Crystal structures of the human IgD Fab reveal insights into CH1 domain diversity

Antibodies of the IgD isotype remain the least well characterised of the mammalian immunoglobulin isotypes. Here we report three-dimensional structures for the Fab region of IgD, based on four different crystal structures, at resolutions of 1.45-2.75 Å. These IgD Fab crystals provide the first high-resolution views of the unique Cδ1 domain. Structural comparisons identify regions of conformational diversity within the Cδ1 domain, as well as among the homologous domains of Cα1, Cγ1 and Cµ1. The IgD Fab structure also possesses a unique conformation of the upper hinge region, which may contribute to the overall disposition of the very long linker sequence between the Fab and Fc regions found in human IgD. Structural similarities observed between IgD and IgG, and differences with IgA and IgM, are consistent with predicted evolutionary relationships for the mammalian antibody isotypes.


Introduction
Described as an enigmatic antibody isotype, and previously thought to have evolved only relatively recently, immunoglobulin D (IgD) is now appreciated to be as ancient as IgM (Ohta, Flajnik, 2006;Chen, Cerutti, 2010). However, in contrast to IgM, the Fab, hinge and Fc regions of IgD display such remarkable structural diversity between species that it has been dubbed "the adaptive immune system's plaything" (Ohta, Flajnick, 2006). For example, each heavy chain in murine IgD-Fc comprises only a single constant domain, in human IgD-Fc there are two constant domains, while in the Fc regions of IgD from bony fish there are up to 13 constant domains. Human IgD has a 64 residue hinge between the Fabs and Fc region whereas IgD from cold blooded vertebrates lacks a hinge; a longer Fc region is instead suggested to provide flexibility. Furthermore, in many bony fish and artiodactyl species, the Cδ1 domain is even replaced by a Cμ1 or Cμ1-like domain (Ohta, Flajnik, 2006;Chen, Cerutti, 2011;Edholm et al, 2011).
Compared with other isotypes, the role of IgD in humans is less well understood. Membrane-bound IgD, as part of the B-cell receptor for antigen, appears to play a role in B-cell anergy in the periphery (Duty et al. 2009;Gutzeit et al., 2018). By contrast, a significant population of B-cells in the upper respiratory tract expresses membrane-bound IgD, where secreted IgD is suggested to play a role in mucosal immunity. Secreted IgD not only recognizes commensal bacteria that reside in the upper respiratory tract, but can bind to and activate basophils (Chen 2009;Edholm et al, 2011;Gutzeit et al., 2018). Moreover, commensal bacteria such as Moraxella catarrhalis have also been shown to activate B-cells in a superantigen-like manner through binding and cross-linking the IgD B-cell receptor by the Moraxella IgD-binding protein MID (Jendholm et al., 2009).
IgD is also the antibody isotype that is least well-characterized structurally. X-ray scattering studies have revealed that human IgD is predominantly T-shaped in solution, with a semiextended hinge (Sun et al., 2005). However, in contrast to other isotypes, there is no high resolution structural information available for any region of IgD.
Here we report four crystal structures of the human IgD Fab that provide the first high resolution views of IgD. Consistent with Fab structures for other isotypes, these structures reveal a range of elbow angles between the V H -V L and Cδ1-Cλ domain pairs, in addition to conformational diversity in particular Cδ1 domain loop regions. The Fab structures also reveal that the IgD upper hinge adopts a unique conformation among the human isotypes, which may play a role in determining its overall structure.

Protein expression
The vector encoding the recombinant human IgD Fab was based on a pVITRO plasmid that contained inserts encoding the heavy and light chains of an anti-Phl p 7 (grass pollen allergen) IgD antibody. This plasmid had been made according to a previously described method (Dodev et al., 2014) using a pVITRO plasmid encoding a human anti-Phl p 7 IgE antibody as a template and an IgD insert synthesized commercially. To generate the construct for the IgD Fab, the IgD heavy chain gene was mutated to incorporate a stop codon after the C-terminus of the Cδ1 domain using a QuikChange Lightning Site-Directed Mutagenesis kit (Agilent) according to the manufacturer's instructions. Recombinant human IgD Fab was expressed in stably transfected HEK293F cells cultured in spinner flasks in high glucose DMEM supplemented with 10% FCS, 120 U/mL penicillin, 100µg/mL streptomycin, 50µg/mL hygromycin and 2mM L-glutamine. The supernatant was harvested by centrifugation at 4000 x g, followed by filtration using a 0.45μm filter and the addition of 0.1% (w/v) sodium azide.

Protein purification
The IgD Fab was purified by affinity chromatography using the recombinant Phl p 7 antigen, which was produced as described previously (Mitropoulou et al., 2018) and then coupled to NHS-activated sepharose (Cytiva), according to the manufacturer's instructions. After the supernatant was loaded, the column was washed with PBS containing 0.1% (w/v) sodium azide and the Fab was eluted using 0.1M glycine pH 2.5. Fractions were pooled, concentrated and further purified by size-exclusion chromatography using a Superdex 200 Increase column that had been equilibrated with TBS and 0.1% (w/v) sodium azide.

Crystallization
For crystallization trials, the IgD Fab was buffer-exchanged into a buffer containing 100mM Tris pH 7.5 and 50mM NaCl and concentrated to 4.1mg/mL. All crystals were grown at 18°C in SWISSCI MRC 96 well plates using a reservoir volume of 100µL and drops comprising 100nL protein and 100nL reservoir solution. Fab 1 crystals were grown in 20% (w/v) PEG 3350 and 0.2M sodium phosphate, and were cryoprotected with 28% (w/v) PEG 3350 and 17% (v/v) ethylene glycol before flash-cooling in liquid nitrogen. Fab 2 crystals were grown in 0.1M BTP pH 6.5, 20% (w/v) PEG 3350 and 0.2M sodium nitrate, and were cryoprotected with 28% (w/v) PEG 3350 and 17% (v/v) ethylene glycol. Fab 3 crystals were grown in 16% (w/v) PEG 6000 and 0.1M tri-sodium citrate, and were cryoprotected with 25% (w/v) PEG 6000, 0.1M tri-sodium citrate and 17% (v/v) ethylene glycol. Fab 4 crystals were grown in 0.1M Tris pH 8.5, 18% (w/v) PEG 8000 and 0.2M lithium sulphate, and were cryoprotected with 0.1M Tris pH 8.6, 25% (w/v) PEG 8000 and 15% (v/v) ethylene glycol. the CCP4 suite (Winn et al., 2011;Evans and Murshudov, 2013). The Fab 3 structure was solved first by molecular replacement with PHASER (McCoy et al., 2007), using Fab protein atoms from Protein Data Bank (PDB) entry 5OTJ (Mitropoulou et al., 2018) as a search model. Structures Fab 1 , Fab 2 and Fab 4 were then solved by molecular replacement using protein atoms from the Fab 3 structure as a search model. The structures were refined with PHENIX (Liebschner et al., 2019) and manual model building was performed with Coot (Emsley et al., 2010). The quality of the structures was assessed with MolProbity (Williams et al., 2018). Data processing and refinement statistics are summarized in Table  1. Fab elbow angles were calculated using the RBOW program (Stanfield et al., 2006). Figures were produced with PyMOL (The PyMOL Molecular Graphics System, Version 1.1r1, Schrödinger, LLC).

Overall structure of the human IgD Fab
We solved four crystal structures of a human anti-Phl p 7 IgD Fab: Fab 1 (2.75Å resolution) solved in space group P 1 with two Fab molecules (Fab 1A and Fab 1B ) in the asymmetric Davies et al. Page 4 Mol Immunol. Author manuscript; available in PMC 2023 July 01.

Europe PMC Funders Author Manuscripts
Europe PMC Funders Author Manuscripts unit, Fab 2 and Fab 3 (1.55Å and 1.45Å resolution, respectively), solved in space group P 2 1 with one molecule each in the asymmetric unit, and Fab 4 (2.1Å resolution) solved in space group P 2 1 2 1 2 1 with one molecule in the asymmetric unit.
Like Fab fragments in other antibody isotypes (Stanfield et al., 2006), the crystal structures of the IgD Fab reveal a range of Fab elbow angles between the V H -V L and Cδ1-Cλ domain pairs, from 128.3° (Fab 3 ) to 228.3° (Fab 1B ) (Fig. 1A). The ABangle tool (Dunbar et al., 2013) was used to measure the set of five angles (HL, HC1, LC1, HC2 and LC2) and one distance ( Using the highest resolution structure (Fab 3 ) as the reference, Cα atoms from the V H domain of each structure were superposed with an RMSD range of 0.41-0.63Å (Fig. 1B).
The conformations of CDRH1-3 were similar, and the most significant difference between the structures was a conformational change about Gly66 in the C"-D loop, which altered the position of His65. Cα atoms from the V L domain were superposed with an RMSD range of 0.35-0.50Å (Fig. 1C). The most significant difference between the structures was a conformational change about Gly70 in the DE loop, which altered the positions of Ser69 and Thr71. Cα atoms from the Cδ1 domain were superposed with an RMSD range of 0.48-0.78Å (Figs. 1D and 1E), revealing modest conformational differences in the AB and EF loops, which are discussed in further detail below. The Cδ1 domain BC, CD, DE and FG loops adopted similar conformations in all five molecules. Cα atoms from the Cλ domain were superposed with an RMSD range of 0.21-0.63Å (Fig. 1F). The structures revealed some variation in the DE loop, which altered the positions of the Ser173 and Asn174 Cα atoms by ~1.8Å. Although some of the loop regions in some of the Fab structures form crystal packing interactions, the different conformations could not be attributed to differences in crystal contacts.

Structural diversity in the Cδ1 domain
Structural diversity in the Cδ1 domain is seen in the AB (residues 133-146) and EF (residues 193-202) loop regions, particularly around Arg137 (AB loop) and Trp198 (EF loop). In Fab 2 and Fab 4 , Trp198 packs against Leu147 and Leu195, the plane of the Arg137 guanidinium group is parallel to the plane of the Trp198 indole group, and the Arg137 side chain forms a hydrogen bond with the Trp198 main chain ( Fig. 2A). By contrast, in Fab 1B and Fab 3 , the plane of the Trp198 indole group is almost perpendicular to its position in Fab 2 and Fab 4 ; in Fab 1B and Fab 3 the Trp198 side chain occupies a similar space, but the overall orientation of the indole group differs. (Figs. 2B and 2C). The position of Trp198 in Fab 1B and Fab 3 would clash with the position of Arg137 observed in Fab 2 and Fab 4 (Fig. 2B). Instead, in Fab 1B and Fab 3 , the Arg137 side chain faces Pro222, and forms hydrogen bonds with Trp221 and Pro222 main chain atoms (Fig. 2C). In these molecules, Arg137 still contacts the Trp198 side chain, but the position of the Arg137 Cα atom is shifted by ~3.4Å compared with Fab 2 and Fab 4 (Fig. 2B). The plane of the Trp198 indole group is rotated in Fab 1A compared with its position in Fab 1B and Fab 3 , and Arg137 is partially disordered, but the Arg137 Cα atom is shifted as in Fab 1B and Fab 3 (Fig. 2B).
In Fab 1A , Fab 1B and Fab 3 , other AB loop residues also undergo conformational changes compared with Fab 2 and Fab 4 . For example, even though the disulfide bond with Cys216 from the λ light chain is unaffected, the position of the Cys136 Cα atom differs by up to 1.9A. His138 is disordered in Fab 3 but undergoes a substantial conformational change in Fab 1A and Fab 1B , with the position of its Cα atom shifted by ~3.9Å compared with that in Fab 2 and Fab 4 . Part of the AB loop is also disordered in Fab 1A , Fab 1B and Fab 3 (Fig. 2D); although the AB loop is ordered in Fab 2 and Fab 4 , the structure of the loop differs between these two Fab molecules ( Fig. 2A).
In the EF loop, the most significant difference between the five Fab molecules, apart from the conformational change involving Trp198, is the position adopted by Arg199 (Fig. 2D).
In Fab 2 , Fab 3 and Fab 4 , the Arg199 side chain extends away from the Cδ1 domain. By contrast, in Fab 1A and Fab 1B , the Arg199 side chain is oriented towards the surface of the Cδ1 domain.
Residues 198-206, which form part of the Cδ1 domain EF loop and β-strand F, are reported to form part of the binding site for the Moraxella catarrhalis IgD-binding protein MID (Samuelsson et al., 2006). However, it remains to be determined whether this structural diversity in the Cδ1 domain, and the MID binding site, has any functional implications.

Conformational diversity in human C H 1 domain loop regions
A set of crystal structures solved for other human isotypes (IgA1, IgG1-4 and IgM) was selected and analysed to assess their structural diversity. Like the IgD Cδ1 domain (Figs. 1D and 1E), the core of the immunoglobulin fold is conserved in Cα1 (subclass IgA1), Cγ1 (subclasses IgG1-IgG4) and Cμ1 domains. Conformational diversity in the C H 1 domain loops was observed in each isotype and subclass as follows: in the AB and DE loops in IgA1

The Cδ1 domain BC, DE and FG loop conformations show structural similarities between IgD and IgG and differences with IgA and IgM
The precise evolutionary relationships between the five human antibody isotypes are still not fully understood. However, IgD is suggested to have given rise to the IgY isotype (in amphibians, birds and reptiles), from which mammalian IgG and IgE are thought to have evolved (Estevez et al., 2016;Hordvik et al., 1999;Zhang et al., 2017). IgM is the ancestor of IgA, which evolved along a different path (Zhang et al., 2017). Consistent with these proposed evolutionary relationships, the BC, DE and FG loops regions of IgD are similar to IgG but different from IgA and IgM. Davies et al. Page 6 Mol Immunol. Author manuscript; available in PMC 2023 July 01.
The C H 1 BC loop comprises residues 153-158 (Figs. 3 and 4; residue numbering is according to the IgD Fab structures). In all structures that were compared from each set (IgA1, IgD, IgG1-4 and IgM), a proline residue is found at position 156, a cis peptide bond is found between residue 155 and Pro156 and the loop conformation for these residues is similar (Figs. 5A and 5B). The Cα1 (IgA1) and Cγ1 (IgG1-4) BC loops contain a second proline residue at position 158. The Cα1 domain contains a single-residue insertion (glutamic acid) between residues 157 and 158 and, even though the overall position of residue 158 is altered due to this insertion, a cis peptide bond is found between residue 157 and Pro158 in both Cα1 and Cγ1 domains (Figs. 5A-C). On the other hand, the Cμ1 BC loop contains a serine residue at position 158 and here the loop adopts a different conformation to that found in the Cα1 and Cγ1 domains, with a trans peptide bond between Asp157 and Ser158 (Fig. 5A). Like Cμ1, a serine residue is found at position 158 in the Cδ1 domain. However, in contrast to the Cμ1 BC loop, the Cδ1 BC loop contains a cis peptide bond between Thr157 and Ser158, and the loop adopts the same conformation as in the Cγ1 domain in all IgG subclasses (Figs. 5B and 5C). That IgD is more IgG-like could suggest that a second cis peptide bond is important to maintain the structure of the C H 1 BC loop, which appears to be more conformationally restricted when another proline residue is present.  (Fig. 5E).
The C H 1 FG loop comprises residues 209-212 (Figs. 3 and 4). Like the DE loop, the FG loop contains a conserved residue in all isotypes, a histidine at position 209 and, like the conserved tyrosine residue in the DE loop, His209 also packs against the BC loop (Fig. 5D). Within the short FG loop, IgG and IgM contain proline residues at position 211 and 210, respectively. In IgA1, a proline is found adjacent to the loop, at position 213. By contrast, a serine residue is found at this position in IgA2. A proline residue is likewise absent in IgD. Like the DE loop, the conformation of the FG loop shows structural diversity in different isotypes. However, here again the Cδ1 FG loop adopts a similar conformation to the Cγ1 FG loop in all IgG subclasses (Fig. 5F).
Packing interactions with the conserved residues Tyr185 and His209, and the presence of a proline residue at position 156, could restrict the conformation of the C H 1 BC loop. The BC loop is close to the linker between the V H and C H 1 domains, with the implication that substantial conformational diversity within the BC loop could reduce the range of Fab elbow angles available to the V H -V L and Cδ1-C L domain pairs.

In IgD, the junction between the C H 1 domain and upper hinge adopts a unique structure
In IgA1, IgD and IgG1-4, the junction between the C H 1 domain and upper hinge is located within β-strand G. In IgM, the sequence of the Fab extends just beyond this strand (Fig.  4). When a set of crystal structures was compared for each isotype, it was found that the structure of the C-terminal region of the C H 1 domain and junction with the hinge was similar within the set. Furthermore, the structure of this region was conserved between IgA1 and IgG1-4, and was similar in IgM (Fig. 6A). By contrast, the structure of the C-terminal region of the C H 1 domain and junction with the hinge differed substantially in IgD (Fig.  6A).
In all five independent views of the IgD Fab the direction of the main chain is altered for IgD, compared with the other isotypes, at residue Phe219 at the C-terminus of the Cδ1 domain, and a kink is introduced into β-strand G (Fig. 6A). As a result, Arg220, the first residue of the IgD upper hinge, faces away from the Cδ1 domain instead of packing against the C H 1 domain like the structurally conserved residue Val220 found in IgA, IgG and IgM. Instead, in IgD, Trp221 from the upper hinge occupies a structurally equivalent position to Val220 and forms more substantial interactions with conserved C H 1 domain residues: Trp221 is sandwiched between Pro131 and Leu147 (Ile147 in IgA1 and Val147 in IgM), and contacts Trp163, in addition to Cys149 and Cys205 that form a disulfide bond (Figs. 6A and 6B). Furthermore, Pro225 from the upper hinge packs against Ser127 from the λ light chain (Fig. 6C). A similar interaction could also form between Pro225 and a κ light chain, in which Ser127 is replaced by aspartic acid.
Trp221 is close to a region of the Cδ1 domain that is observed in multiple conformations. In Fab 1B and Fab 3 , the Arg137 side chain forms a hydrogen bond with the Trp221 main chain; this interaction is absent in Fab 1A , Fab 2 and Fab 4 as Arg137 either adopts a different conformation or is partially disordered. Thus, the conformation adopted by Trp221 appears to be independent of conformational changes in the Cδ1 AB loop (Figs. 2B and 2C). Furthermore, the kink in β-strand G at the C-terminus of the Cδ1 domain (Phe219) appears to be unique to IgD; it was not observed in the set of structures of the other human isotypes.
In intact IgG1 and IgG4 structures (Blech et al., 2019;Saphire et al., 2001;Scapin et al., 2015), the Val220 main chain at the C-terminus of the Cγ1 domain adopts a conserved conformation, and the structure of the upper hinge differs substantially after this residue (Fig. 6C). By contrast, the conformational change about Phe219, and the resulting packing interactions formed between Trp221 from the upper hinge and the Cδ1 domain, and Pro225 and the λ light chain, could restrict the flexibility and conformation of the upper hinge in IgD.
Of the human antibody isotypes, IgG3 and IgD have the longest hinge regions with 62 and 64 residues, respectively. The IgG3 hinge contains eleven cysteine residues that form inter-chain disulfide bonds. Solution-based biophysical techniques have shown that IgG3, and by implication the IgG3 hinge, adopts an extended, elongated structure in solution (Spiteri et al., 2021). By contrast, the IgD hinge contains a single cysteine residue and solution-based techniques have shown that IgD is predominantly T-shaped, yet flexible, with Davies et al. Page 8 Mol Immunol. Author manuscript; available in PMC 2023 July 01.

Europe PMC Funders Author Manuscripts
a semi-extended hinge (Sun et al., 2005). Although the hinge regions of these two isotypes are markedly different, the unique conformation of the upper hinge in IgD, and the packing interactions with the Cδ1 domain and λ light chain, could affect the overall disposition of the Fab fragments with respect to the Fc region: a conformationally restricted upper hinge could contribute to the predominant T-shape of human IgD.
Trp221 and Pro225 from the upper hinge are conserved in primate species including chimpanzee, baboon, mangabey, rhesus macaque, cynomolgous macaque (Rogers et al., 2006) as well as orangutan and gibbon (Fig. S1). It remains to be determined whether IgD antibodies in which Trp221 and Pro225 are conserved adopt the predominantly T-shaped structure observed for human IgD.

Conclusion
In summary, we solved crystal structures of the IgD Fab, providing five independent views of the structure of the Cδ1 domain. These structures reveal conformational diversity in specific regions of the Cδ1 domain, particularly in the AB and EF loops. Comparisons with C H 1 domain structures of other isotypes provide support for the idea of an evolutionary link between IgD and IgG. Among the human isotypes, the structures also reveal features unique to IgD, such as the upper hinge region, that could play a role in determining its overall structure.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.  (A) Cδ1 domain AB loop structure in Fab 2 (green) and Fab 4 (yellow). The plane of the Arg137 guanidinium group is parallel to the plane of the Trp198 indole group. Arg137 forms a hydrogen bond (depicted by a black line) with the Trp198 main chain. (B) In Fab 1B (blue) and Fab 3 (grey), the plane of the Trp198 indole group is almost perpendicular to its position in Fab 2 (green) and Fab 4 (yellow). In Fab 1A (purple), Fab 1B and Fab 3 , the position of the R137 Cα atom is shifted by ~3.4Å (depicted by a black line) compared with its position in Fab 2 and Fab 4 . (C) In Fab 1B (blue) and Fab 3 (grey), the Arg137 side chain forms hydrogen  Sequences were obtained from UniProt (The UniProt Consortium, 2021) with the following accession numbers: IgA1, P01876; IgA2, P01877; IgD, P01880; IgE, P01854; IgG1, P01857; IgG2, P01859; IgG3, P01860; IgG4, P01861; IgM, P01871. For all sequences, the numbering system from the IgD Fab structures has been used. The sequence alignment was performed using Clustal Omega (Madeira et al., 2022). Small manual adjustments were made to the sequence alignment after structural comparisons were performed to ensure better agreement between the alignment and structures. Highly conserved amino acids of the 'central pin' residues of Ig-domains (Williams and Barclay, 1988) are in bold and coloured black. Other noteworthy, conserved residues in C H 1 domains are coloured blue. The IgM Cμ1 FG loop has two fewer residues compared with IgD and IgG and a two-residue gap was inserted into the IgM sequence after position 212. In the Clustal Omega sequence alignment, a gap that had been inserted in the IgM Cμ1 β-strand G at position 217 was removed. The IgA Cα1 FG loop has one fewer residue compared with IgD and IgG and a single residue gap was inserted into the IgA sequence after position 213. Davies et al. Page 19 Mol Immunol. Author manuscript; available in PMC 2023 July 01. (A) C H 1 BC loop structure in IgA1 (PDB entry 3M8O, yellow) and IgM (PDB entry 1DN0, purple). The cis peptide bond between residue 155 and Pro156 is shown. In IgA1, a cis peptide bond is also found between the glutamic acid insertion and Pro158. As the loop structure is similar within each isotype, a single representative structure has been shown.
(B) C H 1 BC loop structure in IgG1 (PDB entry 6WFY, blue) and IgD (Fab 3 , grey). The cis peptide bond between residue 155 and Pro156 is shown. In both IgG1 and IgD, a cis peptide bond is found between residues 157 and 158. As the overall loop structure is similar within each IgG subclass, and between IgD and IgG isotypes, one representative structure has been shown for each of the Cδ1 and Cγ1 domains. is sandwiched between Pro131 and Leu147 and contacts Trp163, Cys149 and Cys205. (C) In intact IgG1 (PDB entry 1HZH, blue and yellow) and IgG4 (PDB entry 5DK3, green and purple; 6GFE, cyan and salmon) structures, the position of Val220 is conserved and the conformation of the hinge differs substantially only after this residue. In IgD (Fab 3 , grey), the junction between the Cδ1 domain and upper hinge adopts a different structure. In addition to the packing interactions formed by Trp221, Pro225 from the IgD upper hinge contacts Ser127 from the light chain. For IgD, the Fab 3 structure is shown and the surface representations for the heavy and light chains are colored light grey and dark grey, respectively. For all structures, β-strand G Cα atoms are shown. The Cδ1 and Vλ domains have been shown as a surface representation. For clarity, part of the Cδ1 AB and EF loops have not been shown. Data processing and refinement statistics.