**1. Introduction**

Out of all the antibody classes, the most abundant isoform in the human body is IgA [1]. It is the predominant antibody found in the mucosa—a vast extracellular environment constantly exposed to antigens from both pathogens and commensal bacteria where it is secreted in the form of secretory IgA (SIgA). This is a multi-component molecule comprised of two IgA monomers linked together by a joining chain (J chain) and covalently attached to the secretory component (SC) [2]. The secretory component wraps around the antibody complex and confers resistance to proteolytic degradation, along with protection in low pH environments. The ability to form this stable structure is exclusive to mucosal antibodies IgA and IgM and presents a unique advantage to

traditional IgG therapy which currently dominates the immunotherapy market [3]. Despite this, there are currently no licenced SIgA products available for use.

Being adapted to mucosal secretions means SIgA has an intrinsic advantage as an oral therapeutic targeting mucosal pathogens compared to other antibody classes. In airway infections, SIgA has been shown to neutralise influenza virus and prevent virus-induced pathology in the upper respiratory tract, at times better than IgG [4, 5]. This is likely to be due to its ability to bind antigens with high avidity on the mucosal surface and prevent adherence to the epithelium, a method called immune exclusion [3]. SIgA is anchored to the mucosal surface by interacting with mucins (glycoproteins which make up the mucus layer) via the antibody's secretory component, ensuring a layer of protection against potential pathogens [6]. More recently, the respiratory application of SIgA therapy was highlighted against COVID-19, with a monoclonal SIgA, but not IgG, potently neutralising SARS-CoV-2 virus [7]. This was attributed to the increased avidity of SIgA arising from its polymeric and flexible Fab regions.

SIgA is also a key immune component of gut mucosa, where it is secreted as the first line of defence [8]. This is an exceedingly complex environment where part of the interplay between host and gut microbiota can be disrupted by intestinal pathogens to establish infection (such as *Escherichia coli*, *Campylobacter jejuni* and *Clostridium difficile*) [9]. There is a global need to engineer therapeutics for diarrhoeal diseases which are a leading cause of death in developing countries, the majority being of young children [10]. In the search for therapeutics against such infections, SIgA has emerged as an attractive candidate. Administered via the oral route, SIgA against enterotoxigenic *E. coli* (ETEC) reduced instances of diarrheal disease by hindering bacterial adhesion to the gut lumen in a nonhuman primate model [11]. When compared to an IgG counterpart, SIgA neutralised a *C. difficile* toxin up to 100 times more effectively, although this was not replicated by all SIgA subtypes [12]. SIgA has also been demonstrated to be more stable in a low pH simulated intestinal fluid environment than the IgG1 variant [13]. The protective barrier arising from the interaction between SIgA and the mucus is effective at preventing bacteria from anchoring and colonising the gut lumen—the neutralised bacteria are then unable to cause disease.

With functional efficacy and applications against a wide range of mucosal pathogens, it is understandable why SIgA emerges as an attractive candidate in the growing field of antibody therapy. However, there are a host of factors which must be addressed before this antibody type can fulfil its therapeutic potential. Issues relating to SIgA production, purification and glycosylation must be addressed—the good news is there are already extensive efforts to do so.

## **2. Structure and assembly**

SIgA is a complex molecule which relies on all 4 components assembling to form a functional antibody. This can be an issue when producing SIgA *in vitro* as partially assembled forms will also be produced (such as monomers and dimers). Furthermore, there are 2 subtypes of IgA which mainly differ in their hinge region: IgA1 with a long O-glycosylated hinge and IgA2 with a short non-glycosylated hinge [14]. Hinge length is associated with differences in flexibility and the ability to reach more distant epitopes at the expense of increased protease-mediated breakdown susceptibility [15]. As a result, the IgA subtype used for a specific therapeutic may be dependent on its application—for example, a potential SIgA therapeutic administered to the bacteria-rich gut mucosa might benefit from being IgA2 in order to evade hinge

degradation by bacterial proteases. On the other hand, IgA1 may offer high stability due to the presence of strong covalent bonds between its heavy and light chains, otherwise absent in IgA2.

Engineering the IgA heavy chain in order to reduce sensitivity to bacterial proteases has highlighted which antibody domains are necessary for protease activity [13]. For example, three amino acids in the CH3 domain were found to confer susceptibility to breakdown by a *Neiseria meningitidis* protease. Efforts to engineer a IgA1/IgA2 hybrid with half of IgA1's long hinge has elucidated the specific proteases which bind to each half of the long hinge [16]. For example, proteases produced by Neisseria *gonorrhoeae* and *Neisseria meningitidis* are able to cleave the hybrid hinge, whereas the *Haemophilus influenzae Type 1* protease did not.

Protein structure can be stabilised by covalent bonds or non-covalent forces. A strategy in antibody engineering is introducing covalent bonds to antibody domains in the form of disulphide bridges. These make the antibody complex more stable and less prone to breakdown. For example, a single amino acid mutation (P221R) will sterically allow new covalent bonding in IgA2 between the heavy and light chain interaction—this leads to a more robust antibody and less free light chain [17].

The incorporation of J chain has been identified as a bottleneck in SIgA assembly [18]. This small 15 kDa polypeptide covalently attaches to the Fc region of opposite IgA monomers via two key cysteines to make dimeric IgA. However, excess J chain can lead to high molecular weight aggregation due to the protein's two free thiol groups [19]. Fine-tuning the expression of J chain relative to the other SIgA components, for example by putting it under the control of a stronger promoter, can help optimise dimeric and secretory forms of the antibody [20].

The final step to make a SIgA molecule is attachment of the secretory component. *In vivo* this happens as dimeric IgA is trancytosed from the lamina propria into mucosal secretions after binding to the polymeric immunoglobulin receptor [21]; *in cellulo* strategies rely on the SIgA complex assembling through co-expression of multiple genes or *in vitro* by incubating free secretory component with purified dimeric IgA [22].

There has been interest in engineering IgA fused with other immunoglobulin forms, particularly with camelid VHH nanobodies, replacing the variable light and heavy chains. These nanobodies are small (~15 kDa compared to the ~55 kDa Fab domain) and eliminate the need for a light chain. This enables the targeting of epitopes in deeper antigenic clefts and also simplifies production [23]. VHH-IgA fusions demonstrated increased functionality whilst retaining the ability to dimerise via the J chain and bind to secretory component. For example, secretory VHH-IgA fusions were protective against ETEC infection unlike related VHH-IgG fusions [24]. Another engineered fusion antibody which has been explored is an IgG backbone with IgA Fc sequences inserted—this demonstrated binding to both the IgA receptor (FcαRI) and IgG receptors (FcγRI/FcγRIIa/ FcγRIIb), possessing both classes of effector function [25].

With the recent increased application of advanced data science and machine learning in biotechnology, the coming years will potentially be very exciting for antibody engineering. Increased access to technology such as AlphaFold by Google, an open-source software to predict the 3D structure of protein sequences, will facilitate the bridging between *in vitro* and digital modelling [26]. Future approaches to improve antibody stability may employ computational models to identify antibody conformations with increased stability [27]. Indeed, proof-of-concept studies that apply different types of modelling software to antibody design are already underway,

either by designing new proteins using an existing sequence dataset, or improving a 3D model and predicting the sequence which would give rise to it [28]. Such strategies have the potential to take native antibody sequences and generate novel sequences with better antigen binding affinity, for example [29]. It is important to note that this strategy requires a robust data set to base new designs on, and that improved antibody characteristics are based on simulation data which may not completely translate *in vitro/in vivo.* However, this is a relatively new technology. More experimental realworld data will both help guide and validate algorithms—unfortunately this information is sparse for IgA compared to IgG. Nevertheless, *in silico* antibody engineering has the power to drastically reduce the hours spent in laborious real-world antibody screening [30]. It is only a matter of when, not if, it will be applied to mainstream SIgA production.
