Research Article

Structure of the secretory immunoglobulin A core

See allHide authors and affiliations

Science  28 Feb 2020:
Vol. 367, Issue 6481, pp. 1008-1014
DOI: 10.1126/science.aaz5807

Hefty structures of IgA and IgM complexes

Immunoglobulin M (IgM) and IgA are antibody isotypes that can form higher-order secretory complexes (sIgM and sIgA), which allows them to effectively bind and neutralize antigens with low-affinity repetitive epitopes, such as those found on the surface of many bacteria and viruses. The assembly and transport of these molecules is also dependent on the joining chain (J-chain) and the polymeric immunoglobulin receptor (pIgR) secretory component (SC). The architecture of these complex, multimeric structures has remained elusive. Li et al. resolved cryo–electron microscopy structures of the sIgM-Fc pentamer in complex with the J-chain and SC. Using similar techniques, Kumar et al. visualized dimeric, tetrameric, and pentameric structures of secretory sIgA-Fc interacting with the J-chain and SC. Both groups report highly similar mechanisms wherein the J-chain serves as a template for antibody oligomerization. An unanticipated, amyloid-like assembly of the oligomerized structure is present in both cases, with the J-chain conferring asymmetry for pIgR binding and transcytosis. These studies may inform structure-based engineering of these molecules for future therapeutic purposes.

Science, this issue p. 1014, p. 1008

Abstract

Secretory immunoglobulin A (sIgA) represents the immune system’s first line of defense against mucosal pathogens. IgAs are transported across the epithelium, as dimers and higher-order polymers, by the polymeric immunoglobulin receptor (pIgR). Upon reaching the luminal side, sIgAs mediate host protection and pathogen neutralization. In recent years, an increasing amount of attention has been given to IgA as a novel therapeutic antibody. However, despite extensive studies, sIgA structures have remained elusive. Here, we determine the atomic resolution structures of dimeric, tetrameric, and pentameric IgA-Fc linked by the joining chain (JC) and in complex with the secretory component of the pIgR. We suggest a mechanism in which the JC templates IgA oligomerization and imparts asymmetry for pIgR binding and transcytosis. This framework will inform the design of future IgA-based therapeutics.

Secreted polymeric immunoglobulins A and M (pIgA and pIgM, respectively) play vital roles in protecting the ~400 m2 of human mucosa from invasion by pathogens. IgA and IgM contain 18-residue tailpiece extensions on their heavy chains that bestow polymer-forming capabilities (1). IgM can independently oligomerize to form hexamers, whereas IgA polymerization requires the 137-residue joining chain (JC), a protein with no known structural homologs (2). The predominant IgA oligomer is a dimer, though a smaller fraction of higher-order polymers up to pentamers have been described (1). At the core of pIgA is a JC-clasped, tail-to-tail dimer stabilized by two disulfides between the JC and one tailpiece of each monomer (3). Additional monomers are linked to this precursor via tailpiece-to-tailpiece–mediated disulfides (3, 4). Although less abundant than the dimer, higher-order polymers display better neutralizing capabilities, specifically for low-affinity antigens (5).

The pIgA that is synthesized by mucosal plasma cells undergoes transcytosis, crossing the epithelium to reach the external secretions and perform its protective function. To initiate this process, pIgA binds the ectodomain of basolaterally expressed pIg receptor (pIgR), an interaction that requires the JC and is stabilized through a disulfide bond between the receptor and one Fc (6). Once bound, the receptor and pIgA undergo transcytosis through the epithelial cell to the mucosa. Upon transcytosis, proteolytic cleavage by unidentified protease(s) releases the secretory component (SC) of the receptor, which remains covalently attached to pIgA to form secretory IgA (sIgA) (7, 8). In this mature form, sIgA performs its antimicrobial, neutralization, and protective functions (8).

In recent years, extensive efforts have been made to develop vaccines that induce immunity via the mucosal route by eliciting sIgA responses (1). In the field of biotechnology, a major focus has been to engineer antibodies to target disease-relevant tissues, and therapeutic IgAs may allow delivery to mucosal tissues inaccessible to traditional IgG-based therapeutics (1). Thus, a detailed structural characterization of sIgA is critical for the development of future therapeutics and vaccines.

In this study, we use cryo–electron microscopy (cryo-EM) to determine the architecture of sIgA-Fc dimers, tetramers, and pentamers at atomic resolution. We propose a mechanism of JC-templated IgA oligomerization and describe a model for pIgR recognition. These findings will inform the design of unexplored IgA-based therapeutics with the potential for tissue-specific targeting and mucosal delivery.

Reconstitution of sIgA complexes

Previous cryo-EM studies of pIgA indicated that these molecules are refractory to high-resolution structure determination due to the inherent flexibility of the hinge regions as well as the tailpieces interconnecting monomers (9). We therefore reasoned that truncation of IgA to the Fc core and the addition of the SC would result in the formation of a more rigid molecule amenable to high-resolution structure determination. In humans, there are two IgA isotypes, IgA1 and IgA2. Both isotypes can form dimers and higher-order oligomers, though there is a propensity for recombinant IgA1 to form dimers and for IgA2m2 to form higher-order oligomers (9). Therefore, we coexpressed human IgA1 or IgA2m2 Fc with JC to form dimers or tetramers and pentamers, respectively (figs. S1 and S2), and subsequently assembled them with the SC. Single-particle cryo-EM was used to determine the structures of the resulting dimeric, tetrameric (two classes), and pentameric sIgAs at 2.9-, 3.0-, 2.9-, and 3.0-Å resolution [Fourier shell correlation (FSC) = 0.143], respectively (figs. S1 to S3 and table S1). Two classes were obtained for the tetramer; the Fcs were slightly bent toward each other in class 1 and were more planar in class 2 (fig. S2). The difference was subtle and did not affect the pIgA/SC interaction interface. The local resolution was highest at the central core, which contains the key interaction interfaces (fig. S4).

Overall architecture of sIgA

sIgA1 adopted a tail-to-tail planar dimer with the two Fcs positioned at an ~110° angle and held tightly by the JC, which functioned as a clasp (Fig. 1). The SC made extensive contacts with both Fcs and the JC, binding diagonally across the ~50° gap between the two monomers. The structure of the core sIgA1 dimer was relatively unchanged in the sIgA2m2 tetramer and pentamer, consistent with the 94% sequence identity between the Fcs of the two isotypes (fig. S5). Therefore higher-order polymers are assembled through incorporation of two or three additional Fcs in-plane to the original dimer (Fig. 2). The pentamer adopted an asymmetric pentagon containing an ~50° gap between Fc1 and Fc2 (Fig. 2), reminiscent of the JC-containing IgM pentamer (10). The SC bound at the gap, cross-bridging two Fcs in a manner similar to that of the IgM-binding apoptosis inhibitor of macrophage protein (10), suggesting a common binding mode.

Fig. 1 Cryo-EM structure of dimeric sIgA1.

(A) Top, back, and front view schematics of subunit arrangements in the dimer. (B) Cryo-EM reconstruction of dimeric sIgA1. Transparent maps overlaid with the model are shown.

Fig. 2 Cryo-EM structures of tetrameric and pentameric sIgA2m2.

(A) Top and back view schematics of subunit arrangements in the tetramer. (B) Cryo-EM reconstruction of tetrameric sIgA2m2. Transparent maps overlaid with the model are shown. (C) Same as in (A) but for pentameric sIgA2m2. (D) Same as in (B) but for pentameric sIgA2m2.

Structure of the JC

We used the 2.9-Å map of the sIgA1 dimer (fig. S6) to de novo build a model containing all but six residues of the 137–amino acid JC (residues 5 to 94 and 97 to 137), as well as the tailpieces of each Fc. The JC was almost entirely composed of β-sheets and loops (Fig. 3, A and B), consistent with secondary structure predictions and circular dichroism studies (11). Previous biochemical studies identified intramolecular disulfides between JCC13 and JCC101, JCC72 and JCC92, and JCC109 and JCC134, and intermolecular disulfides between JCC15 and Fc2C471 and JCC69 and Fc1C471 (3). Our structure confirmed these disulfides (Fig. 3C) and additionally revealed a much larger network of noncovalent interactions. The JC intricately joined the two Fcs of the dimer core though its three β-hairpins and four-stranded, twisted β-sandwich. It lay at an angle bridging the top of Fc2 (defined as the SC-binding face) through β-hairpins 1 and 2, to the bottom of Fc1 through β-hairpin 3, creating an asymmetric dimer (Fig. 3A). The JC β-hairpins made predominantly hydrophobic interactions with the two Fcs (Fig. 3D). β-hairpins 2 and 3 interacted with the same region of Fc2 and Fc1, respectively, with their β-strands packing against FcM433 and FcF443 and their loops contacting FcL258 (Fig. 3D). This same Fc-interaction site is also exploited by the human Fcα receptor I (FcαRI) and the Staphylococcus aureus SSL7 toxin (12, 13), suggesting a hot spot for Fcα recognition (fig. S7). The dimer was further stabilized through interaction of the Fc tailpieces with the JC, where Fc1 and Fc2 contributed two parallel β-strands each, extending the bottom and top sheets, respectively, of the twisted JC β-sandwich (Fig. 4A). Despite its small size, the JC created an extensive, ~3600-Å2 interface with the Fcs in the dimer, underscoring its apropos title of “joining” chain.

Fig. 3 Structure of the JC.

(A) Top, front, and bottom views of dimeric IgA1 with the JC colored rainbow from the N terminus (blue; model starts at residue 5) to the C terminus (red). The SC is omitted for clarity. (B) Topology diagram of the JC with both intramolecular and intermolecular disulfide bonds shown. (C) Magnified views of box c in (A) showing JC intramolecular (left) and intermolecular (right) disulfide bonds. (D) Magnified views of β-hairpins 1, 2, and 3 from boxes a, b, and d, respectively, in (A) and the interactions they make with the Fcs.

Fig. 4 Mechanism of IgA oligomerization.

Top view of (A) dimeric IgA1, (B) tetrameric IgA2m2, and (C) pentameric IgA2m2. The SC is omitted for clarity. (D) Magnification of the left boxed region in (C) highlighting Fc–Fc contacts at the Cα2–Cα3 interface. (E) Magnification of the central boxed region in (C) highlighting the JC β-sandwich extension shown as an open book preparation of the Fc tailpieces with repeating internal residues labeled and side chains shown as sticks. (F) Magnification of the central boxed region in (C) showing repeating external residues of the β-sandwich. (G) Cross-section of the extended β-sandwich of the pentamer with Fc5 omitted for clarity.

Mechanism of IgA oligomerization

The JC-stabilized dimer core acted as a building block for larger polymers, consistent with previous studies showing dissociation of tetrameric IgA into a JC-containing dimer and two monomers upon mild reduction (4). In the tetramer, an additional set of two parallel β-strands, originating from the Fc3 and Fc4 tailpieces, continued the extension of the twisted β-sandwich, with Fc3 adding to the bottom sheet and Fc4 to the top (Fig. 4B). However, in the pentamer, the β-sheet extension continued, with Fc5 adding one tailpiece as a parallel strand to the top sheet and the other tailpiece as a parallel strand to the bottom sheet to cap the β-sandwich (Fig. 4C). Fc–Fc contacts occurred at the Cα2–Cα3 interface in both the tetramer and pentamer using residues similarly involved in JC β-hairpin and FcαRI binding (Fig. 4D and fig. S7). The arrangement of Fc tailpieces into the JC β-sandwich resulted in a striking repetitive pattern of equivalent residues packed at the core of pIgA, running parallel within each β-sheet and antiparallel within the β-sandwich (Fig. 4, E to G). In this arrangement, polar residues were solvent exposed, whereas hydrophobic residues were buried at the β-sandwich interface (Fig. 4, E to G). Thus, we propose a model for IgA oligomerization in which the JC acts as a template for the incorporation of IgA monomers via their tailpieces. This creates a molecular zipper of hydrophobic side chains, which stabilizes the polymer (Fig. 4G).

Structure of the SC and interaction with pIgA

The SC comprises five Ig-like domains, of which D1 is both necessary and sufficient for pIgA binding, whereas D2 to D5 provide affinity enhancement (14, 15). Previous structural analysis of the SC showed that in the absence of pIgA, the SC adopts a closed conformation stabilized by the interaction among D1, D4, and D5 (fig. S8A) (15). Our structures revealed that the SC underwent a large conformational change and interacted through domains D1 and D5 with both Fcs of the IgA dimer core and the JC (Fig. 5A and fig. S8A). The asymmetry imparted by the JC allowed one-to-one binding of the asymmetric SC to the dimer. The interaction with D1 was noncovalent and involved all three complementarity determining region (CDR)–like loops, whereas the interaction with D5 was mediated by a single disulfide bond (Fig. 5). Although D2 to D4 did not make direct contacts with pIgA, they provided the correct spacing to allow the interactions of D1 and D5 with the dimer. D2 to D5 were arranged in a head-to-tail manner, whereas a sharp 180° turn in the D1–D2 linker allowed D1 to position its three CDRs to bind pIgA (Fig. 5A). The described kinked SC conformation was stabilized via an aromatic-rich D1–D3 interface (Fig. 5A). Approximately one-third of the ~1055 Å2 of D1 buried surface area resulted from interactions with the JC C terminus, consistent with the requirement of the JC for pIgR binding and transcytosis (16).

Fig. 5 Recognition of pIgA by the SC.

(A) sIgA1 dimer with IgA shown as a transparent surface with cartoon representations underneath. Domains D1 to D5 of the SC are color coded as indicated with D1 CDR1, CDR2, and CDR3 highlighted in green. Magnification of the aromatic-rich D1–D3 interface shown in the inset. (B) Interaction of SC D5 and Fc2 with inset magnifying the D5C468–Fc2C311 disulfide bond. (C) Relative positioning of D1 CDR1 (box a), CDR2 (box b), and CDR3 (box c) and their interactions with pIgA. Domains D2 to D5 are omitted for clarity. (D) Magnification of box a from (C) to show D1 CDR1 interactions with Fc1 and JC. Side chains are shown as sticks and polar interactions as dashed yellow lines. (E) Magnification of box b from (C) showing interactions among D1 CDR1, CDR2, CDR3, and Fc1. (F) Magnification of box c from (C) showing interactions among D1 CDR3, Fc1, Fc2, and JC.

Discussion

This work presents the first atomic-level resolution structures of pIgA bound to the elusive JC and sheds light on the mechanism involved in higher-order polymer formation. It also presents for the first time the architecture of sIgA and describes at the atomic level the molecular basis for recognition of pIgA by the SC of the pIgR. Our structures show that upon pIgA binding, the SC undergoes a large conformational change. In the closed SC conformation, the majority of the Fc1-binding D1 residues are at the D1–D4–D5 interface and at least partially buried, whereas the JC and Fc2 tailpiece-binding residues are solvent accessible (Fig. 6, A and B). Expanding on the hypothesis of Stadtmueller et al. (15), we propose a model in which the closed pIgR conformation solvent-accessible residues make initial contact with the JC and Fc2 tailpiece. This leads to a conformational change in the receptor that breaks the D1–D4–D5 interface and exposes the remaining contact residues in D1 to bind Fc1 (Fig. 6C). The open pIgR conformation is stabilized through a new D1–D3 interface that creates optimal spacing between D1 and D5 to enable secondary site binding to Fc2. Reduction of the D5C468–D5C502 intramolecular disulfide frees D5C468 to form the intermolecular disulfide with Fc2AC311, locking pIgR into a bent, pIgA-bound conformation (fig. S8). When pIgR is proteolytically cleaved upon transcytosis, SC remains covalently attached to pIgA, leading to the release of sIgA at the mucosa.

Fig. 6 Model for pIgA recognition by pIgR.

(A) Structure of the free SC (PDB 5D4K) in a closed conformation stabilized by a D1–D4–D5 interface. CDR residues of D1 involved in pIgA binding are colored green and their side chains are shown as sticks. (B) Magnification of the box in (A) highlighting pIgA contact residues of D1 that are solvent accessible or buried. (C) Model of the conformational change in pIgR that converts it from the closed, unbound state (left, free SC) to the open, pIgA-bound state (right, dimer structure aligned to D1 of free SC).

Both the SC and pIgA are modified with N-linked glycans, which not only facilitate proper protein folding, but also mediate the attachment and neutralization of pathogens (1, 8). Our structures revealed ordered N-linked glycans on JCN49, FcN337, FcN459, SCN65, SCN72, SCN168, SCN403, SCN451, and SCN481, which are not in contact with the protein and are placed away from any SC–pIgA interaction interfaces (fig. S9). These glycosylation sites could be relevant in facilitating host and pathogen lectin binding.

Extensive mutational analysis has previously been performed on both pIgA and the pIgR, assessing both binding and transcytosis. As seen in our structures, the C-terminal 25 residues of the JC are critical for SC binding, whereas JCC15 and JCC69 are important for efficient transcytosis, likely due to stabilization of the pIgA assembly through their covalent bonds to the Fc tailpieces (16). Mutagenesis of the human SC D1 residues N30, R31, H32, and R34 in the highly homologous rabbit SC either significantly reduced or abolished SC binding to human pIgA (17), consistent with the identification of these as key interaction residues in our structures. Additionally, swapping of the rabbit SC D1 CDR2 or CDR3 loops with the corresponding structurally similar yet sequence diverse CDRs from D2 also abolished binding (17), consistent with both CDRs making important contacts to pIgA in our structures. Mutation of FcC311, FcP440–F443, or deletion of the Fc tailpieces results in a significant impairment in transcytosis (18), a functional consequence highlighting the important roles these regions play in SC binding, as observed in our structures. Taken together, these mutagenesis studies underscore the functional importance of the interactions described in our structures.

The coreceptor Mac-1 (CD11b/CD18) is required to bind and activate FcαRI signaling by sIgA but not dimeric IgA, possibly due to steric hindrance by SC (12, 19). Our sIgA dimer structure showed that JC binding to the Fcs overlapped with two of the four FcαRI-binding sites (fig. S7). Similarly, in the tetramer, only two sites were available because the Fc–Fc-mediated contacts also occurred at the Cα2–Cα3 interface. We saw no evidence of additional steric hindrance between the SC and the two accessible FcαRI-binding sites in the dimer or tetramer. In the pentamer, no FcαRI-binding sites were accessible, suggesting that pentameric IgA cannot bind the receptor. Because the oligomeric state of sIgA used in the Mac-1 studies is unknown (19), it is difficult to interpret the requirement for Mac-1 in FcαRI signaling.

Although many recombinant IgGs have been approved for therapeutic use in fields such as oncology, immunology, and ophthalmology, an IgA-based therapeutic has yet to be approved, likely in part due to the short serum half-life of this isotype (9). The structural characterization of sIgA provided here opens the possibility for engineering half-life extension properties into pIgA while maintaining its fundamental mucosal tissue–targeting properties. Furthermore, using the structural information as a framework, recombinant sIgA could be designed for potential oral delivery of a therapeutic antibody because the SC is known to provide stability and protection to pIgA from proteolytic degradation (8). Our studies also raise the question of whether the JC-containing pentameric IgM is formed through a similar oligomerization mechanism and if pIgR binding is analogous. High sequence conservation between IgA and IgM tailpieces (fig. S5B) suggests a shared mechanism of polymerization. However, the presence of an additional constant domain, Cμ4, in IgM may require a modified mode of pIgR recognition. Additional structural studies will be needed to address these fundamentally important questions in mucosal immunity.

Supplementary Materials

science.sciencemag.org/content/367/6481/1008/suppl/DC1

Materials and Methods

Figs. S1 to S9

Table S1

References (2034)

References and Notes

Acknowledgments: We thank N. Lombana for ordering constructs, Research Materials Group for protein expression, F. Farahi and A. Gill for small-scale purification, A. Estevez and M. Kschonsak for data collection and discussions, and A. Rohou and B. Barad for assistance with data processing. Funding: This work was funded by Genentech, Inc. Author contributions: N.K.: conceptualization; data curation; formal analysis; investigation; validation; visualization; writing original draft; writing, review, and editing; C.P.A.: data curation; investigation; supervision; C.C.: supervision; writing original draft; writing, review, and editing; M.L.M: conceptualization; investigation; formal analysis; project administration; supervision; validation; visualization; writing original draft; writing, review, and editing; Competing interests: All authors are employees of Genentech, Inc., a member of the Roche Group, and may hold stock and options. C.C. and M.L.M. are inventors on unpublished patent applications that relate to IgA antibodies and IgG-IgA fusion molecules and methods of making and using them. Data and materials availability: All maps and coordinate models have been deposited to the Worldwide Protein Data Bank: sIgA1 dimer (EMD-20749, PDB ID 6UE7); sIgA2m2 tetramer class 1 (EMD-20750, PDB ID 6UE8); sIgA2m2 tetramer class 2 (EMD-20751, PDB ID 6UE9); sIgA2m2 pentamer (EMD-20752, PDB ID 6UEA). Materials used in this study are available under a material transfer agreement from Genentech, Inc.

Stay Connected to Science

Navigate This Article