IgA is unique in being able to form a diverse range of polymeric structures. Increases in the levels of dimeric IgA1 (dIgA1) in serum have been implicated in diseases such as IgA nephropathy. We have determined the solution structure for dIgA1 by synchrotron x-ray and neutron scattering and analytical ultracentrifugation. The Guinier radius of gyration (RG) of 7.60–8.65 nm indicated that the two monomers within dIgA1 are arranged in an extended conformation. The distance distribution curve P(r) gave an overall length (L) of 22–26 nm. These results were confirmed by the sedimentation coefficient and frictional ratio of dIgA1. Constrained scattering modeling starting from the IgA1 monomer solution structure revealed a near-planar dimer structure for dIgA1. The two Fc regions form a slightly bent arrangement in which they form end-to-end contacts, and the J chain was located at this interface. This structure was refined by optimizing the position of the four Fab regions. From this, the best-fit solution structures show that the four Fab Ag-binding sites are independent of one another, and the two Fc regions are accessible to receptor binding. This arrangement allows dIgA1 to initiate specific immune responses by binding to FcαRI receptors, while still retaining Ag-binding ability, and to be selectively transported to mucosal surfaces by binding to the polymeric Ig receptor to form secretory IgA. A mechanism for the involvement of dIgA1 oligomers in the pathology of IgA nephropathy is discussed in the light of this near-planar structure.

Immunoglobulin A is the most abundant Ab class found in mucosal surfaces and is also the most heterogeneous human Ab class, being present in two subclasses IgA1 and IgA2 of which there are at least two different allotypes. This is reflected in the diverse functional properties of IgA, which are still not fully understood (1). Unlike any other Ig class, IgA occurs in different oliogomeric states, circulating in serum predominantly as monomeric IgA at 3 mg/ml, and at mucosal surfaces as polymeric IgA, mainly as dimeric IgA covalently linked with secretory component (SC)3(2). In serum, IgA is the second most abundant Ab class, with ∼90% being IgA1. Monomeric and dimeric IgA (dIgA) interact with the cell surface FcαR (FcαRI; CD89) to perform immune functions including the linking of the humoral and cellular immune response (3). In mucosal linings, lymphoid cells produce dIgA locally, which is secreted into the mucosa as secretory IgA (SIgA). Transport of dIgA across the epithelium into luminal secretions, and the consequent formation of SIgA is facilitated by the polymeric Ig receptor (pIgR). dIgA binds to pIgR at the basolateral surface of the epithelial cell, whereupon it is transcytosed across the cell and released into the lumen following cleavage of pIgR at the apical membrane. SIgA is released, with the extracellular pIgR domains (now called SC) remaining bound to the IgA dimer. The solution structure of SC has recently been elucidated (4).

An increased IgA level in serum can be associated with IgA nephropathy (IgAN) which is the most common form of chronic glomerulonephritis worldwide. IgAN is a renal disease involving the inflammation of the glomeruli in the kidneys, and is characterized by the mesangial deposition of polymeric IgA1. IgAN is caused by either an abnormality in IgA1 itself or in the production of IgA1, rather than an intrinsic abnormality in the glomerulus (5). Some reported characteristics of IgA in IgAN include abnormal O-glycosylation and an increased λ/κ L chain ratio (6). Serum from IgA myeloma patients can be rich in dIgA1 and polymeric IgA, although this does not lead to IgAN. In IgA myeloma and in other diseases characterized by high levels of dIgA1 and polymeric IgA, these forms of IgA are thought to play an important role in some of the pathology associated with these diseases (7).

Polymeric IgA consists of two (or sometimes three or four) IgA monomers bound covalently through a joining chain (J chain) (Fig. 1). Monomeric IgA has a 12-domain structure arranged as two H and two L chains (Fig. 1). Crystal structures are known for the murine Fab region of IgA and the human Fc region of IgA1 bound to FcαRI (8, 9). The topology of each of the IgA domains is a β-sheet sandwich comprised of the DEBA and GFC β-strands for the seven-stranded C-type domains and the DEBA and GFCC′C″ β-strands for the nine-stranded V-type domain (10). IgA1 and IgA2 differ significantly in the hinge region where IgA2 lacks a Pro-, Ser-, and Thr-rich sequence of 13 aa that is present in IgA1 (11, 12). IgA2 also differs in that in the commonest allotype (IgA2m(1)), the L chains are not covalently linked to the H chains but to each other, and IgA2 is held together only by noncovalent interactions (13). Both IgA subclasses have an additional 18-aa C-terminal peptide on each H chain called the tailpiece (Fig. 1). Cys471 in the tailpiece of the two IgA monomers forms a disulphide bridge with the J chain (14). Thus, the tailpiece is important for the dimerization of IgA1 and IgA2 (15). The J chain is an 18-kDa Ig-related glycoprotein containing 137 residues and one N-linked carbohydrate.

FIGURE 1.

The domain structure and sequence of dIgA1. dIgA1 is formed by two IgA1 monomers that are covalently linked with the J chain. Each IgA1 H chain contains the VH, CH1, CH2, and CH3 domains, and each L chain contains the VL and CL domains. The V and C domains are highlighted with gray and white backgrounds, respectively. The CDRs (black crescents), the 23-residue hinge peptide (dashed line), and the 18-residue C-terminal tailpiece (dashed line) are highlighted. Inter-H chain disulphide bridges at Cys241-Cys241 and two at Cys242-Cys299 are shown as three black lines. Cys471 in one tailpiece of each Fc fragment is disulphide-bridged with either Cys14 or Cys68 in the J chain. The N-linked oligosaccharide sites at Asn263 and Asn459 are denoted by filled symbols (•). The approximately five O-linked oligosaccharides in each hinge are denoted by open symbols (○).

FIGURE 1.

The domain structure and sequence of dIgA1. dIgA1 is formed by two IgA1 monomers that are covalently linked with the J chain. Each IgA1 H chain contains the VH, CH1, CH2, and CH3 domains, and each L chain contains the VL and CL domains. The V and C domains are highlighted with gray and white backgrounds, respectively. The CDRs (black crescents), the 23-residue hinge peptide (dashed line), and the 18-residue C-terminal tailpiece (dashed line) are highlighted. Inter-H chain disulphide bridges at Cys241-Cys241 and two at Cys242-Cys299 are shown as three black lines. Cys471 in one tailpiece of each Fc fragment is disulphide-bridged with either Cys14 or Cys68 in the J chain. The N-linked oligosaccharide sites at Asn263 and Asn459 are denoted by filled symbols (•). The approximately five O-linked oligosaccharides in each hinge are denoted by open symbols (○).

Close modal

The three-dimensional structures of the different forms of polymeric IgA are not well-understood. Structures for monomeric human IgA1 and IgA2 together with the covalent adduct of IgA1 with human serum albumin (HSA) have been determined by constrained scattering modeling (16, 17, 18). The overall structure of human myeloma dIgA has been observed by electron microscopy as a tail-to-tail arrangement of the Fc regions in the two IgA monomers (19, 20, 21, 22). However, the conditions of electron microscopy work using heavy metal stains in vacuo can be harsh, and the solution arrangement of the monomers in dIgA1 was unknown up to now. The size of dIgA1, together with its high glycosylation and long flexible hinge and tailpiece regions, suggests that its crystal structure will be hard to achieve. In such a situation, x-ray and neutron scattering and analytical ultracentrifugation combined with constrained modeling, will result in a medium resolution structure determination that will reveal molecular details within dIgA (23). By this, we show that myeloma dIgA1 is formed as a near-planar molecule in solution with a bent arrangement of Fc regions. This structure rationalizes the accommodation of the J chain within dIgA1, as well as effector functions of dIgA1, particularly its interaction with FcαRI receptors and SC in relation to the formation of SIgA. Insight is also provided into the role of dIgA1 in the development of IgAN.

A monoclonal dIgA1 with κ-class L chain was isolated from a human myeloma serum using a combination of thiophilic chromatography and jacalin-agarose lectin affinity chromatography (24, 25). Samples were subjected to size-exclusion chromatography to remove nonspecific aggregates and checked by reducing and nonreducing SDS-PAGE to confirm sample integrities before and after data collection. For x-ray scattering and analytical ultracentrifugation experiments, dIgA1 was dialyzed into Dulbecco’s PBS supplemented with EDTA and sodium azide as antibacterial preservatives (12.5 mM sodium phosphate, 140 mM NaCl, 0.5 mM EDTA, 0.02% NaN3, pH 7.4). For neutron scattering, the buffer was Dulbecco’s PBS as above, dialyzed at 6°C into 100% 2H2O for 36 h with four buffer changes. The dIgA1 amino acid and carbohydrate composition for data analyses corresponds to two IgA1 molecules as previously described (16). The human J chain sequence was taken from the SWISSPROT sequence code P01591, to which another biantennary complex-type oligosaccharide was added at Asn49. This resulted in the following: molecular mass, 344.8 kDa, an unhydrated volume of 439.4 nm3, a hydrated volume of 580.2 nm3, an absorption coefficient at 280 nm of 12.3 (1%, 1 cm), and a partial specific volume of 0.724 ml/g.

Solution scattering is a diffraction technique that studies the overall structure of biological macromolecules in random orientations in solution (23). X-ray scattering data were obtained on the Beamline ID02 at the European Synchrotron Radiation Facility (ESRF; Grenoble, France) with a ring energy of 6.0 GeV in single-bunch mode with currents from 11 to 12 mA to reduce the incident x-ray flux. The sample-to-detector distance of 3.0 m yielded a Q range from 0.07 nm−1 to 2.1 nm−1 (Q = 4 π sin θ/λ; 2θ = scattering angle; λ = wavelength). Samples were contained in water-cooled Perspex cells at 15°C, of path thickness 1 mm and mica windows of thickness 25 μm at concentrations of 0.58 and 1.16 mg/ml. Samples were measured in 10 time frames, each of 2 s, these being optimal to eliminate radiation damage effects. Neutron scattering data were obtained on Instrument LOQ in two beam sessions at the pulsed neutron source ISIS at the Rutherford Appleton Laboratory. Neutrons were derived from proton beam currents of ∼180 μA. Samples and buffers were measured in 2-mm-thick rectangular quartz Hellma cells positioned in a thermostated rack at 15°C. Data acquisitions lasted 4–13 h using concentrations of 0.83 and 1.16 mg/ml. Other details, including calibrations and data reduction, the Guinier analyses to determine the radius of gyration RG and the radius of gyration of the cross-sectional structure RXS, and the calculation of the distance distribution function P(r) using GNOM, are described previously (17). For GNOM, the dIgA1 x-ray I(Q) curve contained 524 data points between Q values of 0.98 and 2.02 nm−1 and was fitted with a Dmax set as 26 nm. The dIgA1 neutron I(Q) curve contained 61 data points between Q values of 0.21–2.1 nm−1 and was fitted with a Dmax set as 23.5 nm.

Analytical ultracentrifugation studies macromolecular structures in solution by following their sedimentation behavior on subjecting these to a high centrifugal force (23). Sedimentation equilibrium data were acquired over 45 h using six-sector cells with column heights of 2 mm at rotor speeds of 5,000, 8,000, 11,000, 14,000, 17,000, and 20,000 rpm at 20°C on a Beckman XL-I instrument equipped with an AnTi50 rotor. Absorbance scans at 280 nm and interference scans were recorded at six concentrations between 0.07 and 0.97 mg/ml. Molecular weights were determined on the assumption of a single species using ORIGIN version 4.1 (Microcal). The buffer density was calculated to be 1.00543 g/ml and the viscosity was taken as 0.01002 cp. Sedimentation velocity data were acquired at concentrations of 0.53 and 1.00 mg/ml over 16 h at rotor speeds of 10,000, 15,000, and 20,000 rpm in two-sector cells with column heights of 12 mm. Data were analyzed using DCDT+g(s*) time-derivative analyses and SEDFIT v9.4 based on the continuous c(s) distribution model with a resolution set as 150, while the cell meniscus and bottom, the frictional ratio of 1.638, and the partial specific volume , buffer density and viscosity were held fixed, and the baseline was allowed to float. Other details are as described previously (17).

The scattering and sedimentation data were subjected to constrained modeling to determine solution structures (23). The full coordinates of monomeric IgA1 (16) was used to generate dimer models. A planar starting model for the dimer was created by arranging the two monomers back to back. The J chain was not included. A dummy atom position was defined midway between the two Fc regions to act as the origin for the rotations and translations required to generate dIgA1 models. One IgA1 monomer was then rotated about the dummy atom, while the other monomer remained stationary. The rotations were in steps of 10° between 0° and 180° on each of the x-, y-, and z-axes, hence this resulted in 193 or 6859 models for each rotational search. Monomer separations were assessed by setting this as 1.40, 2.78, and 4.60 nm in three searches in which the two starting monomers shared a common y-axis. This separation was that between the two Cα atoms of Lys457 at the base of the two Fc regions, and the dummy atom is located at the mean position of all four Lys454 Cα atoms. In two final searches, the two IgA1 monomers had a separation of 2.78 nm, but the IgA1 monomer to be rotated was translated 2.0 and 4.0 nm along the x-axis.

The coordinate models were converted into Debye spheres to compute their scattering curves. A cube side length of 0.531 nm and a cutoff of four atoms consistently gave models within 2% of their total unhydrated volume of 417.2 nm3, i.e., that without the J chain as this had not been modeled, and gave a total of 2787 unhydrated spheres. After hydration, the number of hydrated spheres n was 3678. The values of n and the modeled RG, RXS-1, and RXS-2 parameters were calculated in the same Q ranges used for the experimental Guinier analyses. Models that fit within 5% of n and the x-ray Guinier values were ranked using the goodness-of-fit R factor of the model. A flat-background correction of 0.5% of I(0) was applied to the neutron fits to allow for residual incoherent scattering. The sedimentation coefficients s°20,w were calculated from the coordinate models using HYDROPRO version 7c. To take into account the hydration shell surrounding dIgA1, the recommended value of 0.31 nm for the atomic element radius for all atoms was used as an empirical approximation of this. Other details, including those of calibration studies that validate these modeling methods, are given elsewhere (16, 17). The 10 best-fit dIgA1 α-carbon coordinate models were deposited in the Protein Data Bank with the accession code 2QTJ. The α-carbon coordinate model of the related pentameric human IgM structure was deposited with the accession code 2RCJ (24).

Conventional purifications of polymeric IgA from normal human serum are inadequate for purifying the polymeric forms of IgA. A procedure based on thiophilic resins with a binding capacity of over 28 mg of Ig/ml of resin allows rapid purification of Igs which bind selectively to the resin while almost all the other serum proteins do not (25, 26). dIgA1 was purified from anonymized myeloma serum selected for its high polymer content. The myeloma serum had an Ig content of IgA at 15 g/L with a dimer to monomer ratio of 2:1, while IgG was present at 7g/L and IgM was at 1.5 g/L. dIgA1 was shown by nonreducing SDS-PAGE to be well-resolved from monomeric and trimeric IgA1 species in subsequent gel filtration (26). Its characterization is described in detail elsewhere (26). Reducing SDS-PAGE showed the expected bands corresponding to the IgA1 H chain at 62 kDa, and the L chain at 30 kDa. Monoclonal anti-κ L chain and anti-human J chain Abs were used in Western blots to confirm the presence of dIgA1. In Sephacryl S300 size-exclusion gel filtration, dIgA1 eluted as a single symmetrical peak, which was concentrated for data collection.

The averaged solution arrangement of the monomers within dIgA1 was analyzed by comparing x-ray scattering data I(Q) for dIgA1 with previous x-ray data for serum IgA (16, 17). Data were collected at concentrations between 0.6 and 1.2 mg/ml at the ESRF. The comparison of data from a single time-frame exposure with the average over 10 consecutive time frames showed that the Guinier analyses were unaffected by the exposure time, indicating that radiation damage or x-ray induced aggregation were absent. The time-averaged runs were thus used for data analyses. At the lowest Q values, these analyses resulted in linear plots in Q ranges of 0.10–0.17 nm−1 that gave RG values within satisfactory Q.RG limits (Fig. 2). The x-ray RG value for dIgA1 was 8.65 ± 0.27 nm (eight values) (Fig. 2 A). This is significantly larger than the RG value of 6.20 ± 0.13 nm for monomeric IgA1 (16) and 5.18 ± 0.09 nm for monomeric IgA2m(1) (17). The anisotropy ratio RG:RO (where RO is the RG value of the sphere with the same volume as the hydrated glycoprotein) for dIgA1 was 2.16, where that for monomeric IgA1 was 1.99, that for monomeric IgA2 was 1.66 and that for the IgA1-HSA complex was 2.13 (16, 17, 18). This showed that dimer formation had increased the degree of elongation of the IgA1 structure to form a more extended one.

FIGURE 2.

X-ray and neutron Guinier analyses for dIgA1. The Q.RG and Q.RXS ranges used to determine the RG, RXS-1, and RXS-2 values are represented by filled circles between the arrowed data points. A and B, The x-ray and neutron Q ranges used for the RG analyses were 0.10–0.17 nm−1. C and D, The Q ranges used to calculate the RXS-1 and RXS-2 values were 0.20–0.28 nm−1 and 0.56–1.04 nm−1, respectively. In A, C, and D, the x-ray fits for dIgA1 correspond to concentrations of 1.16 (upper) and 0.58 mg/ml (lower). In B, the neutron fits for dIgA1 correspond to 1.16 (upper) and 0.83 mg/ml (lower).

FIGURE 2.

X-ray and neutron Guinier analyses for dIgA1. The Q.RG and Q.RXS ranges used to determine the RG, RXS-1, and RXS-2 values are represented by filled circles between the arrowed data points. A and B, The x-ray and neutron Q ranges used for the RG analyses were 0.10–0.17 nm−1. C and D, The Q ranges used to calculate the RXS-1 and RXS-2 values were 0.20–0.28 nm−1 and 0.56–1.04 nm−1, respectively. In A, C, and D, the x-ray fits for dIgA1 correspond to concentrations of 1.16 (upper) and 0.58 mg/ml (lower). In B, the neutron fits for dIgA1 correspond to 1.16 (upper) and 0.83 mg/ml (lower).

Close modal

The shorter axes of dIgA1 were monitored by x-ray cross-sectional Guinier analyses that resulted in the RXS-1 and RXS-2 parameters. The ln (I(Q).Q) vs Q2 plots showed two linear regions in Q ranges of 0.20–0.28 nm−1 and 0.56 to 1.04 nm−1 (Fig. 2, C and D). The RXS-1 value of 3.94 ± 0.18 nm for dIgA1 is almost double that for monomeric IgA1, IgA2, and the IgA1-HSA complex which are 2.20, 2.47, and 2.31 nm, respectively (16, 17, 18). This indicates that the IgA1 dimer is significantly different in shape from the monomer. The RXS-2 value of 1.43 ± 0.07 nm is close to the RXS-2 of 1.56, 1.47, and 1.31 nm for IgA1, IgA2, and IgA1-HSA (16, 17, 18). The similarity of these four RXS-2 values suggests that the Fab regions retain their distinct identities within the monomeric and dimeric structures. If dIgA1 is crudely approximated as an elliptical cylinder in shape, the combination of the RG and RXS-1 values resulted in an overall length L for dIgA1 of 25 nm. The comparison of this length with that of 20.1 nm for IgA1 and 16.1 nm for IgA2 (16, 17) showed that the increased size of dIgA1 does not correspond to a doubling in length of the monomers.

Complementary neutron scattering data was obtained for dIgA1 at 0.6–1.5 mg/ml in 100% 2H2O buffer as a control for x-ray-induced radiation damage. This also acts as a control of internal scattering inhomogeneity effects caused by the estimated 9.6% carbohydrate content of dIgA1. A high negative protein-solvent scattering contrast is now used in place of the high positive contrast observed by x-rays. Neutron scattering also provided a different view of dIgA1 in that the hydration shell surrounding dIgA1 is visible by x-ray scattering, but not by neutron scattering (27). The neutron RG value of dIgA1 was 7.60 ± 0.05 nm (two values) (Fig. 2,B). This value supports the x-ray RG determination. However, this is 1 nm less than the x-ray RG value, which is a greater difference than that seen previously with IgA1 and IgA2 (16, 17). This is attributed to the positioning of most of the hydration shell at large distances from the centre of the dIgA1 structure. The neutron anisotropy ratio RG:RO was 2.08, and confirmed the x-ray value. As the consequence of poor counting statistics at large Q, no neutron RXS-1 or RXS-2 values were determined for dIgA1. Neutrons also lead to the molecular mass, as there is a linear relationship between the LOQ Guinier I(0)/c values for proteins measured in 2H2O buffers and molecular mass, where molecular mass = I(0)/c × 9.105 (16). The dIgA1 neutron Guinier I(0)/c value of 0.33 ± 0.05 resulted in a molecular mass value of 297 ± 45 kDa. This measurement is within error of the composition-derived value of 345 kDa for dIgA1, showing that the scattering data is consistent with the expected composition of dIgA1 with 9 N-linked and 20 O-linked oligosaccharides (Fig. 1).

Structural dimensions are provided by the transformation of the I(Q) curve into the distance distribution function P(r) curve. The mean RG values determined from the x-ray and neutron P(r) curves were 8.67 ± 0.17 nm (six values) and 7.47 nm (two values), respectively. These were consistent with the Guinier RG values. The x-ray P(r) curve for dIgA1 was of better quality than the neutron P(r) curve for reason of better signal-noise ratios (Fig. 3), nonetheless the two curves were reproducible in that a double peak was observed with maxima of M1 and M2 that were located at similar r values of 4.9–5.1 and 9.9–10.1 nm, respectively. These two peaks correspond to an abundance of interatomic vectors within the dIgA1 solution structure. The peak M1 at 4.9–5.1 nm was also evident in the P(r) curve for monomeric IgA1 and IgA2 and for myeloma IgA1-HSA at 3.7, 4.5, and 4.9 nm, respectively (16, 17, 18). M1 is assigned to the most commonly occurring distance within a single Fab or Fc region, each of approximate length 8 nm. Hence, M1 in Fig. 3 is consistent with a well-defined Fab region within the solution structure of dIgA1. M2 at 9.9–10.1 nm is of greater intensity than M1 when this is compared with monomeric IgA1 and IgA2 (16, 17). In the absence of molecular modeling (below), the most likely reason for this intensity difference is that there are a large number of interatomic vectors close to 10 nm between the four Fab regions that are held in a comparatively inflexible arrangement. The maximum length, L, is determined from where the P(r) curve reaches zero at large r values. For dIgA1, L was found to be 26 nm (x-ray) and 23 nm (neutron). The shorter L value with neutrons indicates that a significant proportion of the hydration shell is located at the periphery of the dIgA1 structure. Simple geometric considerations of planar end-to-end arrangements of two monomer structures suggested by Fig. 1 readily lead to L values close to 26 nm. Hence, the dIgA1 solution structure is concluded to be formed as an extended arrangement of two monomers joined at the end of their Fc regions, and not as a more compact solution structure that might resemble a bouquet of four Fab flowers positioned on two Fc stalks.

FIGURE 3.

X-ray and neutron distance distribution functions P(r) for dIgA1. A and B show the x-ray and neutron P(r) curves respectively. The concentrations of dIgA1 were 1.15 (x-rays) and 1.16 mg/ml (neutrons). The maxima M1 and M2 of the P(r) curve depict two frequently occurring distances within the dIgA1 structure. The positions of M1 and M2 are 4.94 and 9.88 nm, respectively (x-rays), and 5.06 and 10.14 nm, respectively (neutrons). The length of IgA1 is denoted by L at the r value where the P(r) curve reaches 0. The L values for dIgA1 were 26 (x-rays) and 22 nm (neutrons).

FIGURE 3.

X-ray and neutron distance distribution functions P(r) for dIgA1. A and B show the x-ray and neutron P(r) curves respectively. The concentrations of dIgA1 were 1.15 (x-rays) and 1.16 mg/ml (neutrons). The maxima M1 and M2 of the P(r) curve depict two frequently occurring distances within the dIgA1 structure. The positions of M1 and M2 are 4.94 and 9.88 nm, respectively (x-rays), and 5.06 and 10.14 nm, respectively (neutrons). The length of IgA1 is denoted by L at the r value where the P(r) curve reaches 0. The L values for dIgA1 were 26 (x-rays) and 22 nm (neutrons).

Close modal

To confirm the molecular mass of dIgA1, sedimentation equilibrium experiments were performed at six concentrations between 0.07 and 0.97 mg/ml and at six rotor speeds using both interference optics and absorbance optics at 280 nm (Materials and Methods). On the assumption that a single species was present, sedimentation equilibrium fits were conducted with nine curves at each dIgA1 concentration using both data types. The individual fits gave random fit residuals and molecular mass values that ranged between 280 and 345 kDa. Their extrapolation to zero concentration gave a mean molecular mass of 335 (±15) kDa (data not shown). These are within error of the molecular mass of 345 kDa calculated from the sequence of dIgA1 if the 29 glycosylation sites in this are all occupied by biantennary oligosaccharides. Had the glycosylation of dIgA1 corresponded to 29 tetra-antennary oligosaccharides, the molecular mass would have been 357 kDa. This outcome confirmed the neutron molecular mass determination (see above).

The sedimentation coefficient s°20,w provides an independent measure of macromolecular elongation to the RG value. For dIgA1, sedimentation velocity experiments were performed at three speeds at two concentrations (Materials and Methods). The c(s) distribution plots using SEDFIT software identify the species present. All the experimental and fitted dIgA1 boundaries showed good visual agreement with satisfactory root mean square deviation values (Fig. 4, A and B). The resulting c(s) distributions in Fig. 4, C and D, showed that only a single species of dIgA1 was present with a s°20,w value of 9.7 S. The conversion of the c(s) plots to molecular mass distributions c(M) showed that this corresponded to a molecular mass of 342 (±11) kDa. This agreed well with the sedimentation equilibrium value of 335 (±15) kDa and sequence-derived molecular mass of 345 kDa. No higher oligomers were detected in the c(s) plots. An analysis based on 4–20 scans at the center of the sedimentation profile using the g(s*) method in DCDT+ software confirmed the SEDFIT results. Here, the dIgA1 scans gave a s°20,w value of 9.8 S (interference) and 9.9 S (absorbance) (Fig. 5). The SEDFIT and DCDT+ determinations agree well with previous literature reporting a s°20,w value of 10 S (28), 9.65 S (21), and 9.2–9.3 S (29). The s°20,w value leads to the frictional ratio f/fO, where fO is the f value of the sphere with the same volume as the hydrated glycoprotein. Its value for dIgA1 of 1.63 is slightly larger than those of 1.56 found for monomeric IgA1 and 1.53 for monomeric IgA2 (17). This indicates that dIgA1 is more elongated than IgA1, which is consistent with the x-ray and neutron scattering results.

FIGURE 4.

Sedimentation velocity c(s) distribution analyses of dIgA1 using SEDFIT. A and B, The black circles represent the experimental data and the continuous white lines represent the fits, where only every sixth scan of the 120 scans used in the Lamm fits is shown for reason of clarity. The interference scans for dIgA1 at a concentration of 0.96 mg/ml and a rotor speed of 20,000 rpm are shown in A, while the corresponding absorbance scans at 280 nm are shown in B. The scans were recorded at 5-min intervals. C and D, The corresponding c(s) plots are shown from which the sedimentation coefficient for dIgA1 was determined to be 9.7 S.

FIGURE 4.

Sedimentation velocity c(s) distribution analyses of dIgA1 using SEDFIT. A and B, The black circles represent the experimental data and the continuous white lines represent the fits, where only every sixth scan of the 120 scans used in the Lamm fits is shown for reason of clarity. The interference scans for dIgA1 at a concentration of 0.96 mg/ml and a rotor speed of 20,000 rpm are shown in A, while the corresponding absorbance scans at 280 nm are shown in B. The scans were recorded at 5-min intervals. C and D, The corresponding c(s) plots are shown from which the sedimentation coefficient for dIgA1 was determined to be 9.7 S.

Close modal
FIGURE 5.

Time-derivative sedimentation velocity analyses of dIgA1 using DCDT+. In the g(s*) plots, the experimental data are represented by circles, and the fits are shown by the continuous black lines. The goodness-of-fit residuals are shown above each g(s*) analysis. The arrowed s20,w values for dIgA1 using A interference and B absorbance optics are 9.8 and 9.9 S, respectively. The fits are based on data for dIgA1 at 0.96 mg/ml and a rotor speed of 20,000 rpm as in Fig. 4. The fits were determined from totals of 4–20 scans midway through the sedimentation experiment shown in Fig. 4, A and B.

FIGURE 5.

Time-derivative sedimentation velocity analyses of dIgA1 using DCDT+. In the g(s*) plots, the experimental data are represented by circles, and the fits are shown by the continuous black lines. The goodness-of-fit residuals are shown above each g(s*) analysis. The arrowed s20,w values for dIgA1 using A interference and B absorbance optics are 9.8 and 9.9 S, respectively. The fits are based on data for dIgA1 at 0.96 mg/ml and a rotor speed of 20,000 rpm as in Fig. 4. The fits were determined from totals of 4–20 scans midway through the sedimentation experiment shown in Fig. 4, A and B.

Close modal

Constrained modeling of the x-ray data was used to determine solution structures for dIgA1. This was initiated using the monomeric IgA1 solution structure model and the dIgA1 composition (16). This IgA1 structure was assumed to be unchanged when incorporated into dIgA1. The 18-kDa J chain domain was not included because it comprises 5.1% by mass of dIgA1, and it is located centrally in the dIgA1 structure where it will minimally perturb the scattering curve (Fig. 1). The trial dIgA1 models assumed that the two monomers are connected end-to-end through their Fc regions as shown schematically in Fig. 1. This is consistent with the 26-nm length from Fig. 3,A and the dimensions of the IgA1 monomer. Five searches were performed in this first cycle of fits. All five searches held monomer 1 fixed (Fig. 6 A), and rotated monomer 2 about the dummy atom about the x-, y-, and z-axes in 10° steps (Materials and Methods). Search 1 systematically explored all orientations between monomers 1 and 2. Searches 2–5 repeated this to explore the effect of increasing the Fc-Fc separation between the IgA1 monomers. Searches 2–5 allowed for the possibility that the J chain separates the two ends of the Fc regions. Thus, in searches 1–3, monomers 1 and 2 were separated by 1.40, 2.78, and 4.60 nm, respectively between the base of the two Fc regions arranged coaxially with respect to each other (Materials and Methods). In searches 4 and 5, using a separation of 2.78 nm, monomer 2 was translated 2.0 and 4.0 nm, respectively, along the x-axis to explore asymmetry in dIgA1. Each search yielded 6859 models. The scattering curves and their parameters were calculated from the models for comparison with the experimental data.

FIGURE 6.

The starting model for the dIgA1 searches and the 1410 best-fit rotational solutions for the dIgA1 structure. A, The starting dIgA1 molecule with a separation of 1.40 nm is shown in relation to the x-, y-, and z-axes. The x- and y-axes are in the plane of the paper, and the z-axis is perpendicular to this. The dummy atom acting as the center of rotation between monomers 1 and 2 is shown as an open circle. Monomer 1 is held fixed in the searches. B–D, Projections in the XY, XZ, and YZ planes are shown of the 1410 best-fit structures from the 1.40 nm search that satisfy the absence of steric overlap and RG filters (Table I). The best-fit 1410 models of dIgA1 from the 1.40 nm search are shown as open circles, within which the 54 best-fit models are shown in light gray, and the 10 best-fit models are shown in dark gray (TableI). The best-fit model is shown as a black diamond. The best-fit model corresponds to a x, y, z rotation of 30°, 170°, 30°. The poor fit models show x-, y-, or z-axis rotations close to 90° (▴, ▾, and ▪). E, Comparison of the x-ray R factors with the x-ray RG values in searches 1–5. The 6859 rotational models from search 1 are shown as open circles, while those from searches 2–5 are shown in light gray. The symbols and color coding follows that of B–D, with the best-fit model indicated by a horizontal arrow and the poor-fit models indicated by three vertical arrows. The vertical dashed line indicates the experimental x-ray RG value. F, Comparison of the neutron R factors with the neutron RG values in searches 1–5. Other details follow those in BE.

FIGURE 6.

The starting model for the dIgA1 searches and the 1410 best-fit rotational solutions for the dIgA1 structure. A, The starting dIgA1 molecule with a separation of 1.40 nm is shown in relation to the x-, y-, and z-axes. The x- and y-axes are in the plane of the paper, and the z-axis is perpendicular to this. The dummy atom acting as the center of rotation between monomers 1 and 2 is shown as an open circle. Monomer 1 is held fixed in the searches. B–D, Projections in the XY, XZ, and YZ planes are shown of the 1410 best-fit structures from the 1.40 nm search that satisfy the absence of steric overlap and RG filters (Table I). The best-fit 1410 models of dIgA1 from the 1.40 nm search are shown as open circles, within which the 54 best-fit models are shown in light gray, and the 10 best-fit models are shown in dark gray (TableI). The best-fit model is shown as a black diamond. The best-fit model corresponds to a x, y, z rotation of 30°, 170°, 30°. The poor fit models show x-, y-, or z-axis rotations close to 90° (▴, ▾, and ▪). E, Comparison of the x-ray R factors with the x-ray RG values in searches 1–5. The 6859 rotational models from search 1 are shown as open circles, while those from searches 2–5 are shown in light gray. The symbols and color coding follows that of B–D, with the best-fit model indicated by a horizontal arrow and the poor-fit models indicated by three vertical arrows. The vertical dashed line indicates the experimental x-ray RG value. F, Comparison of the neutron R factors with the neutron RG values in searches 1–5. Other details follow those in BE.

Close modal

dIgA1 was revealed to have a near-planar structure in solution from the 6859 models from search 1 (Fig. 6, BD). This gave the best agreements with the x-ray data. The initial removal of bad-fit models based on the absence of steric overlap and the agreement with the experimental RG values left only 1410 dimer models (Table I). This removed models with x-axis and z-axis rotations of 90° in Fig. 6, ruling out structures with the two monomers perpendicular to each other. The selection of models with RXS-1 and RXS-2 values within 5% of the experimental values gave 54 models (light gray circles in Fig. 6; Table I). By this, all the models with y-axis rotations of 90° were now removed, showing that only near-planar dIgA1 models fit the scattering data. No completely planar structures (x- and y-axis rotations close to 0° or 10° in Fig. 6,B) gave good fits. The best 10 of these 54 models (dark gray circles in Fig. 6) were identified by ranking their R-factor values (Table I). The best-fit dIgA1 model has the lowest R factor of 5.5% (black circle in Fig. 6; Table I). This has a near-planar structure with an x-, y-, and z-axis rotation of 30°, 170°, and 30° (Fig. 7,K). It gave good curve fits out to a Q value of at least 1 nm−1 for both the x-ray and neutron data, and reproduced the M and L values and the major features of the P(r) curves (Fig. 7, A and F). All 10 best dIgA1 models showed a low degree of nonplanarity (x-axis rotations between 0° and 40°; y-axis rotations between 150° and 20°; z-axis rotations between 20° and 40°). The two Fc regions are approximately coaxial, but exhibit a small degree of bend between them.

Table I.

Summary of the x-ray and neutron modeling fits for the dIgA1 solution structure

Filter for Best-Fit ModelsModelsHydrated Spheres (n)X-Ray RG (nm)X-Ray RXS-1 (nm)X-Ray RXS-2 (nm)Neutron RG (nm)X-Ray R FactorNeutron R Factors°20,w (S)
Monomer reorientation fits          
 None 6859 2215–3700 6.51–9.01 1.87–5.37 0.93–2.16 5.97–7.94 5.5–20.6 5.9–14.7 naa 
n, RG (Fig. 61410 3426–3700 8.22–9.01 3.86–5.37 1.21–1.94 7.28–7.94 5.5–11.0 5.9–9.4 na 
n, RG, RXS-1, RXS-2 54 3565–3681 8.22–8.52 3.92–4.33 1.36–1.51 7.30–7.53 5.5–7.5 7.5–9.3 na 
n, RG, RXS-1, RXS-2, R factor 10 3587–3649 8.22–8.52 4.22–4.33 1.42–1.51 7.34–7.53 5.5–5.9 7.5–8.0 9.76–9.89 
 Best fit (Fig. 7 K3612 8.32 4.31 1.51 7.41 5.5 7.6 9.85 
Fab conformational fits          
n, RG, RXS-1, RXS-2, R factor 10 3884–3933 8.70–8.89 3.78–4.11 1.58–1.70 7.49–7.63 5.5–5.7 7.3–8.0 8.84–9.03 
 Best fit (Figs. 7,O and 93932 8.61 4.11 1.66 7.49 5.5 7.7 8.94 
Experimental values na 3606 8.65 ± 0.27 3.94 ± 0.18 1.43 ± 0.07 7.60 ± 0.05 na na 9.70 (SEDFIT)  9.85 (DCDT+
Poor-fit dIgA1 models          
 Fig. 7 L 3616 7.87 4.13 1.57 7.12 10.6 9.7 10.17 
 Fig. 7 M 3630 8.94 5.29 1.65 7.91 10.6 7.7 9.35 
 Fig. 7 N 3559 7.76 3.54 1.61 7.00 10.8 11 10.18 
Filter for Best-Fit ModelsModelsHydrated Spheres (n)X-Ray RG (nm)X-Ray RXS-1 (nm)X-Ray RXS-2 (nm)Neutron RG (nm)X-Ray R FactorNeutron R Factors°20,w (S)
Monomer reorientation fits          
 None 6859 2215–3700 6.51–9.01 1.87–5.37 0.93–2.16 5.97–7.94 5.5–20.6 5.9–14.7 naa 
n, RG (Fig. 61410 3426–3700 8.22–9.01 3.86–5.37 1.21–1.94 7.28–7.94 5.5–11.0 5.9–9.4 na 
n, RG, RXS-1, RXS-2 54 3565–3681 8.22–8.52 3.92–4.33 1.36–1.51 7.30–7.53 5.5–7.5 7.5–9.3 na 
n, RG, RXS-1, RXS-2, R factor 10 3587–3649 8.22–8.52 4.22–4.33 1.42–1.51 7.34–7.53 5.5–5.9 7.5–8.0 9.76–9.89 
 Best fit (Fig. 7 K3612 8.32 4.31 1.51 7.41 5.5 7.6 9.85 
Fab conformational fits          
n, RG, RXS-1, RXS-2, R factor 10 3884–3933 8.70–8.89 3.78–4.11 1.58–1.70 7.49–7.63 5.5–5.7 7.3–8.0 8.84–9.03 
 Best fit (Figs. 7,O and 93932 8.61 4.11 1.66 7.49 5.5 7.7 8.94 
Experimental values na 3606 8.65 ± 0.27 3.94 ± 0.18 1.43 ± 0.07 7.60 ± 0.05 na na 9.70 (SEDFIT)  9.85 (DCDT+
Poor-fit dIgA1 models          
 Fig. 7 L 3616 7.87 4.13 1.57 7.12 10.6 9.7 10.17 
 Fig. 7 M 3630 8.94 5.29 1.65 7.91 10.6 7.7 9.35 
 Fig. 7 N 3559 7.76 3.54 1.61 7.00 10.8 11 10.18 
a

na, Not available.

FIGURE 7.

X-ray and neutron scattering curve fits for the best-fit model for dIgA1 from two fit cycles and three poor-fit models. The symbols follow that of Figs. 6 and 7. A, F, and K show the best-fit dIgA1 model from fit cycle 1. The x-ray and neutron experimental I(Q) (open circles) and P(r) curves (dashed lines) are compared with the modeled best-fit curves (continuous black lines). The P(r) curve is shown as an inset in the top right corner. The ribbon trace of this best-fit model (dark gray) is shown in K, with the x, y, z rotation shown in brackets and the carbohydrate chains in light gray. The tip-to-tip distance between the two Fab regions in an IgA1 monomer is arrowed in K. The curve fits in B, G, and L; C, H, and M; and D, I, and N show the curve fits for three poor-fit models in which the two monomers are perpendicular to each other. E, J, and O show the best-fit dIgA1 model from fit cycle 2. Other details are as above for A, F, and K.

FIGURE 7.

X-ray and neutron scattering curve fits for the best-fit model for dIgA1 from two fit cycles and three poor-fit models. The symbols follow that of Figs. 6 and 7. A, F, and K show the best-fit dIgA1 model from fit cycle 1. The x-ray and neutron experimental I(Q) (open circles) and P(r) curves (dashed lines) are compared with the modeled best-fit curves (continuous black lines). The P(r) curve is shown as an inset in the top right corner. The ribbon trace of this best-fit model (dark gray) is shown in K, with the x, y, z rotation shown in brackets and the carbohydrate chains in light gray. The tip-to-tip distance between the two Fab regions in an IgA1 monomer is arrowed in K. The curve fits in B, G, and L; C, H, and M; and D, I, and N show the curve fits for three poor-fit models in which the two monomers are perpendicular to each other. E, J, and O show the best-fit dIgA1 model from fit cycle 2. Other details are as above for A, F, and K.

Close modal

The best-fit dIgA1 structure accounted for the large difference between the x-ray and neutron RG values. The superimposition of an α-carbon view with both the unhydrated and hydrated sphere models showed that the majority of the hydration spheres occupied voids within the Fab region between the two V and C domain pairs, the O-glycosylated hinge peptides, and the upper part of the Fc region between the two CH2 domains. As these hydration spheres are positioned far from the center of mass, they have a proportionately larger effect on the RG values. This is reflected in the shift of the minimum in the distributions from 8.5 nm for X-rays (Fig. 6,E) to 7.6 nm for neutrons (Fig. 6 F).

This first cycle of curve fits was corroborated by four controls:

1) The near planarity of the deduced IgA dimer structures was clearly supported by the poor I(Q) and P(r) fits in Fig. 7, B–D, obtained with three out-of-plane dIgA1 structures (▴, ▾, and ▪ in Fig. 6, B–F). This argues against a previously discussed perpendicular arrangement for the two Fc regions (20, 21).

2) No improved dIgA1 fits were obtained with the increased separations in searches 2–5. For search 1, the lowest R factors of the best models agree well with the experimental RG (dashed line in Fig. 6,E), RXS-1 and RXS-2 values (data not shown). For searches 2–5 (gray in Fig. 6, E and F), the RG values of the models increased with an increased separation of the Fc regions and consequently deviated away from the experimental RG value.

3) The effect of the J chain on the modeling was tested by adding the D4 domain of SC to the center of the best-fit dIgA1 structure (4). This was located proximate to two IgA1 tailpieces in the plane between two Fc regions to follow the previous modeling of the IgM pentamer (24). It is stressed that this J chain model and its orientation within dIgA1 are arbitrary. There is no crystal structure for J chain, although two models have been proposed (30, 31). This inclusion did not significantly perturb the best-fit model, as it only caused small reductions in the x-ray RG, RXS-1, and RXS-2 values of 0.01–0.07 nm, and the R factor to increase by 1.0%. In addition, search 1 was repeated with the J chain present. Near-planar dIgA1 structures were again obtained, and the best-fit R factor increased by only 0.3%.

4) The dIgA1 searches did not incorporate the IgA1 Fc crystal structure (9) for reason of consistency with the previous modeling of IgA1 and IgA1-HSA. This crystal structure was superimposed upon that in the dIgA1 model to show that this had little effect. The root mean square deviation of the 418 superimposed α-carbon atoms was 0.43 nm, and its overall dimensions were unchanged. The resulting x-ray RG, RXS-1, and RXS-2 values differed from the search 1 best-fit model by only 0.02, 0.06, and 0.07 nm, respectively, and the x-ray R-factor improved slightly from 5.5 to 5.4%.

A second cycle of curve fits was performed to investigate randomized reorientations in the four Fab regions in dIgA1. The dimeric Fc region from the best-fit model of Fig. 7,K was used for this. Constrained modeling searches were performed to follow method 2 used previously for monomeric IgA1 (16). A total of 3000 dIgA1 structures were created by superimposing four Fab regions and four copies of each of 3000 randomized IgA1 hinges onto the Fc dimer. The randomized hinges spanned lengths of 6.5–9.0 nm to test a sufficient range of Fab orientations, and this was verified by the minimum observed in Fig. 8,A. The curve fits showed that, while the same minimum R factor of 5.5% was obtained again, the modeled x-ray RG, RXS-1, and RXS-2 values were now in better agreement with the experimental values (Table I). The fit quality in Fig. 7, E and J, was similar to that of Fig. 7, A and F. Superimposition of the 32 best-fit final models showed that a limited conformational family of best-fit structures had resulted (Fig. 8, B and C) in which the two Fab regions displayed an extended and mostly T-shaped arrangement relative to their adjacent Fc region. All four Fab regions were independent of each other. When compared with monomeric IgA1 (16), the Fab regions were again positioned in approximately the same plane as the Fc region. However, Fig. 9 D shows that the Fab regions are slightly displaced upwards above the Fc fragment in dIgA1. Hence, dimer formation may have caused the Fab regions to move away from the Fc region when compared with the monomer. This suggests that the extended hinge conformation between the Fab and Fc regions has semiflexible connections between them.

FIGURE 8.

The best-fit models for dIgA1 from the Fab reorientation search. A, Comparison of the x-ray R factors with the x-ray RG values for 3000 models. The 54 best-fit models are shown in light gray and the 10 best-fit models are shown in dark gray, of which the best-fit model is shown as a black diamond (arrowed). The vertical dashed line indicates the experimental x-ray RG value. B and C, Two orthogonal views of the superimposition of the 32 best-fit models for dIgA1. The two Fc regions are shown as a black ribbon at the center of dIgA1, while the 32 sets of four Fab regions are shown in black outlines.

FIGURE 8.

The best-fit models for dIgA1 from the Fab reorientation search. A, Comparison of the x-ray R factors with the x-ray RG values for 3000 models. The 54 best-fit models are shown in light gray and the 10 best-fit models are shown in dark gray, of which the best-fit model is shown as a black diamond (arrowed). The vertical dashed line indicates the experimental x-ray RG value. B and C, Two orthogonal views of the superimposition of the 32 best-fit models for dIgA1. The two Fc regions are shown as a black ribbon at the center of dIgA1, while the 32 sets of four Fab regions are shown in black outlines.

Close modal
FIGURE 9.

Molecular architecture of the best-fit model for dIgA1. A, The 10 best-fit dIgA1 models from fit cycle 2 are shown projected onto the XY plane, with the Fc-Fc regions in each one shown in the same orientation in all 10 views. Monomers 1 and 2 are shown in green and blue, respectively. The first model (∗) is the best-fit dIgA1 model from Fig. 7,O. B, The best-fit dIgA1 model of Fig. 7,O is shown as a stereo pair in blue (H chains) and red (L chains). The J chain is shown in green. The hinge region and tailpiece are shown in pink, the CDR residues are shown in black, and the dIgA1 residues in the CH2 and CH3 domains that bind to the FcαRI receptor are shown in orange. Cys311 in the CH2 domain and Cys471 in the tailpiece are shown in yellow, and carbohydrate chains are shown in cyan. C and D, The best-fit dIgA1 model of Fig. 7 O is shown after superimposition of the crystal structure of the human IgA Fc region complexed with two FcαRI receptor molecules onto each Fc region in the dIgA1 model. The FcαRI receptors associated with monomer 1 are shown in green, and those with monomer 2 are shown in yellow. The view in C shows the two IgA1 monomers. That in D corresponds to a rotation by 90° of that in C to display the near-planarity of the dIgA1 structure and the projection of the FcαRI receptors above and below the dIgA1 structure.

FIGURE 9.

Molecular architecture of the best-fit model for dIgA1. A, The 10 best-fit dIgA1 models from fit cycle 2 are shown projected onto the XY plane, with the Fc-Fc regions in each one shown in the same orientation in all 10 views. Monomers 1 and 2 are shown in green and blue, respectively. The first model (∗) is the best-fit dIgA1 model from Fig. 7,O. B, The best-fit dIgA1 model of Fig. 7,O is shown as a stereo pair in blue (H chains) and red (L chains). The J chain is shown in green. The hinge region and tailpiece are shown in pink, the CDR residues are shown in black, and the dIgA1 residues in the CH2 and CH3 domains that bind to the FcαRI receptor are shown in orange. Cys311 in the CH2 domain and Cys471 in the tailpiece are shown in yellow, and carbohydrate chains are shown in cyan. C and D, The best-fit dIgA1 model of Fig. 7 O is shown after superimposition of the crystal structure of the human IgA Fc region complexed with two FcαRI receptor molecules onto each Fc region in the dIgA1 model. The FcαRI receptors associated with monomer 1 are shown in green, and those with monomer 2 are shown in yellow. The view in C shows the two IgA1 monomers. That in D corresponds to a rotation by 90° of that in C to display the near-planarity of the dIgA1 structure and the projection of the FcαRI receptors above and below the dIgA1 structure.

Close modal

The experimental sedimentation coefficient s°20,w value of 9.7 S for dIgA1 was compared with the s°20,w values of the 10 best-fit dIgA1 models (Table I). Using a previously established simulation procedure (Materials and Methods), the calculated s°20,w values for the 10 dIgA1 models ranged between 9.76 and 9.89 S from the first fit cycle and between 8.84 and 9.03 S from the second fit cycle (Table I). The two best-fit dIgA1 models had s°20,w values of 9.85 and 8.94 S, respectively. Although the second cycle of fits gives lower s°20,w values, the modeled values agree well with the experimental s°20,w value of 9.7 S (SEDFIT) and 9.85 S (DCDT+) (Table I). This supports the x-ray scattering best-fit dIgA1 model (Fig. 7, K and O).

dIgA is a building block of SIgA, one of the most important components of the immune system. Dimeric serum IgA is less well-characterized, although an altered structural form of dIgA1 is believed to be central to the pathology of a number of IgA-mediated diseases such as IgAN, the most common form of chronic glomerulonephritis. No differences in function between normal and myeloma dIgA1 have been identified. Although an IgA myeloma was used as the source of purified dIgA1, we have observed no difference in structural or functional properties between this myeloma dIgA1 and dIgA purified from normal serum or indeed, the monomeric myeloma IgA purified from the same sample compared with “normal” monomeric IgA. We have not defined its glycosylation in detail. However, the myeloma dIgA1 bound to jacalin and was eluted in the same way as normal dIgA1 (26). It was cleaved by IgA proteases, other proteases, and by a glycosidase from a Streptococcus species in the same way as normal dIgA1 (Ref. 26 , A. Almogren and M. A. Kerr, unpublished data). On SDS-PAGE, the myeloma dIgA1 had the same mobility as normal dIgA1 (26). Functionally, the myeloma dIgA1 triggered neutrophil respiratory burst through CD89 in the same way as normal dIgA1, and bound to purified SC. Hence, our new structure of dIgA will aid our understanding and subsequent ability to manipulate the mucosal immune system and clarify the pathology of diseases. The size of dIgA1, its high carbohydrate content and its flexibility mean that the crystallization of intact dIgA1 may not be realistic, thus constrained scattering modeling presently offers the only route to a molecular structure. Experimentally, we have shown that dIgA1 has a near-planar structure in solution with the two monomers orientated end-to-end through their Fc regions. The two Fc regions are slightly bent in their relative arrangement (Fig. 9). Because the dIgA1 structure was determined by fitting models against experimental scattering data, this is not a prediction method, and the final models qualify for deposition in the Protein Data Bank.

Although scattering modeling is not able to determine unique molecular structures, it is able to rule out poorer-fit models (Table I). Both search cycles successfully converged on one small family of related best-fit structures. One striking outcome is the close end-to-end proximity of the two Fc regions in dIgA1. The J chain does not act as a spacer, as sometimes suggested in cartoons of the IgA dimer. Likewise, the two Fc regions do not overlap as discussed in earlier reviews (32). A bent Fc arrangement was modeled, which has also been noted by electron microscopy (19). The length of the Fc-Fc region within dIgA1 in Fig. 9,B is close to 14.4 nm. This agrees well with the 14.0–15.5 nm, 13.8 (±2.6) nm and 12.5 nm determinations from electron microscopy (19, 20, 21, 28). The J chain is best accommodated in dIgA1 within a crevice formed by the bent Fc regions, although a crystal structure will be needed to confirm this. If the J chain had separated the two Fc regions, this would add 2.0–2.5 nm to the length of the Fc-Fc regions. This outcome was ruled out by our Searches 2 and 3. Our location of the J chain in dIgA1 is consistent with our scattering modeling of pentameric IgM (24). There, the J chain was best fitted within the planar disc of the five IgM Fc regions, and not as a spacer. The early electron microscopy studies of dIgA1 indicated flexible planar end-to-end structures (20, 21, 22, 28), but it was not known whether these near-planar structures exist in solution. The constrained modeling showed that the best-fit dIgA1 structures are near planar. Interestingly a near-planar dIgA1 structure would be consistent with the planar IgM structure (24). Models showing perpendicular arrangements of the two monomers or the four Fab regions were eliminated during the rejection of poor-fit models (Figs. 6, BD, and 7, B–D).

The solution structures for monomeric IgA1, the IgA1-HSA complex and dIgA1 all demonstrated similar T-shaped structures for the two Fab regions relative to the Fc region (16, 18). A well-separated Fab-Fc structure in all three cases was deduced from the observation of the M2 peak in the Pr curves, meaning that the two Fab regions are consistently seen on average to be extended away from the Fc region. Hinge flexibility of the Fab region relative to the Fc region is nonetheless expected, as observed from the variable conformations seen in electron microscopy studies of dIgA1, and the altered position of the Fab regions on going from the monomer to the dimer (Figs. 7,K and 9,A). This is in accordance with the ability of IgA1 to bind efficiently to a range of Ags. The maintenance of these extended hinge conformations can be rationalized in terms of its primary structure (PVPSTPPTPSPPTPPTPSPSCCH). Even though the hinge is comparatively long at 23 residues, the presence of O-linked NeuNAc.Gal.GalNAc trisaccharides, multiple prolines and the absence of glycines are expected to maintain their extended conformations. Interestingly, human IgD also possesses an O-glycosylated segment within its 64-residue hinge, and this has a similar T-shaped structure to that for IgA1 (24). As the four Fab regions are well separated within dIgA1, each Ag-binding site at the CDRs (CDR in Fig. 9 B) is potentially able to interact with repeated epitopes independently of each other. The valency of dIgA1 is therefore predicted to be four, and its avidity is predicted to be higher compared with its monomer. This assumes that the linearity of the dIgA1 structure has no bearing on its capacity to bind independently to four Ag molecules.

A near-planar dIgA1 structure would be well-adapted for interactions with its receptors. By this, dIgA1 is able to approach the host cell membrane surface for effective interactions with a cell surface receptor, most notably FcαRI which is responsible for IgA-mediated phagocytosis, oxidative burst, and Ab-dependent cellular cytoxicity among other roles (33, 34). The dIgA1 model shows that receptor binding would occur independently of Ag binding. The dIgA1 residues involved in binding to FcαRI are highlighted in orange in Fig. 9,B (9, 35). By superimposition of the Fc crystal structure complexed with two FcαRI molecules (9) upon the dIgA1 model, it is seen that the two FcαRI receptors will project out of the plane of dIgA1 but on opposite sides of this dIgA1 plane (green in Fig. 9, C and D). The superimposition of a second crystal structure onto the other monomer within dIgA1 shows likewise that the third and fourth FcαRI receptors project outward on both sides of this plane (yellow in Figs. 9, C and D). Previously it has been shown that two FcαRI receptors can bind simultaneously to the monomeric Fc region in solution, and it has been speculated that this bivalent binding occurs at cell surfaces (9, 17, 36). Fig. 9, C and D, likewise leads to the speculation that dIgA1 is able to bind up to two FcαRI receptors on a given cell surface, even though its predicted valency for receptors is four. The spatial separation between the C-terminal α-carbon atoms of the two FcαRI receptors bound to monomeric and dimeric IgA1 is similar at 12.3 and 10.2 nm, respectively, meaning that the same arrangement of FcαRI receptors will theoretically bind equally well to monomers or dimmers of IgA1. However, for reason of comparatively low receptor binding affinities in relation to the IgA1 serum concentrations, it seems more likely that a 1:1 stoichiometry results in vivo (33, 34).

The appearance of dIgA1 as a near-planar structure with a bent double Fc region leads to a possible mechanism for the formation of SIgA1 from dIgA1 and SC. The bend between the two Fc regions in the final models of Fig. 6 offers a covalently bound location for the J chain within the crevice that is created by this bend. This location may enable the J chain in dIgA1 to be sufficiently exposed to bind to a central region in SC, this being the first event in the selective transportation of dIgA1 across epithelial cells to form SIgA (37). The interaction between the C-terminal region of the J chain in dIgA1 and SC is noncovalent and represents a critical step for transport of this complex to mucosal surfaces to form SIgA (37). A “zipper effect” binding of SC to dIgA1 has been proposed, where the N-terminal D1 domain of SC first interacts noncovalently with dIgA1 (possibly with the Fc region), a central region in SC then interacts with the C-terminal region of the J chain, and dIgA1 and SC are brought into the correct position for covalent binding between Cys311 in the CH2 domain of dIgA1 and Cys502 in the D5 domain of pIgR. A J-shaped solution structure for free SC has been recently determined (4). It is possible that the center of this may form an initial contact with the J chain in dIgA1, then SC will unfold toward both ends of dIgA1 as the D1 and D5 domains interact successively across the two Fc regions. The wrapping of the D1–D5 domains around the Fc regions in dIgA1 then provides SIgA1 with a greater ability to resist proteolysis in the harsh environment of mucosal surfaces.

The near-planar dIgA1 structure provides insight into the formation of other polymeric IgAs. The structure of dimeric IgA2 is predicted to be similar to that of dIgA1 because the two monomers are joined through the base of their two Fc regions to the J chain and there are only minor sequence differences between the IgA1 and IgA2 isotypes (11). The hinge region is not involved in the formation of the IgA2 dimer. However, the binding site of SC to dimeric IgA2 differs from that to dIgA1, as SC binds covalently to dIgA1 but may not do so to dimeric IgA2 (18). The two dIgA isotypes may therefore differ in a manner that remains to be elucidated, most probably mediated by the closer proximity of the Fab regions to SC in secretory IgA2. The antigenic reach of the two isotypes will differ for reason of the different hinge structures, as previously discussed (16, 17), and dimer formation is expected to accentuate this difference. The formation of trimeric and tetrameric IgA structures (38) can be visualized by analogy with the structure of planar pentameric IgM in which its Fc regions are disulphide linked through the Cys414 residues in the Cμ3 domains and the Cys575 residues in the tailpieces (24). Hexameric IgM structures without J chain have been described in which a sixth IgM monomer is inserted within this plane (39). Cys311 in IgA1 (Fig. 1) is located at the identical sequence and structural position to that of Cys414 in IgM (9). By analogy with IgM, it is relatively straightforward to insert further monomers within this Fc plane to form larger IgA oligomers. Using molecular graphics based on the IgA1 Fc crystal structure and the IgM scattering structure, it was possible to create trimeric (and tetrameric) assemblies of three (and four) coplanar Fc regions starting from the dimer with the J chain present, but only if the Fab regions are absent. If the Fab regions and hinges are present, they will readily lead to significant steric conflicts with the assembly of Fc trimers and tetramers of IgA1 unless these Fab regions are significantly displaced out of the plane containing the three or four Fc regions.

IgAN is defined by the deposition of IgA in the glomerular mesangium of the kidney and this is predominantly polymeric IgA1 (5). A high level of polymeric IgA1 such as in IgA myelomas does not necessarily lead to IgAN. A reduced galactosylation of the hinge O-linked oligosaccharides has been implicated in the pathology of IgAN. It is possible that this change may destabilize the extended conformation of the IgA1 hinge and cause the Fab regions to be displaced. Hence glycosylation changes may provide one possible mechanism for IgA aggregate formation by permitting the Fc regions to self-associate. Another possibility for IgA aggregation arises from the consequences of the cleavage of the IgA1 hinge by bacterial proteases. This was seen to cause the cleaved Fc fragments derived from dimeric and trimer IgA1 (but not monomeric IgA1) to form disulphide-linked multimers (26). Our molecular graphics modeling of planar Fc rings of IgA by analogy with IgM (see above) lead to the hypothesis that, if the IgA1 hinges become more flexible through deglycosylation and cause the Fab regions to move away, or the Fab regions are partially or completely removed by bacterial proteolysis, then near-planar assemblies of trimers, tetramers, and higher oligomers of IgA can be generated starting from the double Fc region of dIgA1 and its J chain. If this hypothesis is correct, these Fc oligomers can be stabilized by disulphide bridges through Cys311 at the surface of the CH2 domain in IgA, and may lead to the IgA deposits seen in IgAN.

The elucidation of a near-planar solution structure of dIgA1 has yielded a better understanding of the function of polymeric forms of IgA. This opens the way for the constrained modeling of SIgA1 and SIgA2. Knowledge of the monomer arrangement within the dimer has provided insights into the assembly of this and higher oligomers of IgA1, and the way that this potentially interacts with the FcαRI receptor and SC. However, many unsolved issues remain, including the exact binding motifs within dIgA1 that interact with SC in the formation of SIgA, and the structure and composition of pathological dIgA1 forms that are implicated in diseases such as IgAN.

We thank Dr. R. E. Saunders for excellent computational support, J. Gor for ultracentrifugation support, Dr. S. Finet (ESRF, Grenoble, France) and Drs. R. K. Heenan and S. M. King (ISIS, Rutherford Appleton Laboratory) for instrumental support, and Dr. A. Robertson for assistance with the initial scattering data collection.

The authors have no financial conflict of interest.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1

This work was supported by the Biotechnology and Biological Sciences Research Council.

3

Abbreviations used in this paper: SC, secretory component; dIgA1, dimeric IgA1; SIgA, secretory IgA; pIgR, polymeric Ig receptor; IgAN, IgA nephropathy; HSA, human serum albumin.

1
Russell, M. W., M. Kilian, M. E. Lamm.
1999
. Biological activities of IgA. P. L. Ogra, and J. Mestecky, and E. Lamm, and W. Strober, and J. Bienenstock, and J. R. McGhee, eds.
Mucosal Immunology
2nd Ed.
225
-240. Academic Press, San Diego.
2
Kerr, M. A..
1990
. The structure and function of human IgA.
Biochem. J.
271
:
285
-296.
3
Morton, H. C., M. van Egmond, J. G. J. van de Winkel.
1996
. Structure and function of human IgA Fc receptors (FcαR).
Crit. Rev. Immunol.
16
:
423
-440.
4
Bonner, A., C. Perrier, B. Corthésy, S. J. Perkins.
2007
. The solution structure of human secretory component and implications for biological function.
J. Biol. Chem.
282
:
16969
-16980.
5
Barratt, J., J. Feehally.
2005
. IgA nephropathy.
J. Am. Soc. Nephrol.
16
:
2088
-2097.
6
Coppo, R., A. Amore.
2004
. Aberrant glycosylation in IgA nephropathy.
Kidney Int.
65
:
1544
-1547.
7
Mestecky, J., H. Suzuki, T. Yanagihara, Z. Moldoveanu, M. Tomana, K. Matousovic, B. A. Julian, J. Novak.
2007
. IgA nephropathy: current views of immune complex formation.
Contrib. Nephrol.
157
:
56
-63.
8
Suh, S. W., T. N. Bhat, M. A. Navia, G. H. Cohen, D. N. Rao, S. Rudikoff, D. R. Davies.
1986
. The galactan-binding immunoglobulin Fab J539: an X-ray diffraction study at 2.6 Å resolution.
Proteins
1
:
74
-80.
9
Herr, A. B., E. R. Ballister, P. J. Bjorkman.
2003
. Insights into IgA-mediated immune responses from the crystal structures of human FcαRI and its complex with IgA1-Fc.
Nature
423
:
614
-620.
10
Chothia, C., E. Y. Jones.
1997
. The molecular structure of cell adhesion molecules.
Annu. Rev. Biochem.
66
:
823
-862.
11
Torano, A., F. W. Putnam.
1978
. Complete amino acid sequence of the α 2 heavy chain of a human IgA2 immunoglobulin of the A2m (2) allotype.
Proc. Natl. Acad. Sci. USA
75
:
966
-969.
12
Putnam, F. W., Y. S. V. Lui, T. L. K. Low.
1979
. Primary structure of a human IgA1 immunoglobulin.
J. Biol. Chem.
254
:
2865
-2874.
13
Grey, H. M., C. A. Abel, W. J. Yount, H. G. Kunkel.
1968
. A subclass of human γA-globulins (γA2) which lacks the disulphide bonds linking heavy and light chains.
J. Exp. Med.
128
:
1223
-1236.
14
Garcia-Pardo, A., M. E. Lamm, A. G. Plaut, B. Frangione.
1981
. J chain is covalently bound to both monomer subunits in human secretory IgA.
J. Biol. Chem.
256
:
11734
-11738.
15
Atkin, J. D., R. J. Pleass, R. J. Owens, J. M. Woof.
1996
. Mutagenesis of the human IgA1 heavy chain tailpiece that prevents dimer assembly.
J. Immunol.
157
:
156
-159.
16
Boehm, M. K., J. M. Woof, M. A. Kerr, S. J. Perkins.
1999
. 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.
286
:
1421
-1447.
17
Furtado, P. B., P. W. Whitty, A. Robertson, J. T. Eaton, A. Almogren, M. A. Kerr, J. M. Woof, S. J. Perkins.
2004
. Solution structure determination of human IgA2 by X-ray and neutron scattering and analytical ultracentrifugation and constrained modelling: a comparison with human IgA1.
J. Mol. Biol.
338
:
921
-941.
18
Almogren, A., P. B. Furtado, Z. Sun, S. J. Perkins, M. A. Kerr.
2006
. Biochemical and structural properties of the complex between human immunoglobulin A1 and human serum albumin by X-ray and neutron scattering and analytical ultracentrifugation.
J. Mol. Biol.
356
:
413
-431.
19
Bloth, B., S.-E. Svehag.
1971
. Further studies on the ultrastructure of dimeric IgA of human origin.
J. Exp. Med.
133
:
1035
-1042.
20
Munn, E. A., A. Feinstein, A. J. Munro.
1971
. Electron microscope examination of free IgA molecules and of their complexes with antigen.
Nature
231
:
527
-529.
21
Feinstein, A., E. A. Munn, N. E. Richardson.
1971
. The three-dimensional conformation of M and A globulin molecules.
Ann. NY Acad. Sci.
190
:
104
-121.
22
Nezlin, R..
1998
.
The Immunoglobulins: Structure and Function
Academic Press, New York.
23
Perkins, S. J., A. I. Okemefuna, A. N. Fernando, A. Bonner, H. E. Gilbert, P. B. Furtado.
2008
. X-ray and neutron scattering data and their constrained molecular modelling.
Methods Cell Biol.
84
:
375
-423.
24
Perkins, S. J., A. Nealis, B. J. Sutton, A. Feinstein.
1991
. Solution structure of human and mouse immunoglobulin M by synchrotron X-ray scattering and molecular graphics modelling: a possible mechanism for complement activation.
J. Mol. Biol.
221
:
1345
-1366.
25
Sun, Z., A. Almogren, P. B. Furtado, B. Chowdhury, M. A. Kerr, S. J. Perkins.
2005
. Semi-extended solution structure of human myeloma immunoglobulin D determined by constrained X-ray scattering.
J. Mol. Biol.
353
:
155
-173.
26
Almogren, A., M. A. Kerr.
2008
. Irreversible aggregation of the Fc fragment derived from polymeric but not monomeric serum IgA1- implications in IgA-mediated disease.
Mol. Immunol.
45
:
87
-94.
27
Perkins, S. J..
2001
. X-ray and neutron scattering analyses of hydration shells: a molecular interpretation based on sequence predictions and modelling fits.
Biophys. Chem.
93
:
129
-139.
28
Dourmashkin, R. R., G. Virella, R. M. E. Parkhouse.
1971
. Electron microscopy of human and mouse myeloma serum IgA.
J. Mol. Biol.
56
:
207
-208.
29
Bjork, I., E. Lindh.
1974
. Gross conformation of human secretory immunoglobulin A and its component parts.
Eur. J. Biochem.
45
:
135
-145.
30
Zikan, J., J. Novotny, T. L. Trapane, M. E. Koshland, D. W. Urry, J. C. Bennett, J. Mestecky.
1985
. Secondary structure of the immunoglobulin J chain.
Proc. Natl. Acad. Sci. USA
82
:
5905
-5909.
31
Frutiger, S., G. J. Hughes, N. Paquet, R. Lüthy, J. C. Jaton.
1992
. Disulphide bond assignment in human J chain and its covalent pairing with IgM.
Biochemistry
31
:
12643
-12647.
32
Mestecky, J., J. R. McGhee.
1987
. Immunoglobulin A (IgA): molecular and cellular interactions involved in IgA biosynthesis and immune response.
Adv. Immunol.
40
:
153
-183.
33
Monteiro, R. C., J. G. van de Winkle.
2003
. IgA Fc receptors.
Annu. Rev. Immunol.
21
:
177
-204.
34
Woof, J. M., M. van Egmond, M. A. Kerr.
2005
. Fc receptors. J. R. Mestecky, and J. Bienenstock, and M. E. Lamm, and L. Mayer, and J. R. McGhee, and W. Strober, eds.
Mucosal Immunology
3rd Ed.
251
-266. Academic Press, San Diego.
35
Pleass, R. J., J. I. Dunlop, C. M. Anderson, J. M. Woof.
1999
. Identification of residues in the CH2/CH3 domain interface of IgA essential for interaction with the human Fcα receptor (FcαRI) CD89.
J. Biol. Chem.
274
:
23508
-23514.
36
Herr, A. B., C. L. White, C. Milburn, C. Wu, P. J. Bjorkman.
2003
. Bivalent binding of IgA1 to FcαRI suggests a mechanism for cytokine activation of IgA phagocytosis.
J. Mol. Biol.
327
:
645
-657.
37
Johansen, F.-E., R. Braathen, P. Brandzaeg.
2001
. The J chain is essential for polymeric Ig receptor-mediated epithelial transport of IgA.
J. Immunol.
167
:
5185
-5192.
38
Mestecky, J., I. Moro, M. A. Kerr, J. M. Woof.
2005
. Mucosal immunoglobulins. J. R. Mestecky, and J. Bienenstock, and M. E. Lamm, and L. Mayer, and J. R. McGhee, and J. R. , and W. Strober, eds.
Mucosal Immunology
3rd Ed.
153
-181. Academic Press, San Diego.
39
Davis, A. C., K. H. Roux, M. J. Shulman.
1988
. On the structure of polymeric IgM.
Eur. J. Immunol.
18
:
1001
-1008.