IgG1 and 6.24–6.50 S for deglycosylated IgG1 (Table 2). These values agreed well with the experimental s⁰_20,w-values of 6.16–6.43 S for glycosylated IgG1 and 6.09–6.15 S for deglycosylated IgG1 (Table 1). These agreements corroborated the outcome of the atomistic scattering modeling, given that the mean difference between the modeled and experimental values should typically be ±0.21 S for related macromolecules (73). This modeling was however unable to distinguish changes before and after deglycosylation.

Discussion

This scattering and atomistic modeling study has notably clarified the conformational effect of removing the two glycans chains on the structure of the major IgG1 antibody subclass. Unlike earlier protein structural investigations based on crystallography, NMR, or circular dichroism (CD), our approach provided a more informative outcome on full-sized IgG1 of the changes accompanying deglycosylation of the two conserved Asn²⁹⁷ residues in the Fc region (Fig. 1 B). The complete deglycosylation of IgG1 was validated by a combination of gel filtration, routine mass spectrometry, and AUC. The AUC data showed that the IgG1 samples were monomeric and showed a slight concentration dependence in the s_20,w-values that were extrapolated to give the s⁰_20,w-values. Subsequently, the glycosylated and deglycosylated IgG1 proteins were submitted to abundant SAXS and SANS data collection to establish their Guinier R_G-, R_XS-1- and R_XS-2-values, and their distance distribution curves P(r). Small changes in the R_XS-1-values and larger changes in the M2 parameter that monitored the mean separation of the Fab and Fc regions in IgG1 were seen on deglycosylation. The advent of atomistic scattering modeling using SASSIE (34) based on molecular dynamics and MC simulations gave excellent curve fits based on large stereochemically correct trial conformational libraries for IgG1. The display and interpretation of the output was much facilitated by PCA analyses. Two different conformational best-fit structures for glycosylated and deglycosylated IgG1 were identified by PCA. The clearest view of the final result was determined from wireframe representations of the 100 best-fit structures (Fig. 12). Views of the 100 best-fit x-ray models at four concentrations showed that the glycosylated Fc region occupied a smaller volume (blue wireframe, Fig. 12 A) than the notably larger volume occupied by the deglycosylated Fc regions (magenta wireframe, Fig. 12 A). This outcome showed that the Fab and Fc regions formed better-defined native glycosylated structures compared with the more dispersed and flexible structures seen after deglycosylation. The 100 best-fit neutron models were limited by the reduced precision of the neutron-scattering data, but are consistent with this interpretation (Fig. 12 B). The Kratky plots also suggest greater disorder after deglycosylation (Fig. 11), in keeping with the flexibility shown in Fig. 12 A.

Figure 12.

Views of representative best-fit x-ray and neutron structures. The blue-ribbon cartoon denotes the protein backbone of the starting glycosylated and deglycosylated IgG1 structures. The Fab regions of the 100 best-fit models were superimposed onto these starting structures, thus focusing on movements in the Fc region. The blue and magenta wireframe envelopes denote the space occupied by the glycosylated and deglycosylated Fc regions, respectively, in the 100 best-fit structures for each of the (A) four x-ray and (B) two neutron analyses. In five of the six representations, the magenta wireframes (deglycosylated IgG1) occupy a greater region of space compared with the blue wireframes (glycosylated IgG1). (C) The cartoon representation based on part (b) of (A) showed the larger range of Fc conformations as arrowed (magenta) after deglycosylation compared with that for glycosylated IgG1 (blue). To see this figure in color, go online.

Our results (Fig. 12 A) account for previous functional studies of human IgG1. The Fc region of IgG1 is responsible for interactions with the three classes of human FcγRs and with the globular heads of human C1q. Several studies have reported that both interactions are abrogated after deglycosylation. Thus, the antibody effector functions such as antibody-dependent cytotoxicity and complement-dependent cytotoxicity mediated by FcγRs and C1q are impaired for aglycosylated and deglycosylated antibodies (11,28,74). Deglycosylated IgG1 was used as a negative control in surface plasmon resonance experiments in which several FcγRs were tested and deglycosylated IgG1 failed to bind to these with the exception of the high-affinity FcγRI (75). Our structural investigations explains these findings by showing that the orientation of the Fc region in IgG1 has become disorganized (Fig. 12 A). The cartoon view (Fig. 12 C) showed that the deglycosylated Fc region occupied more conformational space than the glycosylated one. High-resolution crystal structures of glycosylated Fc regions bound to the FcγRs showed none or little interaction with the Fc glycan chains, and receptor binding generally occurred at residues in the lower hinge (22,29,76,77). Accordingly, from our work, IgG1 deglycosylation means that the essential presentation of structurally well-defined C_H2 domain surfaces in the Fc region near the hinge peptides is no longer present. It comes after this conclusion that the glycoengineering of human IgG1 based on either the removal or modification of specific glycosylation patterns in the two Fc glycans will influence receptor binding and in turn the effector functions if human IgG1, such as its core fucosylation (9,25,78) and terminal sialic acids (79).

The advantage of our combined SAXS-SANS-AUC-MC approach is the ability to address the full IgG1 structure, this being the functional native structure. Previous structural studies have focused on the Fc region alone because it is difficult to crystallize the full IgG1 antibody compared with the Fc region alone, and NMR and CD studies are more difficult with the full IgG1 structure because its molecular mass has tripled to over 150 kDa. Previous NMR solution studies of the glycosylated and deglycosylated Fc region (80) showed that the glycan chain at Asn²⁹⁷ stabilizes the loop between β-strands C′ and E in the C_H2 domains and, in turn, positions the two C_H2 domains into a stable orientation seen in 22 Fc crystal structures to orient the FcγR interface on the Fc region for optimal-binding affinity with receptors. Previous crystallography studies of the deglycosylated Fc region showed that the C_H2 domains had reorientated themselves to form a more compact structure. (31). CD solution studies of the glycosylated and deglycosylated Fc region studied the β-sheet secondary structure of its four domains (Fig. 1; (81)). By this, similar β-sheet-rich structures showing a minimum at 217 nm were observed for both forms of the Fc region, showing that this β-sheet structure was preserved with or without the glycan chains. The greater mobility of the Fc region after glycan removal, as observed in this study (Fig. 12 A), would not have been observed by CD. These previous studies on the Fc region alone complement our results showing that the Fc region within intact IgG1 is more flexible after deglycosylation.

The advent of atomistic scattering modeling has resulted in a molecular explanation of the changes induced in IgG1 after glycan removal. Traditionally, solution scattering is a low-resolution method with a resolution of around 2 nm, whereas protein crystallography routinely achieves resolutions that are over 10 times better. Our recent IgG1 modeling analysis used “constrained” modeling based on the Fab and Fc crystal structures and joined by a hinge region that was conformationally randomized using molecular dynamics to give 20,000 trial models (35). Seven different fits from x-ray and neutron data out to Q-values of 1.5 nm⁻¹ for human IgG1 6a and 19a in different NaCl buffers corresponded to clear minima in the R-factor versus R_G graphs. All these revealed an asymmetric solution structure for IgG1, in agreement with the single asymmetric conformation seen in the crystal structure of human IgG1 b12 (19) and that for glycosylated IgG1 from SasCalc (Fig. 12). The best-fit R-factors were 2.8–3.7%. This conformation permitted the Fc region to bind readily to its FcγR and C1q ligands without steric clashes. The follow-up study on human IgG1 using “atomistic” scattering modeling involved further x-ray and neutron data collection out to Q-values of 1.5 nm⁻¹, and fitting these data to 231,492 trial models produced from full molecular dynamics and rapid MC simulations (36). The best-fit R-factors were 2.9%. This improved method likewise gave an asymmetric IgG1 solution structure similar to that seen in the crystal structure of human IgG1 b12 (19), and also that for glycosylated IgG1 from SasCalc (Fig. 12). This likewise accounted for the binding of the Fc region to its FcγR and C1q ligands. Of particular interest is that the previous use of the SCT/SCTPL modeling approach (available in SASSIE-web) had explicitly incorporated hydration shells in a coarse-grained approach (47). The current atomistic modeling study using SasCalc (62) did not include atomistic representation of hydration shells because this is computationally expensive. The final outcomes from all three x-ray modellings were similar when all three showed asymmetric IgG1 solution structures. In this study, the simulations of IgG1 A33 were based on x-ray and neutron data that extended out to Q-values of 1.5 nm⁻¹ and resulted in fits with low R-factors of 1% or less (Table 2). This R-factor improvement is attributed to the improved signal/noise ratio of the scattering curves from the B21 instrument at Diamond. The atomistic modeling approach in combination with high-quality SAXS data with little noise at large Q-values has been of great value in studying structural perturbations in IgG antibodies caused by the removal of its two glycans.

Author contributions

V.A.S. obtained AUC, SAXS, and SANS data and analyzed these, performed the modeling analyses, and wrote the article. J.D., R.P.R., and J.G. assisted with the SANS, SAXS, and AUC data collection, respectively. P.A.D. supervised the work. S.J.P. conceived and coordinated the study, supervised the work, and wrote the article.

Editor: Jill Trewhella.