Of the four human IgG antibody subclasses IgG1–IgG4, IgG4 is of interest in that it does not activate complement and exhibits atypical self-association, including the formation of bispecific antibodies. The solution structures of antibodies are critical to understand function and therapeutic applications. Thus IgG4 was studied by synchrotron X-ray scattering. The Guinier X-ray radius of gyration RG increased from 5.0 nm to 5.1 nm with an increase of concentration. The distance distribution function P(r) revealed a single peak at 0.3 mg/ml, which resolved into two peaks that shifted to smaller r values at 1.3 mg/ml, even though the maximum dimension of IgG4 was unchanged at 17 nm. This indicated a small concentration dependence of the IgG4 solution structure. By analytical ultracentrifugation, no concentration dependence in the sedimentation coefficient of 6.4 S was observed. Constrained scattering modelling resulted in solution structural determinations that showed that IgG4 has an asymmetric solution structure in which one Fab–Fc pair is closer together than the other pair, and the accessibility of one side of the Fc region is masked by the Fab regions. The averaged distances between the two Fab–Fc pairs change by 1–2 nm with the change in IgG4 concentration. The averaged conformation of the Fab regions appear able to hinder complement C1q binding to the Fc region and the self-association of IgG4 through the Fc region. The present results clarify IgG4 function and provide a starting point to investigate antibody stability.

INTRODUCTION

The average serum concentrations of the four human IgG subclasses (IgG1, IgG2, IgG3 and IgG4) are 8 mg/ml, 4 mg/ml, 0.8 mg/ml and 0.4 mg/ml respectively [1]. The four subclasses IgG1–IgG4 primarily differ at their hinges that connect the Fab and Fc regions (Figure 1) [1,2]. Whereas human IgG2 and IgG4 both have hinges of 12 residues in length, human IgG1 and IgG3 have hinges of length 15 or 62 residues respectively. The hinge is composed of upper, middle and lower sections. The upper hinge connects the Fab region to the cysteine-rich middle hinge, while the lower hinge connects the Fc region to the middle hinge. The genetic hinge corresponds to the upper and middle hinges, while the lower hinge is encoded by the CH2 exon, and the three hinge sections in the four subclasses differ in size. The subclasses relate to different activities, in which IgG4 with the shortest upper hinge is unable to activate complement, whereas IgG1 and IgG3 are very effective, and IgG2 is only effective at high concentrations [2]. IgG4 also shows poorer binding to Fc receptors [1]. Additionally, IgG4 differs in being monovalent, binding to two different antigens through two different Fab regions [3,4]. A mutation from proline to serine in the middle hinge shifts the disulfide-bond equilibrium in IgG4 and allows the ready exchange of pairs of heavy and light chains to form bispecific antibodies [3]. Such exchanges may play an important role in the in vivo function of IgG4. The 25–75% loss of this disulfide bond potentially leads to greater conformational freedom of the IgG4 hinge region that may be expected to vary with solution conditions. In the pharmaceutical industry, even though antibodies comprise one-third of drugs, and IgG1 monoclonal antibodies are the most commonly exploited subclass, IgG4 is the basis of several successfully approved biopharmaceuticals for reason of its lack of inflammatory reaction in treatments, including those for myeloid leukaemia (gemtuzumab ozogamicin), paroxysmal nocturnal haemoglobinaemia (eculizumab), and multiple sclerosis (natalizumab) [5,6].

Domain diagram of the human–mouse chimaera IgG4 structure

Figure 1
Domain diagram of the human–mouse chimaera IgG4 structure

The heavy chains consist of VH, CH1, CH2 and CH3 domains, and the light chains each comprise a VL and CL domain. The heavy chains are connected by two Cys–Cys disulfide bridge bonds at Cys243 and Cys246. There is one N-linked oligosaccharide site at Asn314 on the two CH2 domains. The hinge region between the Fab and Fc fragments is composed of 14 residues (KYGPPCPSCPAPEF) between Ser234 and Leu252. Both the Cys243–Cys243 and Cys246–Cys246 bonds are weakened by the Ser245 substitutions.

Figure 1
Domain diagram of the human–mouse chimaera IgG4 structure

The heavy chains consist of VH, CH1, CH2 and CH3 domains, and the light chains each comprise a VL and CL domain. The heavy chains are connected by two Cys–Cys disulfide bridge bonds at Cys243 and Cys246. There is one N-linked oligosaccharide site at Asn314 on the two CH2 domains. The hinge region between the Fab and Fc fragments is composed of 14 residues (KYGPPCPSCPAPEF) between Ser234 and Leu252. Both the Cys243–Cys243 and Cys246–Cys246 bonds are weakened by the Ser245 substitutions.

Structural studies of the human IgG subclasses are important for functional and therapeutic applications. Knowledge of the relative positions of the Fab and Fc regions in solution is important for both antigen and receptor interactions [1,2,7,8]. In this context, protein crystal structures of intact IgG antibodies are relatively few and correspond to single-hinge conformations that are trapped within the crystallographic unit cell [912]. Likewise, a knowledge of antibody aggregation behaviour during manufacture, shipping and storage is critical to minimizing unwanted immunogenic effects in vivo [13,14]. This understanding is based largely on experimental studies of solution parameters such as temperature and process environment [15,16]. Therefore antibody structures in solution are required under a wide range of conditions. So far, such structure determinations of IgG antibodies have been restricted to relatively few examples. Neutron-scattering structures for the bovine IgG1 and IgG2 subclasses were determined by constrained modelling based on Fab and Fc crystal structures [17]. Low-resolution solution conformations for human IgG4 were obtained by hydrodynamic ellipse and bead modelling of ultracentrifugation data [1821]. Other approaches have included electron microscopy in vacuo, which showed that IgG4 demonstrated less hinge flexibility compared with IgG1 or IgG3 and was less able to form bivalent ring dimers [22].

In the present paper, we describe solution-structure determinations for IgG4 by X-ray scattering and analytical ultracentrifugation combined with constrained modelling based on the human IgG1 crystal structure [10]. This combination of methods yields more detailed antibody solution structures than those from hydrodynamic bead modelling alone [23,24]. By this approach, solution structures have been determined for pentameric human IgM (PDB code 2RCJ), human IgA1 (PDB code 1IGA) and IgA2 (PDB code 1R70), human IgD (PDB code 1ZVO) and a murine Crry–Fc chimaeric therapeutic antibody (PDB code 1NTL) [17,2529]. Because X-ray scattering data can now be measured within seconds [30], it is possible to follow protein conformations as a function of different solution conditions [31]. In the present study, in combination with analytical ultracentrifugation to establish the monodispersity of IgG4 [32], we show that the averaged solution structure of IgG4 possesses semi-extended hinges and exhibits a detectable conformational dependence on concentration. We show that the Fab regions mask one side of the Fc region in IgG4, unlike the arrangement seen in the IgG1 crystal structure, and this may perturb the functional ability of IgG4 to bind complement and to self-associate.

EXPERIMENTAL

Purification and composition of IgG4

A human–mouse chimaeric IgG4 B72.3 that binds to tumour cells was generously supplied by Dr Jim Davies, Lonza Biologics. IgG4 was further purified by gel-filtration chromatography using a Superose 6/300 column (GE Healthcare) to remove large non-specific aggregates, then dialysed extensively at 4 °C against PBS (10.6 mM sodium phosphate, 1.5 mM potassium phosphate, 138 mM NaCl and 2.7 mM KCl, pH 7.4). In this IgG4, the heavy- and light-chain amino acid sequences of the human–mouse chimaera Fab region have been reported in its PDB structure file (PDB code 1BBJ [33]), and included its human κ light chain. In this PDB file, the C-terminal cysteine residue is truncated in the CL sequence (see Supplementary Figure S1 at http://www.biochemj.org/bj/432/bj4320101add.htm); this is expected to be bridged with Cys148 in the CH1 domain. The engineered variable domains VH and VL originated from a mouse antibody, whereas the remaining sequence is from human IgG4. The CH1 domain, hinge and Fc region sequences were taken from its GenBank® sequence (accession code AAG00912). The carbohydrate composition was assumed to be similar to that for N-linked oligosaccharides in human IgG, in which each of the Asn314 residues in the Fc region (Figure 1) contain a complex-type oligosaccharide with a Man3GlcNAc2 core and two NeuNAc-Gal-GlcNAc antennae. This resulted in a IgG4 calculated molecular mass of 145.5 kDa, an unhydrated volume of 187.2 nm3, a hydrated volume of 246.6 nm3 (based on a hydration of 0.3 g of water/g of glycoprotein and an electrostricted volume of 0.0245 nm3 per bound water molecule), a partial specific volume v of 0.728 ml/g, and an absorption coefficient at 280 nm (1%, 1 cm path length) of 14.7 [34]. The buffer density of PBS was taken as the default value of 1.00543 g/ml; this value was corroborated at 20 °C using an Anton Paar DMA 5000 density meter to be 1.00526 g/ml.

Sedimentation-velocity data for IgG4

Analytical ultracentrifugation data for IgG4 were obtained on two Beckman XL-I instruments equipped with AnTi50 and AnTi60 rotors. Sedimentation-velocity data were acquired at 20 °C at rotor speeds of 50000 rev./min and 60000 rev./min in two-sector cells with column heights of 12 mm for approx. 6 h. Sedimentation analysis was performed using direct boundary Lamm fits of up to 300 scans using SEDFIT (version 11.71) [35,36]. SEDFIT resulted in size-distribution analyses c(s), the algorithm for which assumes that all species have the same frictional ratio f/fo. The final SEDFIT analyses used a fixed resolution of 200 and optimized the c(s) fit by floating f/fo 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.

X-ray scattering data for IgG4

X-ray scattering data were obtained during one beam session in four-bunch mode at the ESRF (European Synchrotron Radiation Facility), Grenoble, France, operating with a ring energy of 6.0 GeV on the Beamline ID02 [37]. Storage ring currents ranged from 31 mA to 39 mA. Data were acquired using a recently installed fibre-optically coupled high sensitivity and dynamic range CCD detector (FReLoN). Together with a smaller beam stop and the use of a detector resolution of 512×512 pixels, this allowed a shorter sample-to-detector distance of 2.0 m to be used and resulted in higher signal-to-noise ratios [38]. IgG4 was studied at four concentrations between 0.3 mg/ml and 1.3 mg/ml, in which samples of volume 100 μl were measured in flow cells to reduce radiation damage by moving the sample continuously during beam exposure. Sets of ten time frames each, with frames of duration 0.1 s or 0.2 s each, were acquired in quadruplicate as a control of reproducibility. Online checks during data acquisition confirmed the absence of radiation damage, after which the ten frames were averaged. Other details including the data-reduction procedure are described elsewhere [39]. A second X-ray beam session utilized IgG4 at 0.3 mg/ml only, which was used to verify the above data set, but not used here.

In a given solute–solvent contrast, the radius of gyration RG is a measure of structural elongation if the internal inhomogeneity of scattering densities within the protein has no effect. Guinier analyses at low Q (where Q = 4π sin θ/λ; 2θ is the scattering angle and λ is the wavelength) gives the RG, and the forward scattering at zero angle I(0) [40]:

 
formula

This expression is valid in a Q·RG range up to 1.5. If the structure is elongated, the mean radius of gyration of cross-sectional structure RXS and the mean cross-sectional intensity at zero angle [I(Q)Q]Q→0 is obtained from:

 
formula

For immunoglobulins, the cross-sectional plot exhibits two regions, a steeper innermost one and a flatter outermost one [41], and the two analyses are identified by RXS−1 and RXS−2 respectively. The RG and RXS analyses were performed using an interactive PERL script program SCTPL7 (J.T. Eaton and S.J. Perkins, unpublished software) on Silicon Graphics OCTANE Workstations. 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 the program GNOM [42]:

 
formula

P(r) corresponds to the distribution of distances r between volume elements. This yields the maximum dimension of the macromolecule L and its most commonly occurring distance vector M in real space. For this, the X-ray I(Q) curve utilized up to 246 data points in the Q range between 0.06 nm−1 and 1.75 nm−1.

Debye scattering and sedimentation-coefficient modelling of IgG4

The homology model for IgG4 was constructed from the crystal structures of Fab B72.3 and human IgG1 (PDB codes 1BBJ and 1HZH) [11,33]. The Fab B72.3 crystal structure incorporated the IgG4 residues up to Val232. A homology model for IgG4 between residues Cys243 and Leu462 was created using SWISS-MODEL software available at http://swissmodel.expasy.org/ [43]. There is a 94.1% sequence identity between the two sequences. The nine-residue peptide used for randomization was V232ESKYGPPC243 (Supplementary Figure S1), in which the central seven residues that correspond to the upper hinge were conformationally randomized using molecular dynamics, and the outermost residues were anchor points used to superimpose the hinge on to the Fab and Fc regions in order to create the full IgG4 model. The nine-residue peptide was first created in an extended β-strand conformation. Following method 2 used for human IgA1 [26], two libraries of 5000 randomized hinge conformations were generated. The nine-residue peptides were constrained to be of minimum lengths either (i) 2.2–3.1 nm or (ii) 3.1–3.15 nm, where the latter hinges were almost fully extended in length. A minimum hinge length was required in order to prevent the hinge adopting abnormally short structures. Each hinge in duplicate was used to join the two Fab and Fc regions to create 10000 IgG4 models. The centre-of-mass of each Fab or Fc region is defined as the midpoint of the line joining the midpoint of the four α-carbon atoms in the disulfide bridges of the β-strands B and F (i) between the VH and VL domains, and (ii) between the CH1 and CL domains.

Each full IgG4 model was used to calculate its X-ray scattering curve for comparison with the four experimental IgG4 curves (Table 1) using Debye sphere models [44]. A cube-side length of 0.536 nm in combination with a cutoff of four atoms was used to convert the atomic co-ordinates into Debye sphere models with 1189 spheres that corresponded to the unhydrated structure. Because hydration shells are visible by X-rays, a hydration shell corresponding to 0.3 g of water/g of protein was created using HYPRO [45], where the optimal total of hydrated spheres in the IgG4 model was 1566 (Table 1). The X-ray scattering curve I(Q) was calculated using the Debye equation adapted to spheres [44]. Details are given elsewhere [46]. Steric overlap between the Fab and Fc regions in the models was assessed using the number of spheres N in the models after grid transformation, where models showing less than 95% of the required total of 1566 spheres were discarded. Over 66% of the 10000 models met this ‘absence of steric overlap’ criterion. Next, the X-ray RG and RXS−1 values were calculated from the modelled curves in the same Q ranges used for the experimental Guinier fits. This allowed for any approximations inherent in the use of the Q·RG fit range up to 1.5. Models that passed RG and RXS−1 filters of±5% were then ranked using a goodness-of-fit R-factor defined by analogy with protein crystallography and based on the experimental curves in the Q range extending to 1.8 nm−1 (Table 1). The distances d1, d2 and d3 were determined from the centres-of-mass of the Fab and Fc regions (excluding hydrogen atoms) using a script written in PYTHON. The three angles between the Fab and Fc regions were defined as the angle of intersection from the dot product between two vectors, also using a PYTHON script. Each vector was the long axis through each Fab or Fc region, where the axes were defined as the lines passing through the centres of gravity between each cluster of four cysteine α-carbon atoms at the two ends of the Fab and Fc regions (one cluster at each end of each Fab or Fc region, corresponding to the conserved disulfide bridge in each immunoglobulin fold domain). Surface accessibilities were calculated using the Lee–Richards method using a PYTHON script to allow for different probe radii between 0.14 nm and 5 nm. Artwork was prepared using PyMOL software (De Lano Scientific; http://www.pymol.org/).

Table 1
X-ray scattering and sedimentation coefficient data for IgG4 and their modelling

n.a., not available.

(a) Experimental data 
 RG (nm) RXS−1 (nm) RXS−2 (nm) Dmax (nm)  s20,w (S) 
IgG4 (1.30 mg/ml) 4.98±0.08; 5.08±0.12a 2.55±0.03 1.41±0.02 17  6.44b 
IgG4 (0.98 mg/ml) 5.00±0.09; 5.03a 2.56±0.01 1.44±0.02 17  6.44b 
IgG4 (0.65 mg/ml) 5.03±0.03; 5.04a 2.57±0.03 1.46±0.06 17  n.a. 
IgG4 (0.30 mg/ml) 4.92±0.25; 4.96a 2.55±0.08 1.26±0.05 17  n.a. 
(b) Modelling of IgG4c 
 RG (nm) RXS−1 (nm) RXS−2 (nm) Dmax (nm) R factor (%) s20,w (S) 
IgG4 (1.30 mg/ml) 4.87 2.61 1.35 17d 2.0 6.69, 6.46e 
IgG4 (0.98 mg/ml) 4.93 2.55 1.49 17d 2.0 6.51, 6.67e 
IgG4 (0.65 mg/ml) 4.84 2.60 1.46 17d 2.3 6.71, 6.54e 
IgG4 (0.30 mg/ml) 4.91 2.54 1.22 17d 3.8 6.65, 6.41e 
(a) Experimental data 
 RG (nm) RXS−1 (nm) RXS−2 (nm) Dmax (nm)  s20,w (S) 
IgG4 (1.30 mg/ml) 4.98±0.08; 5.08±0.12a 2.55±0.03 1.41±0.02 17  6.44b 
IgG4 (0.98 mg/ml) 5.00±0.09; 5.03a 2.56±0.01 1.44±0.02 17  6.44b 
IgG4 (0.65 mg/ml) 5.03±0.03; 5.04a 2.57±0.03 1.46±0.06 17  n.a. 
IgG4 (0.30 mg/ml) 4.92±0.25; 4.96a 2.55±0.08 1.26±0.05 17  n.a. 
(b) Modelling of IgG4c 
 RG (nm) RXS−1 (nm) RXS−2 (nm) Dmax (nm) R factor (%) s20,w (S) 
IgG4 (1.30 mg/ml) 4.87 2.61 1.35 17d 2.0 6.69, 6.46e 
IgG4 (0.98 mg/ml) 4.93 2.55 1.49 17d 2.0 6.51, 6.67e 
IgG4 (0.65 mg/ml) 4.84 2.60 1.46 17d 2.3 6.71, 6.54e 
IgG4 (0.30 mg/ml) 4.91 2.54 1.22 17d 3.8 6.65, 6.41e 
a

The first value is from the Guinier RG analyses (Figure 4a) and the second one is from the GNOM P(r) analyses (Figure 5a).

b

The values were measured at 1.6 mg/ml and 0.8 mg/ml.

c

The full description of the modelling is reported in Supplementary Table S1 at http://www.biochemj.org/bj/432/bj4320101add.htm.

e

The first value corresponds to that from HYDRO; the second is from HYDROPRO.

Sedimentation coefficients s °20,w were calculated directly from the hydrated Debye sphere models using the program HYDRO [47]. They were also calculated from the atomic co-ordinates in the HYDROPRO shell modelling program using the default value of 0.31 nm for the atomic-element radius for all atoms to represent the hydration shell [48]. Previous applications of these calculations to antibodies are reviewed elsewhere [24,49].

PDB accession number

The best-fit IgG4 α-carbon co-ordinate models have been deposited in the PDB with the accession codes 3M7X (1.3 mg/ml), 3M7Y (0.98 mg/ml), 3M7Z (0.65 mg/ml) and 3M80 (0.30 mg/ml).

RESULTS

Purification and analytical ultracentrifugation of IgG4

A chimaeric mouse–human IgG4 antibody that possesses Ser245 and not Pro245 in its hinge (alternatives names: Ser228 in [2]; Ser331 in [20]; and Ser241 in [21]) was used. In this chimaera, the mouse VH and VL domains are combined with the human CL, CH1, CH2 and CH3 domains (Figure 1). IgG4 was subjected to gel filtration to remove potential non-specific aggregates immediately prior to ultracentrifugation or scattering experiments, and eluted as a single peak with a trailing edge (Figure 2a). Purified IgG4 showed four bands on SDS/PAGE (4–20% gel) under non-reducing conditions. These indicate the observation of the intact IgG4 molecule (apparent molecular mass 190 kDa) together with separated pairs of disulfide-linked heavy–light chains (apparent molecular mass 90 kDa) and the individual heavy and light chains on their own (apparent molecular masses 60 kDa and 25 kDa) (Figure 2b). This is consistent with the presence of the two known isomeric forms of IgG4, one formed by interchain Cys–Cys bridges at Cys243 and Cys246, and the other lacking inter-heavy-chain disulfides (Figure 1) [3,50]. Only heavy chains (apparent molecular mass 55 kDa) and light chains (apparent molecular mass 25 kDa) were observed in reducing conditions (Figure 2c). This purified IgG4 was found to be stable upon storage at 4 °C for at least 1 month.

Purification of IgG4

Figure 2
Purification of IgG4

(a) Elution peak from a Superdex 200 gel-filtration column, from which fractions 14–19 were pooled. mAu, milli-absorbance units. (b) Non-reduced SDS/PAGE analysis of IgG4. The molecular-mass markers in the left-hand lane are denoted in kDa, while the right lane corresponds to IgG4. H, heavy chain; L, light chain. (c) The corresponding reduced SDS/PAGE analysis of IgG4 is shown.

Figure 2
Purification of IgG4

(a) Elution peak from a Superdex 200 gel-filtration column, from which fractions 14–19 were pooled. mAu, milli-absorbance units. (b) Non-reduced SDS/PAGE analysis of IgG4. The molecular-mass markers in the left-hand lane are denoted in kDa, while the right lane corresponds to IgG4. H, heavy chain; L, light chain. (c) The corresponding reduced SDS/PAGE analysis of IgG4 is shown.

Sedimentation-velocity experiments were performed as a check of monodispersity and for shape determinations. Runs were performed for IgG4 concentrations of 0.8 mg/ml (5.5 μM) and 1.6 mg/ml (11 μM) (see the Experimental section). The SEDFIT analysis of Figure 3 involved fits to as many as 150 scans, and the good agreement between the experimental boundary scans and fitted lines is clear. A regular sedimentation profile was observed, from which only a single peak was observed in the size-distribution analysis c(s) in SEDFIT, with no other significant peaks observed. This single peak indicated that the IgG4 sample was monodispersed in solution, with no evidence for the separation of the two isomeric forms [50] or the existence of smaller species that had appeared in the harsh environment of the non-reducing SDS/PAGE analysis (Figure 2b). This agrees with previous reports of IgG4 stability [3]. This also agrees with the <1% dimers or oligomers seen for IgG4 when this was subjected to size-exclusion chromatography [51]. The major peak was observed at s20,w values of 6.44 S. Since the sedimentation coefficient s20,w is a structural parameter of macromolecular elongation in solution, it can be compared with those for other related antibody molecules. These s20,w values are comparable with those reported for human IgA1 of 6.15–6.20 S [26], human IgA2 of 6.42–6.57 S [28] and human IgG4 of 6.20–6.80 S [1821]. The calculation of the frictional coefficient ratio f/f0 for IgG4 normalizes the shape parameter by comparing this with the f0 for a sphere of the same volume. From SEDFIT analyses, this was determined to be 1.37. From the value of 6.44 S and the composition, this ratio is 1.30 for IgG4, which is less than that of 1.56 for IgA1 and 1.53 for IgA2, and indicates that IgG4 is more compact in shape than IgA.

Sedimentation-velocity analyses of IgG4

Figure 3
Sedimentation-velocity analyses of IgG4

The experimentally observed sedimentation boundaries at (a) 1.6 mg/ml and (b) 0.8 mg/ml were recorded using a rotor speed of 50000 rev./min. Thirty boundaries (black) are shown from 150 scans, these being taken at intervals of every fifth scan for clarity and fitted as shown (white lines). The size-distribution analyses c(s) revealed single peaks at sedimentation coefficients of 6.44 S in both cases.

Figure 3
Sedimentation-velocity analyses of IgG4

The experimentally observed sedimentation boundaries at (a) 1.6 mg/ml and (b) 0.8 mg/ml were recorded using a rotor speed of 50000 rev./min. Thirty boundaries (black) are shown from 150 scans, these being taken at intervals of every fifth scan for clarity and fitted as shown (white lines). The size-distribution analyses c(s) revealed single peaks at sedimentation coefficients of 6.44 S in both cases.

X-ray scattering of IgG4

The solution arrangement of the Fab and Fc fragments of IgG4 was analysed by synchrotron X-ray scattering. Data collection was carried out between 0.3 and 1.3 mg/ml (2–9 μM) at the ESRF, using time-frame analyses to ensure the absence of radiation-damage effects. Guinier analyses to measure the degree of elongation of IgG4 resulted in linear plots in three distinct regions of the I(Q) curves, as expected for antibodies, from which the RG, RXS−1 and RXS−2 values were obtained within satisfactory Q·RG and Q·RXS limits (Figure 4). Only trace amounts of aggregates were present, as shown by the slight upturn in the curves at the smallest Q values. These aggregates occurred before the Q range used for the Guinier analyses and did not perturb these analyses. The mean X-ray RG value for IgG4 varied between 4.92 nm and 5.03 nm, with a mean of 4.99 nm (Table 1). On dilution, these increased slightly from 4.98 nm to 5.03 nm, then decreased to 4.92 nm. The RXS−1 values were also similar in the concentration range, giving a mean value of 2.56 nm, and likewise the RXS−2 values were similar, giving a mean value of 1.40 nm (Table 1). These RG, RXS−1 and RXS−2 values are comparable with those for other antibodies such as those of IgA1 and IgA2, indicating that their general structural morphology in terms of structurally distinct Fab and Fc regions are similar. The anisotropy ratio RG/RO (where RO is the RG value of the sphere with the same volume as the hydrated glycoprotein) was calculated from the X-ray RG values of IgG4 to be 1.63–1.67. Comparison of this ratio with the value of 1.99 determined for human IgA1 and 1.66 for human IgA2 [26,28] showed that IgG4 is less elongated than IgA1, but is comparable in elongation with IgA2 in solution.

Guinier RG and RXS analyses for IgG4

Figure 4
Guinier RG and RXS analyses for IgG4

Results are shown using four different concentrations of IgG4 (from top to bottom, 1.3 mg/ml, 0.98 mg/ml, 0.65 mg/ml and 0.3 mg/ml). The filled circles between the arrowed data points represent the Q·RG and Q·RXS ranges used to determine the RG and RXS values. The Q range used for RG values is 0.7–1.4 nm−1 and those for RXS−1 and RXS−2 are 0.8–1.3 nm−1 and 0.9–1.4 nm−1 respectively.

Figure 4
Guinier RG and RXS analyses for IgG4

Results are shown using four different concentrations of IgG4 (from top to bottom, 1.3 mg/ml, 0.98 mg/ml, 0.65 mg/ml and 0.3 mg/ml). The filled circles between the arrowed data points represent the Q·RG and Q·RXS ranges used to determine the RG and RXS values. The Q range used for RG values is 0.7–1.4 nm−1 and those for RXS−1 and RXS−2 are 0.8–1.3 nm−1 and 0.9–1.4 nm−1 respectively.

The distance distribution function P(r) curve provides structural information in real space. From the P(r) analysis, the RG values were determined to range from 4.96 nm to 5.08 nm at the four concentrations, with a mean of 5.03 nm. These are within the error of the mean RG value of 4.99 nm from the Guinier analyses and are therefore consistent with these (Table 1). The maximum length L of IgG4 was determined from the value of r when the P(r) curve intersects P(r) = 0 to be 17.0 nm at all four concentrations (Table 1). The similarity of the s20,wRG and L values showed that the overall shape of IgG4 remains unchanged as the concentration is varied. The peaks in the P(r) curves correspond to the most frequently occurring interatomic distances within the structure. For IgG4, a single peak M1 was obtained at an r value of 5.3 nm at 0.3 mg/ml; as the IgG4 concentration increases, a second peak M2 becomes visible (Figure 5a). The two peaks show reproducible concentration dependences (Figure 5b). In the cases of other human antibodies, only a single peak M1 was experimentally observed in the X-ray P(r) curve for human IgD, whereas two peaks M1 and M2 were observed for IgA1 and IgA2 [26,28,29]. The previous explanation for two P(r) peaks for IgA1 and IgA2 was based on postulating relatively rigid hinge structures in IgA1 and IgA2 and their predominantly T-shaped structures. The single P(r) peak for IgD was attributed to either the existence of a different conformation in solution or the presence of sufficient hinge flexibility to result in multiple Fab and Fc arrangements, either of which would have caused peak M2 to disappear. These explanations are corroborated by scattering curve simulations based on the static crystal structures of intact murine IgG1 and IgG2a (PDB codes 1IGY and 1IGT [11,12]). The different Y-shaped or T-shaped Fab arrangements within IgG lead either to a single or double peak in their P(r) curves respectively [28]. For IgG4, it is concluded that the movement of the M1 and M2 peaks with concentration reflect alterations in the Fab and Fc arrangement within its structure that do not affect its overall shape.

Distance-distribution analyses P(r) for IgG4

Figure 5
Distance-distribution analyses P(r) for IgG4

(a) The concentration dependence of the P(r) curves is shown for 1.3 mg/ml, 0.98 mg/ml, 0.65 mg/ml and 0.30 mg/ml, from top to bottom. The peak maxima at M1 and M2 are indicated by arrows. The maximum length of IgG4 at 17 nm is denoted by L. (b) The concentration dependence of peaks M1 (circles) and M2 (triangles) is shown for repeated measurements at each of the four concentrations studied.

Figure 5
Distance-distribution analyses P(r) for IgG4

(a) The concentration dependence of the P(r) curves is shown for 1.3 mg/ml, 0.98 mg/ml, 0.65 mg/ml and 0.30 mg/ml, from top to bottom. The peak maxima at M1 and M2 are indicated by arrows. The maximum length of IgG4 at 17 nm is denoted by L. (b) The concentration dependence of peaks M1 (circles) and M2 (triangles) is shown for repeated measurements at each of the four concentrations studied.

The similarity of the RG values when set against the altered M1 and M2 values for IgG4 upon dilution implies that the scattering curve at 0.3 mg/ml is altered. Further inspection of the RG, RXS−1 and RXS−2 values substantiated this. The single M1 peak at 0.3 mg/ml was correlated with a slightly lower RG value of 0.1 nm, an almost unchanged RXS−1 value and a lower RXS−2 value of approx. 0.2 nm when compared with the other three concentrations (Table 1).

Molecular modelling of IgG4 and its hinge region

In order to determine IgG4 solution structures that would account for the movement of the two peaks in the P(r) curves, a starting model of IgG4 was constructed from the crystal structure of Fab B72.3 that corresponded to the mouse–human chimaera in use [33] and a homology model for the Fc region that was constructed from the human IgG1 crystal structure [11]. The sequence identity between the IgG1 and IgG4 Fc regions was 94.1% (Supplementary Figures S1f and S1g). No residue insertions or deletions were required in the modelling, and the main template β-strand structures and locations of the highly conserved cysteine residues were retained. A Ramachandran plot of the homology model using the structure validation program PROCHECK showed that all of the residues fell within sterically acceptable values. There is a single oligosaccharide chain at Asn314 of IgG1 that is conserved at Asn314 in IgG4. The two complex-type oligosaccharide structures seen in the human IgG1 crystal structure with nine residues in each were retained unmodified in the IgG4 models. Both Asn314 residues are located at the top of the CH2 domain in the Fc region, and the oligosaccharide chains point into the space between the two CH2 domains.

The IgG4 hinge was modelled after examination of the IgG1 crystal structure. Although the Fab region formally terminates at Ser234 (Figure 1 and Supplementary Figure S1d), the end of the terminal β-strand G in this immunoglobulin domain is Arg231, after which there are no steric restraints on the tripeptide V232ES234. The full hinge itself is formally defined by the residues E233SKYGPPCPSCPAPEFLGGP255 [2], within which the Fc region formally starts at Leu252 (Figure 1 and Supplementary Figure S1). Overall, this hinge sequence is characterized by six proline residues and two Cys243 and Cys246 disulfide bridges (using the numbering in Supplementary Figure S1). The latter are expected to tether the two hinge sequences to make these less flexible, especially between Cys243 and the Fc region. If the full hinge was conformationally randomized in the modelling, this would result in the separation of the two disulfide bridges. Accordingly, the middle and lower hinge residues C243PSCPAPEFLGGP255 were considered to be invariant and not modelled. It was concluded that IgG4 hinge variability would be best modelled by conformational sampling of a nine-residue peptide comprising the upper hinge E233SKYGPP242 and the two terminal residues Val232 and Cys243 that were used for connections with the Fab and Fc regions (Supplementary Figure S1). Because this sequence V232ESKYGPPC243 is located asymmetrically in relation to the Fc structure, this meant that the IgG4 model does not have 2-fold symmetry.

Randomized conformational search for IgG4 solution structures

A total of 10000 IgG4 conformationally randomized structures was assembled to model the IgG4 solution structure. Molecular dynamics resulted in a sufficiently wide range of orientations and translations of the two Fab and Fc fragments in IgG4 models for evaluation, while not necessarily satisfying the allowed regions in Ramachandran plots (see the Experimental section). In these, no account was taken of the two isomeric disulfide forms of IgG4 because these showed similar sedimentation properties (Figure 3) to indicate that they have similar solution structures. The simulation with 5000 hinges of length 3.1–3.15 nm generated 5000 trial IgG4 structures with no 2-fold axial symmetry about the Fc region. The resulting modelled scattering curves were compared with the experimental X-ray curves to give goodness-of-fit R-factors ranging between 2.0% and 27.0% (Table 1). When the R-factors were compared with their RG values, the lowest R-factors corresponded to models with RG values close to the experimental RG values [Supplementary Figure S2 (upper row) at http://www.biochemj.org/bj/432/bj4320101add.htm]. The identification of a single minimum R-factor value in Supplementary Figure S2 showed that the modelling strategy was successful in generating a single family of possible solution structures for IgG4. A second simulation with 5000 different IgG4 models with hinges of minimum length 2.2–3.1 nm was also performed. The distributions in Supplementary Figure S2 (lower row) showed that fewer models with larger RG values were obtained, but nonetheless a similar minimum in R-factor values was obtained in all four cases. This indicated the reproducibility of the modelling, and also showed that extended hinges were required for good fits.

Filters based on the experimental scattering data were used with the 10000 IgG4 models to reject unsatisfactory models and identify best-fit models.

(i) The ±5% filter for steric overlap eliminated those models in which the Fab and Fc fragments and the hinge sterically overlapped as the result of inappropriate randomized hinge conformations generated by the simulations. A satisfactory IgG4 sphere model needed a minimum total number N of 1566 spheres in order to correspond to the composition-calculated volume of IgG4.

(ii) A ±5% filter of the RG values calculated from the models in the same Guinier Q ranges used for the four experimental X-ray RG values of 4.92–5.03 nm identified those IgG4 models that fitted the degree of elongation determined by X-rays.

(iii) When the models were ordered on the basis of their lowest R-factors, this resulted in the identification of ten best-fit models for each of the four IgG4 concentrations. These best-fit models are found as a single cluster (green in Supplementary Figure S2), suggesting that a single best-fit family of structures has been identified. The R-factor values were low at 2% compared with other modelling fits [24]. Both the RG and RXS−1 values for these ten best-fit models are within acceptable error of the experimental values (Table 1). In particular, the model for 0.3 mg/ml showed a slightly lower RXS−1 value and a more reduced RXS−2 value when compared with those for the other three concentrations (Table 1).

The outcome of the modelling analyses in terms of best-fit models is summarized in Figure 6. Good curve fits were obtained in all four cases, both for the experimental I(Q) curves and the P(r) curves shown as insets. The four models in Figure 6 display a range of Fab conformations compared with the Fc region. The models do not indicate a simple explanation for the concentration dependence of the two peaks in the P(r) curves. The previous modelling of IgA1, IgA2 and IgD showed that similar T-shaped or Y-shaped structures resulted in P(r) curves with either one or two peaks. Accordingly, the distribution of the distances d1, d2 and d3 between the centres of the Fab and Fc regions in the 100 best-fit IgG4 models was investigated (see Supplementary Figure S3 at http://www.biochemj.org/bj/432/bj4320101add.htm). Comparisons of d1, d2 and d3 showed that only a limited range of distances fitted the data, mostly in the upper-right corners in Supplementary Figures S3(a)–S3(c). This showed that only the more extended hinges in IgG4 fitted the X-ray data best.

X-ray scattering curve fits for the best-fit IgG4 models at four concentrations

Figure 6
X-ray scattering curve fits for the best-fit IgG4 models at four concentrations

The IgG concentrations are (a) 1.3 mg/ml (pink), (b) 0.98 mg/ml (green), (c) 0.65 mg/ml (orange) and (d) 0.30 mg/ml (blue). In the left-hand panels, the experimental data are indicated by circles and the calculated best-fit scattering curve is indicated by a continuous line. The insets correspond to the experimental (black) and best-fit modelled (coloured) P(r) curves, in which the M1 and M2 values from Figure 5 are indicated by arrows. In the right-hand panels, the best-fit IgG4 models are shown, with the Fab regions in blue and the Fc regions in red.

Figure 6
X-ray scattering curve fits for the best-fit IgG4 models at four concentrations

The IgG concentrations are (a) 1.3 mg/ml (pink), (b) 0.98 mg/ml (green), (c) 0.65 mg/ml (orange) and (d) 0.30 mg/ml (blue). In the left-hand panels, the experimental data are indicated by circles and the calculated best-fit scattering curve is indicated by a continuous line. The insets correspond to the experimental (black) and best-fit modelled (coloured) P(r) curves, in which the M1 and M2 values from Figure 5 are indicated by arrows. In the right-hand panels, the best-fit IgG4 models are shown, with the Fab regions in blue and the Fc regions in red.

Even though much scatter is present in Supplementary Figure S3, the distribution of d1 versus d3 distances explained the observed P(r) differences with concentration. The ten best-fit models at 0.3 mg/ml (dark blue) cluster around d3 = 6 nm and d1 = 7 nm. The three higher concentrations show two clusters of best-fit structures at either d3=8 nm and d1=5 nm, or at d3=5 nm and d1=8 nm. Because the two Fab fragments are equivalent in the IgG4 structure, the two clusters correspond, in fact, to the same IgG4 structure, and show that the d3 distances correspond to a more extended structure at higher concentration compared with that at 0.3 mg/ml. The ten best-fit structures at 1.3 mg/ml and 0.3 mg/ml were examined after their Fc regions were superimposed (PDB codes 3M7X and 3M80) with a viewer such as RasMol (http://www.openrasmol.org/). Although the Fab positions vary, the Fab regions appear to be more separated at 1.3 mg/ml than at 0.3 mg/ml, which corresponds to the increase of d3 to 8 nm. The occurrence of two clusters is best explained by the asymmetry of the Fab and Fc regions within the IgG4 structure at higher concentration, much as is seen in the crystal structure of human IgG1. One Fab region cannot approach the near side of the Fc region unless the other Fab region is moved away. This best-fit solution structure for IgG4 is, however, notably different from that for the IgG1 crystal structure, for which the d1, d2 and d3 distances are denoted by a black X symbol in Supplementary Figure S3.

Sedimentation coefficient modelling for human IgG4

The sedimentation coefficient s20,w value of the ten best-fit X-ray scattering IgG4 models were calculated for comparison with the experimental values of 6.44 S (Figure 3). The calculated s20,w values for the ten models for each of the four concentrations ranged from 6.61 S to 6.77 S using HYDRO (Table 1). The corresponding four sets of s20,w values using HYDROPRO ranged from 6.39 S to 6.60 S (Table 1). Given that the calculations are generally accurate to within±0.21 S [24], both sets of s20,w values agree well with the experimental value of 6.44 S. In previous applications of HYDRO and HYDROPRO to antibody structures, large discrepancies had occasionally been observed with the use of HYDRO, but not with HYDROPRO [49]. This discrepancy was not observed here, showing that the independent modelling of each of the X-ray and s20,w data sets have resulted in a self-consistent structural outcome.

DISCUSSION

IgG4 structure and function

Our IgG4 analyses resulted in the determination of an asymmetric antibody conformation in solution that revealed new insight into function (Figure 7). The two Fab regions partially mask the Fc region in IgG4. This structure has implications for (i) C1q binding to IgG4, which is important for complement activation, (ii) the self-association of IgG4, which may be relevant for rheumatoid factor-like IgG self-binding, and (iii) the formation of bispecific IgG4 molecules.

Views of the best-fit IgG4 models at four concentrations

Figure 7
Views of the best-fit IgG4 models at four concentrations

The IgG4 concentrations correspond to (a) 1.3 mg/ml, (b) 0.98 mg/ml and 0.65 mg/ml, and (c) 0.3 mg/ml. Two representative views of the best-fit model at each concentration are shown in order to highlight the Fab–Fab and Fab–Fc separations at different concentrations. The distances d1 between the centres-of-mass of the Fab regions (yellow spheres) are marked as shown, together with the distance d3 between one Fab region and the Fc region. In (d,e), the ten best-fit IgG4 structures at 1.3 mg/ml and 0.3 mg/ml are shown superimposed upon their Fc region (dark blue; top right). In (f), the crystal structure of human IgG1 is shown with its Fc region also viewed edge-on in a similar orientation. The heterotrimeric C1q globular head is depicted in terms of its A, B and C chains in three colours adjacent to the Fc region of IgG1 for size comparison. The positions of Glu335, Lys337 and Lys339 on each of the IgG1 heavy chains are highlighted as shown (red, orange and orange respectively).

Figure 7
Views of the best-fit IgG4 models at four concentrations

The IgG4 concentrations correspond to (a) 1.3 mg/ml, (b) 0.98 mg/ml and 0.65 mg/ml, and (c) 0.3 mg/ml. Two representative views of the best-fit model at each concentration are shown in order to highlight the Fab–Fab and Fab–Fc separations at different concentrations. The distances d1 between the centres-of-mass of the Fab regions (yellow spheres) are marked as shown, together with the distance d3 between one Fab region and the Fc region. In (d,e), the ten best-fit IgG4 structures at 1.3 mg/ml and 0.3 mg/ml are shown superimposed upon their Fc region (dark blue; top right). In (f), the crystal structure of human IgG1 is shown with its Fc region also viewed edge-on in a similar orientation. The heterotrimeric C1q globular head is depicted in terms of its A, B and C chains in three colours adjacent to the Fc region of IgG1 for size comparison. The positions of Glu335, Lys337 and Lys339 on each of the IgG1 heavy chains are highlighted as shown (red, orange and orange respectively).

C1q binding to IgG activates the classical pathway of complement. Unlike IgG1, which possesses three extra residues in the upper hinge, but is otherwise identical with IgG4, IgG4 does not activate complement through its C1q-binding site at the top of the CH2 domain in the Fc region. Glu335, Lys337 and Lys339 are the core residues for C1q binding (Figure 7f), and appear in both the IgG4 and IgG1 sequences (Supplementary Figure S1f) [52]. The difference in C1q binding between IgG4 and IgG1 is clarified by our models. For IgG4, molecular views of each of the four sets of ten superimposed best-fit IgG4 models (PDB codes 3M7X, 3M7Y, 3M7Z and 3M80) shows that the Fab regions are positioned at the top of the Fc region and along one Fc face, but leaves the other Fc face exposed (Figures 7d and 7e). Glu335, Lys337 and Lys339 are located above and below the Fc region as viewed (Figure 7f). It is not possible for the C1q head to interact with the Fc region in IgG4 without encountering steric clash with the Fab regions that are close to the Fc region. This analysis was tested using solvent-accessible surface calculations by the Lee–Richards method [53]. The use of a sphere of radius 0.14 nm to represent a water molecule showed that Glu335, Lys337 and Lys339 were 80–90%, 60% and 60% solvent-accessible respectively in the four best-fit IgG4 models (Table 1). The use of spheres with five radii 0.5–4.0 nm to represent the C1q head of approximate radius 3 nm (Figure 7f) showed that the surface accessibility of these three residues within IgG4 was reduced to 0% even with a sphere of radius 1 nm. Comparative accessibility calculations based on Fc regions only showed that the Fab regions contribute to this reduction, particularly at high concentrations. For IgG1, the single view of the IgG1 crystal structure (PDB code 1HZH) shows that both Fab regions are positioned above and below the Fc region (Figure 7f), where they now offer no hindrance of C1q access to the Fc region [54].

The self-binding of human IgG1 and IgG4 is relevant in the context of auto-antibodies associated with rheumatoid arthritis. In the present study, intact IgG4 at 2–11 μM in solution does not interact with itself. Interestingly, when IgG4 is immobilized on a solid support, this self-associates with IgG4 in the solution phase [51]. This self-association presumably occurs because IgG4 immobilization makes the Fc region more accessible because the Fab regions have become partially displaced. If IgG4 Fc is immobilized on a sensor surface, a small amount of Fc self-association is detected with a dissociation constant Kd of 1 μM [51]. This self-association is inferred to occur because the Fab regions are not present to mask the immobilized Fc regions. These comparisons support the view that the Fc regions in IgG4 at 2–11 μM in solution are masked by the Fab regions.

In addition, even though the overall IgG4 shape is unaffected by concentration, the observed alterations in both Fab–Fc hinge lengths and Fab–Fab separations within IgG4 with concentration may be related to the observation of bispecificity in IgG4. The ready exchange between monospecific and bispecific forms of IgG4 in the reducing environment of red blood cells may be facilitated by molecular crowding at the hinge region in IgG4 caused by the proximity between the Fab and Fc regions [55]. This exchange process also involves the CH3 domains, which also make inter-heavy-chain contacts [55]. When hinge crowding is supplemented by the reduction of disulfide bridges between the two IgG4 heavy chains, steric crowding at the hinge would facilitate heavy-chain exchange between different IgG4 molecules. The transient formation of half-molecules of IgG4 leads to the observed inability of IgG4 antibody to cross-link identical antigens through bivalent binding, as well as the existence of bispecific IgG4 molecules.

Hinge structure of IgG4

The X-ray modelling showed that the IgG4 hinge possesses a mostly extended structure in solution, and that the Fab and Fc regions show modest but detectable alterations with concentration. This asymmetric solution structure for IgG4 is attributed to its hinge sequence. Molecular views show that the hinge is not sufficiently long to permit the three Fab and Fc regions in IgG4 to exhibit unrestricted freedom of conformational rearrangement about the hinge. The upper hinge in IgG4 is only seven residues long, compared with the ten residues in the upper hinge of IgG1, and IgG4 possesses the shortest middle hinge of the four human IgG subclasses [2]. It is not surprising that the IgG4 upper hinge is structurally extended in order to be able to accommodate its Fab and Fc regions. The length of the upper hinge was defined by the α-carbon positions of the flanking residues Arg231 and Pro244. The maximum possible length for ten peptide bonds in an extended conformation is 3.8 nm. The ten bestfit structures from the four IgG4 concentrations gave hinge lengths that increased from 2.8 nm to 3.6 nm with dilution (Table 2). This increase is accompanied by a decreased separation between the Fab regions and an increased separation between the Fab–Fc regions with dilution (Figure 7). In comparison, the corresponding distances between the α-carbons of Lys222 and Pro240 (PDB code 1HZH numbering) in the IgG1 upper hinge are 2.6 nm and 2.7 nm. The shorter length of the IgG1 upper hinge may be the consequence of its extra three residues that facilitate greater conformational possibilities between the Fab and Fc regions (Supplementary Figure S1e).

Table 2
Summary of each set of ten IgG4 best-fit models

Results shown are means±S.D.

 Mean upper hinge length (nm) Fab–Fab angle (°) First Fab–Fc angle (°) Second Fab–Fc angle (°) 
IgG4 (1.30 mg/ml) 2.7± 0.2 104± 43 139± 20 90± 32 
IgG4 (0.98 mg/ml) 3.5± 0.3 96± 47 136± 21 71± 63 
IgG4 (0.65 mg/ml) 3.1± 0.5 76± 39 120± 18 74± 64 
IgG4 (0.30 mg/ml) 3.6± 0.1 105± 43 137± 24 64± 55 
 Mean upper hinge length (nm) Fab–Fab angle (°) First Fab–Fc angle (°) Second Fab–Fc angle (°) 
IgG4 (1.30 mg/ml) 2.7± 0.2 104± 43 139± 20 90± 32 
IgG4 (0.98 mg/ml) 3.5± 0.3 96± 47 136± 21 71± 63 
IgG4 (0.65 mg/ml) 3.1± 0.5 76± 39 120± 18 74± 64 
IgG4 (0.30 mg/ml) 3.6± 0.1 105± 43 137± 24 64± 55 

The solution conformation of IgG4 can be compared with IgG4 structures previously reported by electron microscopy, using angles measured between the Fab and Fc regions. In our IgG4 models, the mean Fab–Fab angle ranged between 76 ° and 105 ° with S.D.s of up to 47 ° (Table 2). Two different sets of Fab–Fc angles resulted because of the asymmetric IgG4 solution models, and these ranged between 120–139 ° and 64–90 °. By electron microscopy, the mean Fab–Fab angle in IgG4 was 128 °±39 ° and the mean Fab–Fc angle was 98 °±25 °, and similar values were reported for IgG1, IgG2 and IgG3 [22]. A more recent study reported mean Fab–Fab angles of 96 °±26 ° and 99 °±17 ° for a murine MMA 383 antibody [56]. It is deduced that the electron microscopy angles are in good accord with those from the scattering modelling of IgG4.

The concentration dependence of the IgG4 structure corresponds to an internal structural rearrangement of the Fab and Fc regions. No change in overall shape is observed from the s20,w and RG values, and IgG4 dissociation into two halves is ruled out from the c(s) analyses. There are two possible explanations for this based on fast non-binding interactions. First, it is possible that greater diffusion–collision rates between IgG4 molecules at 1.3 mg/ml may alter the internal arrangement of the Fab and Fc regions within IgG4 compared with that at 0.3 mg/ml. At 1.3 mg/ml, the average separation between IgG4 molecules will be approx. 58 nm. At 0.3 mg/ml, this average separation will be reduced to approx. 94 nm. These separations are comparable with the 17 nm size of IgG4 molecules (Table 1). Alternatively, given the presence of 272 charged residues (21%) in a total of 1315 residues in IgG4, long-range electrostatic fields between IgG4 molecules at higher concentration may influence the arrangement of Fab and Fc regions within IgG4. Physiologically, the conformation at 1.3 mg/ml is more likely to be relevant, given that serum proteins exist in a crowded environment at 80 mg/ml in blood, and also during conditions of bioprocessing during which reagent concentrations of as high as 20 mg/ml will be encountered [5,6].

The present study represents advances in comparison with earlier similar studies. Previous X-ray scattering experiments at synchrotrons often employed a minimum concentration of 2 mg/ml [19,21]. In the present work, excellent signal-to-noise ratios were attainable even at 0.3 mg/ml (Table 1), this being attributed to improvements in detector technologies. X-ray scattering can be limited by aggregated material that may increase RG values by as much as 5–10% and perturb the modelling [18,20]. In the present study, the absence of aggregates was confirmed by the independent use of ultracentrifugation c(s) analyses and the good agreement between the X-ray and s20,w modelling. A limitation of earlier studies is the use of simple shape models to model the antibody structure [1821]. In the present study, the use of atomic starting models for IgG4 to result in families of best-fit structures resulted in more convincing molecular insight into the solution data.

Stability of therapeutic IgG4

In terms of the industrial bioprocessing of antibodies, the clarification of the solution conformation of IgG4 and the relative movement of the Fab and Fc domains within this may be important for its aggregation and surface-adhesion effects during its manufacture and storage as biopharmaceuticals. These alterations may modulate their susceptibility to interactions with other proteins or container surfaces. Our IgG4 results will enable us to examine the impact of low pH conditions on IgG4 during its elution from Protein A or Protein G resins during large-scale chromatography. This key step is frequently implicated in the loss of antibody to aggregation with nearly half of this class of medicine found to have such issues [57].

Abbreviations

     
  • ESRF

    European Synchrotron Radiation Facility

AUTHOR CONTRIBUTION

Yuki Abe designed and performed experiments, analysed data and wrote the manuscript; Jayesh Gor designed and performed experiments; and Daniel Bracewell, Stephen Perkins and Paul Dalby designed the study, analysed data and wrote the manuscript.

We thank Dr Jim Davies (Lonza Biologics) for generously providing human IgG4, and Dr Alexandra Bonner and Dr Azubuike I. Okemefuna for excellent computational support.

FUNDING

This work was supported by the Innovative Manufacturing Research Centre in Bioprocessing, University College London, London, U.K. Equipment was funded by the Biotechnology and Biological Sciences Research Council, the Henry Smith Charity and the Mercer Fund of the Fight For Sight Charity.

References

References
1
Hamilton
R. G.
The Human IgG Subclasses, Calbiochem Corporation http://www.antibodybeyond.com/books/Calbiochem_Human_IgG_Booklet_CB0051.pdf
2001
2
Brekke
O. H.
Michaelson
T. E.
Sandlie
I.
The structural requirements for complement activation by IgG: does it hinge on the hinge?
Immunol. Today
1995
, vol. 
16
 (pg. 
85
-
90
)
3
Aalberse
R. C.
Schuurman
J.
IgG4 breaking the rules
Immunology
2002
, vol. 
105
 (pg. 
9
-
19
)
4
Rispens
T.
den Bleker
T. H.
Aslberse
R. C.
Hybrid IgG4/IgG4 Fc antibodies form upon ‘Fab-arm’ exchange as demonstrated by SDS-PAGE or size-exclusion chromatography
Mol. Immunol.
2010
, vol. 
47
 (pg. 
1592
-
1594
)
5
Wang
W.
Singh
S.
Zeng
D. L.
King
K.
Nema
S.
Antibody structure, instability and formulation
J. Pharm. Sci.
2007
, vol. 
96
 (pg. 
1
-
26
)
6
Wang
X.
Das
T. K.
Singh
S. K.
Kumar
S.
Potential aggregation prone regions in biotherapeutics: a survey of commercial monoclonal antibodies
MAbs
2009
, vol. 
1
 (pg. 
254
-
267
)
7
Bonner
A.
Furtado
P. B.
Almogren
A.
Kerr
M. A.
Perkins
S. J.
Implications of the near-planar solution structure of human myeloma dimeric IgA1 for mucosal immunity and IgA nephropathy
J. Immunol.
2008
, vol. 
180
 (pg. 
1008
-
1018
)
8
Bonner
A.
Almogren
A.
Furtado
P. B.
Kerr
M. A.
Perkins
S. J.
Location of secretory component on the Fc edge of dimeric IgA1 reveals insight into the role of secretory IgA1 in mucosal immunity
Mucosal Immunol.
2009
, vol. 
2
 (pg. 
74
-
84
)
9
Saphire
E. O.
Parren
P. W.
Pantophlet
R.
Zwick
M. B.
Morris
G. M.
Rudd
P. M.
Dwek
R. A.
Stanfield
R. L.
Burton
D. R.
Wilson
I. A.
Crystal structure of a neutralizing human IgG against HIV-1: a template for vaccine design
Science
2001
, vol. 
293
 (pg. 
1155
-
1159
)
10
Saphire
E. O.
Stanfield
R. L.
Crispin
M. D. M.
Parren
P. W. H. I.
Rudd
P. M.
Dwek
R. A.
Burton
D. R.
Wilson
I. A.
Contrasting IgG structures reveal extreme asymmetry and flexibility
J. Mol. Biol.
2002
, vol. 
319
 (pg. 
9
-
18
)
11
Harris
L. J.
Larson
S. B.
Hasel
K. W.
McPherson
A.
Refined structure of an intact IgG2a monoclonal antibody
Biochemistry
1997
, vol. 
36
 (pg. 
1581
-
1597
)
12
Harris
L. J.
Skaletsky
E.
McPherson
A.
Crystallographic structure of an intact IgG1 monoclonal antibody
J. Mol. Biol.
1998
, vol. 
275
 (pg. 
861
-
872
)
13
Cromwell
M. E. M.
Hilario
E.
Jacobson
F.
Protein aggregation and bioprocessing
AAPS J.
2006
, vol. 
8
 (pg. 
E572
-
E579
)
14
Sukumar
M.
Doyle
B. L.
Combs
J. L.
Pekar
A. H.
Opalescent appearance of an IgG1 antibody at high concentrations and its relationship to noncovalent association
Pharm. Res.
2004
, vol. 
21
 (pg. 
1087
-
1093
)
15
Manning
M. C.
Chou
D. K.
Murphy
B. M.
Payne
R. W.
Katayama
D. S.
Stability of protein pharmaceuticals: an update
Pharm. Res.
2010
, vol. 
27
 (pg. 
544
-
575
)
16
Biddlecombe
J. G.
Smith
G.
Uddin
S.
Mulot
S.
Spencer
D.
Gee
C.
Fish
B. C.
Bracewell
D. G.
Factors influencing antibody stability at solid-liquid interfaces in a high shear environment
Biotechnol. Prog.
2009
, vol. 
25
 (pg. 
1499
-
1507
)
17
Mayans
M. O.
Coadwell
W. J.
Beale
D.
Symons
D.
Perkins
S. J.
Demonstration by pulsed neutron scattering that the arrangement of the Fab and Fc fragments in the overall structures of bovine IgG1 and IgG2 in solution is similar
Biochem. J.
1995
, vol. 
311
 (pg. 
283
-
291
)
18
Gregory
L.
Davis
K. G.
Sheth
B.
Boyd
J.
Jefferis
R.
Nave
C.
Burton
D. R.
The solution conformations of the subclasses of human IgG deduced from sedimentation and small angle X-ray scattering studies
Mol. Immunol.
1987
, vol. 
24
 (pg. 
821
-
829
)
19
Longman
E.
Kreusel
K.
Tendler
S. B.
Fiebrig
I.
King
K.
Adair
J.
O'Shea
P.
Ortega
A.
Garcia de la Torre
J.
Harding
S. E.
Estimating domain orientation of two human antibody IgG4 chimeras by crystallohydrodynamics
Eur. Biophys. J.
2003
, vol. 
32
 (pg. 
503
-
510
)
20
Lu
Y.
Harding
S. E.
Michaelsen
T. E.
Longman
E.
Davis
K. G.
Ortega
A.
Grossmann
J. G.
Sandlie
I.
Garcia de la Torre
J.
Solution conformation of wild-type and mutant IgG3 and IgG4 immunoglobulins using crystallohydrodynamics: possible implications for complement activation
Biophys. J.
2007
, vol. 
93
 (pg. 
3733
-
3744
)
21
Lu
Y.
Harding
S. E.
Rowe
A. J.
Davis
K. G.
Fish
B.
Varley
P.
Gee
C.
Mulot
S.
The effect of a point mutation on the stability of IgG4 as monitored by analytical ultracentrifugation
J. Pharm. Sci.
2008
, vol. 
97
 (pg. 
960
-
969
)
22
Roux
K. H.
Strelets
L.
Michaelsen
T. E.
Flexibility of human IgG subclasses
J. Immunol.
1997
, vol. 
159
 (pg. 
3372
-
3382
)
23
Perkins
S. J.
Bonner
A.
Structure determinations of human and chimaeric antibodies by solution scattering and constrained molecular modelling
Biochem. Soc. Transact.
2008
, vol. 
36
 (pg. 
37
-
42
)
24
Perkins
S. J.
Okemefuna
A. I.
Nan
R.
Li
K.
Bonner
A.
Constrained solution scattering modelling of human antibodies and complement proteins reveals novel biological insights
J. R. Soc. Interface
2009
, vol. 
6
 (pg. 
S679
-
S696
)
25
Perkins
S. J.
Nealis
A. S.
Sutton
B. J.
Feinstein
A.
The solution structure of human and mouse immunoglobulin IgM by synchrotron X-ray scattering and molecular graphics modelling: a possible mechanism for complement activation
J. Mol. Biol.
1991
, vol. 
221
 (pg. 
1345
-
1366
)
26
Boehm
M. K.
Woof
J. M.
Kerr
M. A.
Perkins
S. J.
The Fab and Fc fragments of IgA1 exhibit a different arrangement from that in IgG: a study by X-ray and neutron solution scattering and homology modelling
J. Mol. Biol.
1999
, vol. 
286
 (pg. 
1421
-
1447
)
27
Aslam
M.
Guthridge
J. M.
Hack
B. K.
Quigg
R. J.
Holers
V. M.
Perkins
S. J.
The extended multidomain solution structures of the complement protein Crry and its chimeric conjugate Crry-Ig by scattering, analytical ultracentrifugation and constrained modelling: implications for function and therapy
J. Mol. Biol.
2003
, vol. 
329
 (pg. 
525
-
550
)
28
Furtado
P. B.
Whitty
P. W.
Robertson
A.
Eaton
J. T.
Almogren
A.
Kerr
M. A.
Woof
J. M.
Perkins
S. J.
Solution structure determination of monomeric human IgA2 by X-ray and neutron scattering, analytical ultracentrifugation and constrained modelling: a comparison with monomeric human IgA1
J. Mol. Biol.
2004
, vol. 
338
 (pg. 
921
-
941
)
29
Sun
Z.
Almogren
A.
Furtado
P. B.
Chowdhury
B.
Kerr
M. A.
Perkins
S. J.
Semi-extended solution structure of human myeloma immunoglobulin D determined by constrained X-ray scattering
J. Mol. Biol.
2005
, vol. 
353
 (pg. 
155
-
173
)
30
Perkins
S. J.
Okemefuna
A. I.
Fernando
A. N.
Bonner
A.
Gilbert
H. E.
Furtado
P. B.
X-ray and neutron scattering data and their constrained molecular modelling
Methods Cell Biol.
2008
, vol. 
84
 (pg. 
375
-
423
)
31
Okemefuna
A. I.
Nan
R.
Gor
J.
Perkins
S. J.
Electrostatic interactions contribute to the folded-back conformation of wild-type human Factor H
J. Mol. Biol.
2009
, vol. 
391
 (pg. 
98
-
118
)
32
Cole
J. L.
Lary
J. W.
Moody
T. P.
Laue
T. M.
Analytical ultracentrifugation: sedimentation velocity and sedimentation equilibrium
Methods Cell Biol.
2008
, vol. 
84
 (pg. 
143
-
211
)
33
Brady
R. L.
Hubbard
R. E.
King
D. J.
Low
D. C.
Roberts
S. M.
Todd
R. J.
Crystallization and preliminary X-ray diffraction study of a chimaeric Fab' fragment of antibody binding tumour cells
J. Mol. Biol.
1991
, vol. 
219
 (pg. 
603
-
604
)
34
Perkins
S. J.
Protein volumes and hydration effects: the calculation of partial specific volumes, neutron scattering matchpoints and 280 nm absorption coefficients for proteins and glycoproteins from amino acid sequences
Eur. J. Biochem.
1986
, vol. 
157
 (pg. 
169
-
180
)
35
Schuck
P.
Sedimentation analysis of non-interacting and self-associating solutes using numerical solutions to the Lamm equation
Biophys. J.
1998
, vol. 
75
 (pg. 
1503
-
1512
)
36
Schuck
P.
Size-distribution analysis of macromolecules by sedimentation velocity ultracentrifugation and Lamm equation modeling
Biophys. J.
2000
, vol. 
78
 (pg. 
1606
-
1619
)
37
Narayanan
T.
Diat
O.
Bosecke
P.
SAXS and USAXS on the high brilliance beamline at the ESRF
Nucl. Instrum. Methods Phys. Res. A
2001
, vol. 
467–468
 (pg. 
1005
-
1009
)
38
Li
K.
Okemefuna
A. I.
Gor
J.
Hannan
J. P.
Asokan
R.
Holers
V. M.
Perkins
S. J.
Solution structure of the complex formed between human complement C3d and full length complement receptor Type 2
J. Mol. Biol.
2008
, vol. 
384
 (pg. 
137
-
150
)
39
Gilbert
H. E.
Eaton
J. T.
Hannan
J. P.
Holers
V. M.
Perkins
S. J.
Solution structure of the complex between CR2 SCR 1–2 and C3d of human complement: an X-ray scattering and sedimentation modelling study
J. Mol. Biol.
2005
, vol. 
346
 (pg. 
859
-
873
)
40
Glatter
O.
Kratky
O.
Small Angle X-ray Scattering
1982
New York
Academic Press
41
Pilz
I.
Kratky
O.
Licht
A.
Sela
M.
Shape and volume of anti-poly(D-alanyl) antibodies in the presence and absence of tetra-D-alanine as followed by small-angle X-ray scattering
Biochemistry
1973
, vol. 
12
 (pg. 
4998
-
5005
)
42
Semenyuk
A. V.
Svergun
D. I.
GNOM – a program package for small-angle scattering data-processing
J. Appl. Crystallogr.
1991
, vol. 
24
 (pg. 
537
-
540
)
43
Arnold
K.
Bordoli
L.
Kopp
J.
Schwede
T.
The SWISS-MODEL Workspace: a web-based environment for protein structure homology modelling
Bioinformatics
2006
, vol. 
22
 (pg. 
195
-
201
)
44
Perkins
S. J.
Weiss
H.
Low resolution structural studies of mitochondrial ubiquinol:cytochrome c reductase in detergent solutions by neutron scattering
J. Mol. Biol.
1983
, vol. 
168
 (pg. 
847
-
866
)
45
Ashton
A. W.
Boehm
M. K.
Gallimore
J. R.
Pepys
M. B.
Perkins
S. J.
Pentameric and decameric structures in solution of the serum amyloid P component by X-ray and neutron scattering and molecular modelling analyses
J. Mol. Biol.
1997
, vol. 
272
 (pg. 
408
-
422
)
46
Perkins
S. J.
X-ray and neutron scattering analyses of hydration shells: a molecular interpretation based on sequence predictions and modelling fits
Biophys. Chem.
2001
, vol. 
93
 (pg. 
129
-
139
)
47
Garcia de la Torre
J.
Navarro
S.
Martinez
M. C. L.
Diaz
F. G.
Cascales
J. L.
HYDRO: a computer program for the prediction of hydrodynamic properties of macromolecules
Biophys. J.
1994
, vol. 
67
 (pg. 
530
-
531
)
48
Garcia de la Torre
J.
Huertas
M. L.
Carrasco
B.
Calculation of hydrodynamic properties of globular proteins from their atomic-level structure
Biophys. J.
2000
, vol. 
78
 (pg. 
719
-
730
)
49
Almogren
A.
Furtado
P. B.
Sun
Z.
Perkins
S. J.
Kerr
M. A.
Purification, properties and extended solution structure of the complex formed between human immunoglobulin A1 and human serum albumin by scattering and ultracentrifugation
J. Mol. Biol.
2006
, vol. 
356
 (pg. 
413
-
431
)
50
Bloom
J. W.
Madanat
M. S.
Marriott
D.
Wong
T.
Chan
S.-Y.
Intrachain disulfide bond in the core hinge region of human IgG4
Protein Sci.
1997
, vol. 
6
 (pg. 
407
-
415
)
51
Rispens
T.
Heer
P. O.
Vermeulen
E.
Schuurman
J
Van Der Neut Kolfschoten
M.
Aalberse
R. C.
Human IgG4 binds to IgG4 and conformationally altered IgG1 via Fc-Fc interactions
J. Immunol.
2009
, vol. 
182
 (pg. 
4275
-
4281
)
52
Harris
L. J.
Larson
S. B.
McPherson
A.
Comparison of intact antibody structures and the implications for effector function
Adv. Immunol.
1999
, vol. 
72
 (pg. 
191
-
208
)
53
Lee
B.
Richards
F. M.
The interpretation of protein structures: estimation of static accessibility
J. Mol. Biol.
1971
, vol. 
55
 (pg. 
379
-
400
)
54
Gaboriaud
C.
Juanhuix
J.
Gruez
A.
Lacroix
M.
Darnault
C.
Pignol
D.
Verger
D
Fontecilla-Camps
J. C.
Arlaud
G. J.
The crystal structure of the globular head of complement protein C1q provides a basis for its versatile recognition properties
J. Biol. Chem.
2003
, vol. 
278
 (pg. 
46974
-
46982
)
55
Van Der Neut Kolfschoten
M.
Schuurman
J.
Losen
M.
Bleeker
W. K.
MartínezMartínez
P.
Vermeulen
E.
den Bleker
T. H.
Wiegman
L.
Vink
T.
Aarden
L. A.
, et al. 
Anti-inflammatory activity of human IgG4 antibodies by dynamic Fab arm exchange
Science
2007
, vol. 
317
 (pg. 
1554
-
1557
)
56
Taschner
N.
Müller
S. A.
Alumella
V. R.
Goldie
K. N.
Drake
A. F.
Aebi
U.
Arvinte
T.
Modulation of antigenicity related to changes in antibody flexibility upon lyophilization
J. Mol. Biol.
2001
, vol. 
310
 (pg. 
169
-
179
)
57
Shukla
A. A.
Hubbard
B.
Tressel
T.
Guhan
S.
Low
D.
Downstream processing of monoclonal antibodies: application of platform approaches
J. Chromatogr. B
2007
, vol. 
848
 (pg. 
28
-
39
)

Author notes

The structural co-ordinates reported will appear in the PDB (Protein Data Bank) under accession codes 3M7X, 3M7Y, 3M7Z and 3M80.

Supplementary data