Structural and genetic diversity in antibody repertoires from diverse species

Abstract

The antibody repertoire is the fundamental unit that enables development of antigen specific adaptive immune responses against pathogens. Different species have developed diverse genetic and structural strategies to create their respective antibody repertoires. Here we review the shark, chicken, camel, and cow repertoires as unique examples of structural and genetic diversity. Given the enormous importance of antibodies in medicine and biological research, the novel properties of these antibody repertoires may enable discovery or engineering of antibodies from these non-human species against difficult or important epitopes.

Introduction

The adaptive immune system is a crucial adaptation associated with the evolution of vertebrates. The ability to create repertoires of antigen binding molecules and then select those which bind with high affinity to their cognate antigen in order to remove evading pathogens is the key feature of any adaptive immune system. While it may appear that the antibody response is capable of binding an infinite number of epitopes, this is probably not true. The germline repertoires of different species have evolved both for the ability to create diversity, but also in a Darwinian way to select antibody features specific for each organism’s antigenic load. Indeed, knockout of a single V_H region in mice can result in impaired ability to fight certain infections associated with a key neutralizing epitope [1]. Thus, while different species may be capable of producing billions of different paratopes in their antibody repertoire, these repertoires are still limited by the scaffold and germline genetic composition of their antibody genes. In this regard, various species have evolved novel structural features that presumably were selected based on their unique struggles with specific foreign invaders. Across all species, therefore, the ‘paratope universe’ is certainly dramatically larger than the individual repertoire of any specific species.

A challenge in the creation of the antibody repertoire at the genetic level is that the number of genes required for millions of antigen binding molecules could theoretically exceed the amount of DNA in the genome. In this regard, strategies to diversify antibodies based on the combinatorial rearrangement of genetic elements to produce single antibodies per cell has been accomplished in several species [2–4,5^••]. In the jawless fish (e.g. lamprey and hagfish), the combinatorial association of different variable lymphocyte receptor genes produces an array of antigen receptors based on the leucine rich repeat motif, which is the only known repertoire that does not use the immunoglobulin domain as a structural scaffold [2,6]. In humans and mice V(D)J recombination combinatorially produces genetic diversity by using a multitude of different V, D, and J gene segments, as well as N and P nucleotide junctional diversity, to create the naïve antibody repertoire [7,8]. This recombination event particularly provides diversity within CDR H3 of the paratope which is often a major contact of antigen (Figure 1).

Figure 1.

V(D)J recombination and the structure of human and mouse antibodies. A schematic of the human heavy chain locus is shown, where multiple V_H, D_H, and J_H regions can rearrange in any given B-cell to produce a functional VDJ unit that encodes the heavy chain variable region. A close-up of the antigen binding region (bottom left) illustrates the regions encoded by the V, D, and J gene segments. A full length antibody is shown on the bottom right.

Antibodies are major tools in biotechnology; they function as research tools, diagnostic reagents, and are now an important class of drugs for the pharmaceutical industry [9,10]. Historically, the generation of an antibody was accomplished through immunization and hybridoma techniques [11], which resulted in reagents for ELISAs, western blots, flow cytometry, immunoprecipitation, and other important immunochemical techniques. Many therapeutic antibodies were also originally discovered through these techniques [12]. However, as sophistication in drug discovery has increased, along with an explosion in data in genetics and structural biology, it is clear that antibodies with certain binding and functional properties would be ideal for various specific applications. For example, nearly all currently FDA approved antibodies are high affinity antagonists, however antibodies with agonist, antagonist, modulator, or other activities may be mediated by binding to limited epitopes with exquisite specificity and affinity, and may be highly desireable for certain indications and targets [13]. Additionally, binding to enzymatic active sites, allosteric epitopes, or to important regions on multipass membrane proteins may be difficult using standard techniques with the canonical human or mouse antibody scaffolds. Therefore, alternative paratope structures derived from alternate species may allow unique physicochemical binding characteristics not available in traditionally used mouse or human antibodies.

Certain antigens and epitopes can be particularly challenging to develop antibodies against. It is a well appreciated difficulty to generate antibodies with pharmacologic activity against multipass transmembrane proteins like GPCRs and ion channels [14,15]. While this difficulty may in part be due to the challenges associated with producing stable purified protein in functional form, it is also that the typical flat antibody paratope may not be optimal for interacting with grooves, pores, or other concave epitopes in these receptors. Similarly, antibodies that ‘reach’ into enzymatic active sites are relatively rare. Even certain microorganisms present challenges in eliciting antibodies against neutralizing epitopes. In the case of HIV, while antibodies can be raised against the spike protein gp120, these antibodies are typically not neutralizing, and rarely broadly neutralizing [16,17]. However, extremely rare antibodies with long CDR H3s have been identified which pierce the glycan shield and contact key conserved epitopes on the protein and can neutralize multiple viral clades. Given that the human and mouse repertoires appear limited in its ability to target these epitopes, utilization of other species with alternative antibody scaffolds could lend itself to identifying new antibody tools to study the fine aspects of viral neutralization, and potentially to bind more difficult antigens or epitopes.

Here we review strategies for repertoire diversity in shark, camel, chicken, and cow as unique representatives of novel mechanisms at both the genetic and structural levels. Within vertebrates the differences in genetic and structural strategies to create an antibody repertorie are significant (Figure 2), and with relatively few species studied in detail, it seems probable that further variations have yet to be discovered.

Figure 2.

Evolutionary relationships among species: useful antibody innovations evolved several times in vertebrates. Deuterostome phylogeny highlights the divergence times between some lineages that are discussed in this review (in bold). Agnathans such as the lamprey have lymphocyte classes similar to jawed vertebrates, but employ diverse repertoires of variable lymphocyte receptors from somatically rearranged loci employing activation induced cytidine deaminase (AID). Gnathostomes such as the shark began using immunoglobulin superfamily receptors on lymphocytes and first evolved an immunoglobulin heavy chain in the absence of light chains. Birds use AID to gene convert an initially very restricted recombination activating gene (RAG) generated locus via sequence donation from variable segment pseudogenes. Camelids convergently evolved heavy-chain only antibody variants of IgG, and at least some members of the family Bovidae employ heavy chains with an ultralong CDR H3, creating another diverse disulfide stabilized domain. (Right) Structures of antibody binding fragments from diverse species. Ribbon diagrams of lamprey (PDB, 3E6J), shark (4HGK), chicken (4P48), camel (1ZVY), cow (4K3D), and human (1N8Z) antigen binding fragments. Key distinguishing features are that lampreys utilize a leucine rich repeat scaffold, as opposed to other species which use the immunoglobulin domain. Shark and camelid fragments are devoid of light chains, with camelids often using a longer disulfide-bonded CDR H3. Chickens have novel canonical classes of CDR H1, and disulfides in CDR H3. Cows have evolved an ultralong CDR H3 repertoire with a disulfide bonded knob that sits atop a β-ribbon stalk.

Chicken

The chicken has been a model organism for studying B-cell development for over half a century. The role of the bursa of Fabricius in B-cell development was revealed when bursectomized chicks failed to produce antibodies [18,19]. In fact, the name ‘B-cell’ is derived from this discovery regarding the importance of the bursa. Subsequent work revealed B-cell and repertoire developmental pathways in birds. Chickens have serum IgM, IgA, and IgY, the first two of which are homologs of their mammalian counterparts, however they do not have IgE or IgD [20]. The IgY appears related to both mammalian IgG and IgE [3], and may be an evolutionary common ancestor to both [21].

Genetics of chicken repertoire generation

Chickens utilize a diversity creating process that is distinctly unique compared to humans or mice. Chicken antigen receptor genes undergo a single V(D)J recombination event, followed by gene conversion utilizing multiple upstream V-region pseudogenes [3] (Figure 3a). Rearranged variants of the pseudogenes can further diversify the CDR H3 region by inserting sequence into the D_H region. Interestingly, the gene conversion process is dependent on the activation-induced cytidine deaminase (AID) enzyme [22], the same factor which is required for performing somatic hypermutation (SH) and class switch recombination in other species [23]. While framework diversity is limited in chicken antibodies [24^••], probably due to the need for significant homology at the DNA level for efficient gene conversion, the potential diversity of creating different CDR combinations and content through multiple theoretical gene conversion recombination events is enormous. Although not discussed in detail here, rabbits also use a gene conversion strategy to diversify their antibody repertoire [25].

Figure 3.

Novel processes for producing genetic diversity in chickens and cows. (a) Unlike humans or mice, chickens use a gene conversion strategy to create a diverse antibody repertoire. A single VDJ event occurs on the heavy chain locus, followed by gene conversion using sequence homology from a number of 5′ pseudogenes (ψV_H) which can be in rearranged form and donate diverse genetic fragments into the rearranged V region. (b) Ultralong CDR H3 and repertoire development in the cow. Cow antibodies with ultralong CDR H3s appear to use only one V_H region, V_HBUL, which recombines with a long D region, D_H2 to produce a germline VDJ recombined V-region. The single rearrangement then undergoes somatic hypermutation which may mutate residues to or from cysteine, changing the disulfide patterns of the repertoire, in addition to adding sequence diversity.

Structure of chicken antibodies

While chickens have been a model system for repertoire and lymphocyte development, few detailed structural studies on chicken antibodies have been performed. In repertoire analysis by gene sequencing, Wu et al. analyzed the amino acid content of chicken heavy chains [24^••]. Interestingly, cysteine content was substantially higher in chicken CDRs (9.4%) compared to mice (0.25%) or humans (1.6%). They also identified six families of putatively different disulfide patterns, which may include disulfide bonds within CDR H3, or between CDR H3 and CDR H1 or CDR H2. Tyrosine, an amino acid reportedly important and abundant in antibody CDRs [26–28], is found less frequently in chicken CDR H3 (9.2%) compared to humans (16.8%). While on average chicken CDR H3s are not longer than humans or camelids, certain CDR H3s may form longer and unique disulfide stabilized structures. Selection of high affinity antibodies of these unusual structural subtypes could be accomplished in phage-based discovery systems [24^••]. Future crystallography of examples of these antibody fragments, and their comparison with typical human CDR H3s or the unusual cow antibody CDR H3s described below will be an interesting comparative study in this regard.

To date only three structures are available for chicken antibody fragments. In structural studies of two chicken scFvs Conroy et al. identified new unique canonical classes of CDR L1, and a disulfide bonded CDR H3 [29^•]. The extended loop of L1 along with disulfide bonded CDR H3 could present a unique paratope that is structurally distinct from mouse or human antibodies and, along with the sequence analysis of Wu et al. strongly suggests that chickens may have a novel repertoire of paratopes.

Engineering chicken antibodies

As a model system for immunology, chicken antibodies have been made for multiple research applications [30]. Chicken IgY can be made by several companies using immunization, and phage display of chicken antibody fragments has been accomplished against multiple antigens [31–33]. From a protein engineering standpoint, the DT40 cell line has been manipulated to serve as a host for SH of antigen receptor genes [34], and human pseudogenes have been engineered into DT40 and used to create human repertoires using gene conversion, a process where a donor DNA sequence replaces a homologous sequence in the genome [35]. In this regard, chickens with the antibody loci knocked out have been produced [36^•], which may enable further transgenic engineering of human antibody V-regions or pseudogenes to produce transgenic chickens with human antibodies that utilize the novel gene conversion diversity generating systems of birds.

Camels

Camels have conventional hetero-tetrameric antibodies with identical heavy chains paired with identical light chains. These canonical IgG molecules represent approximately 25% of total circulating antibodies [37]. In a significant departure from other mammals, camels also have an unusual antibody type that is similar to a conventional IgG molecule but has identical heavy chains that lack the C_H1 domain and does not pair with light chains. These heavy-chain only antibodies represent a significant fraction of the immunoglobulins in the serum, constituting up to 75% of total proteins binding to a protein A column. These molecules have been termed heavy-chain antibodies (HCAbs) and have a dedicated variable domain referred to as the V_HH. The V_HH domains are structurally and functionally similar to an F_v fragment derived from typical IgG molecules but have only three CDR variable loops to define the antigen binding surface. HCAbs have a molecular weight of about 90 kDa compared to the conventional IgG molecular weight of about 150 kDa [38]. Another differentiating feature of the V_HH from canonical IgGs are the long loops that comprise the CDR H3. These long CDR H3 regions may enable V_HH’s to interact with and inhibit unusual targets or epitopes not available to the flat binding surface of conventional antibodies such as enzyme active sites or other recessed crevices [39]. Both alpacas and llamas also have HCAbs that are very similar to those found in camels [40].

Camel antibodies have received much interest due to the small size of the V_HH, robust biophysical characteristics and unusual antigen binding surface. Currently 5 V_HH domains are in clinical development by Ablynx (Ghent, Belgium) and many more are in the early research phase in both academia and the pharmaceutical industry [12].

Genetics of camel repertoire generation

Camel HCAbs share the same gene locus as their conventional IgG tetrameric counterpart [41]. In addition, both HCAbs and IgGs have dedicated variable region genes encoded in germline sequences and undergo classical V(D)J recombination [42]. Repertoire diversity is driven by over 30 unique variable region (V_HH) sequences, possible unique splicing events of the mRNA, and promiscuous V genes that can produce either V_H (which will also pair with V_L molecules) or V_HH domains, each of which can undergo SH to produce further diversity [42–44].

Structure of HCAbs

Camel homodimeric HCAbs have an unusual structure compared to conventional IgG molecules. These HCAbs lack the C_H1 domain, and the V_HH variable region is not compatible with pairing to light chains, resulting in an antibody structure with a molecular weight under 100 kDa. A G to A point mutation that disrupts a consensus splicing sequence may be the cause for HCAbs lacking a C_H1 domain. This sequence is located at the 5′ end of the intron between the C_H1 hinge-side of the exon and increases the propensity to splice out the C_H1 portion of the mRNA [45]. Different isoforms of these HCAbs have been identified and classified by the length of the hinge region sequence between the V_HH domain and the C_H2 domain. Shorter hinge length isoforms are referred to as IgG3 and the longer hinge regions as IgG2 [38].

Camel HCAbs have unique characteristics that differentiate them from canonical IgGs. Specific mutations in conserved locations are prevalent in HCAbs to possibly compensate for the lack of C_H1 and light chains. For example, the mutation Leu11Ser, replaces interactions that classically make contact with the C_H1 domain. While protein sequences of V_H and V_HH framework and hypervariable regions reveal similar structural organization, amino acids at key positions in V_HH that define the V_L/V_H interface were substituted at Val37, Phe42, Gly44, Leu 45, Trp47, Glu49, Arg50 and Gly52 with smaller amino acids in V_HH relative to wild-type V_H regions that form heterodimers [46,47]. Additionally, a co-crystal structure of a camel V_HH complexed with lysozyme highlights an unusually large convex H3 derived paratope that bound inside of the enzyme active site [48^••]. The longer H3 loops may provide added surface area to compensate for the loss of the three V_L CDR loops that classically contribute to antigen contacts [39] and can achieve binding affinities similar to conventional IgGs [49]. V_HH CDR H3 loops are not only longer than canonical IgG H3’s but also have a significantly greater prevalence of intramolecular disulfide bonds that potentially define specific paratope topology [50]. Disulfide bonds form between cysteine residues located at H1 and H3 and between H3 and framework region 2 (Figure 4a). Cysteine is generally limited to positions 30, 32 and 33 in H1, position 50 in FR2 and can occur in several different positions throughout H3 [51]. The constrained disulfide structures may help stabilize the H3 antigen-loop complex, potentially leading to stronger antigen interactions. The additional stability of the H3 region may reduce the entropic cost of restricting the conformation of a flexible loop, lowering the energy barrier for antigen contact and association [52^••], and also may help compensate for the lack of electrostatic interaction classically found in IgG H3 loops between Arg or Lys94 and Asp101 (by Kabat numbering) [53,46].

Figure 4.

Protruding CDR H3 structures of camelid and cow antibodies. (a) CDR H3 structure and epitope binding of camel antibody fragments. (Left) Closeup of the structure of a camelid Fv, showing the long CDR H3 loop with a disulfide bond between CDR H3 and CDR H1. Unique epitopes bound by camelid Fvs, including the β-adrenergic GPCR (middle) and enzymatic active site of lysozyme (right). Note the protruding CDR H3 which reach into concave epitopes on each of these antigens. (b) Structures of two cow Fab fragments, BLV1H12 and BLV5B8, showing ultralong CDR H3s that are composed of a β-ribbon ‘stalk’ and disulfide bonded ‘knob’ structural minidomains. (c) Closeup views of the surface of the knob domains of BLV1H12 and BLV5B8. The sequences of the knobs are shown below, and a two-dimensional representation of the loops are shown to the right. The disulfide patterns of the two knobs are different, as indicated above the sequences and in the two dimensional loop schematic.

In addition to the unusual structure that potentially enables unique target access, V_HH domains are thermodynamically stable and reversibly melt without common complications such as aggregation and loss of binding activity after refolding. While capable of retaining activity after refolding from denaturation, V_HH domains do not exceed the thermostability of human subclass V_H3 domains [54].

Engineering HCAbs

The small size, stability and ability to tightly bind to antigens have made the industrialization of camel V_HH domains, termed ‘nanobodies’, of great interest [55]. Therapeutic applications of nanobodies and most foreign proteins may require some degree of engineering to incorporate as much human-like sequence as possible to minimize potential immunogenicity. Several camel V_HH domain sequences have been shown to be homologous with the human V_H3 and V_H4 germline sequences [39,43]. The high degree of homology between the camel V_HH framework sequences with those of the human sequences may aid or even bypass the need to engineer and humanize [44]. Regardless, in an effort to reduce potential immunogenicity, several key conserved residues in the camel FR2 region have been explored by mutating to the human consensus amino acids. Interestingly, biophysical data on several of these engineered molecules such as Glu49Gly and Arg50Leu resulted in more thermodynamically robust constructs while other mutations in FR2 negatively affected antigen binding due to impact of the H3 loop position [55]. H3 loop length is also correlated with the characteristics of surrounding residues on the scaffold and perhaps should be considered when engineering camel V_HH domains [56].

There are several examples of camel antibodies with unique characteristics, including many with clinical applications [57^•]. Of particular note, the small size of the V_HH and its associated longer CDR H3 may allow binding to antigens or epitopes not easily accessible by typical mouse or human antibodies. In this regard, crystallization of the active state β₂ adrenoceptor, a multipass transmembrane GPCR, was facilitated by a specific camel antibody fragment, an achievement that produced a key component of the knowledge that ultimately was awarded the Nobel Prize in chemistry in 2012 in the area of GPCR structure and function [58^•] (Figure 4a, middle).

Standard molecular library techniques, particularly phage display, have been utilized to discover and engineer camelid antibodies [57^•]. Immunization of an animal followed by construction of a nanobody phage display library result in discovery of antigen specific binders [59]. However, removal of large animal immunization can allow rapid molecular evolution techniques in the laboratory if large ‘naïve’ libraries are constructed. Monegal et al., 2009 constructed a large naïve llama phage display library that enabled identification of constituents with high affinity binding to FGFR1 [60]. In a related approach, a synthetic phage display nanobody library was built based on a conserved camel single-domain antibody fragment (V_HH) framework and diversity was introduced into the CDR H3 by randomization using synthetic oligonucleotides [61]. While immunization and phage display approaches have met with success in identifying specific binders, Eschrichia coli display with camel V_HH’s has also resulted in the identification of high affinity binders [62].

Several camel V_HH domain antibodies have been discovered and are in early preclinical development in areas including oncology, infectious, inflammatory and neurodegenerative diseases. Ablynx currently has five humanized camel V_HH domains in clinical development. Clinical targets include von Willibrand’s Factor, TNF-α, IL-6, RANKL and Respiratory Syncytial Virus [12,49,60,63]. Camel V_HH’s have also been identified that can neutralize toxins such as the botulinum neurotoxin serotype E [64].

Shark

Nearly half a billion years ago the adaptive immune system based on immunoglobulins, T cell receptors and the major histocompatibility complex evolved in cartilaginous fish (reviewed in [65]). Thus, sharks represent man’s most distant cousins that share the fundamental components of ‘our’ immune system. There are, however, some interesting differences between the lymphocyte antigen receptors of shark and man.

Sharks have three IgH chain isotypes. IgM (μ) appears developmentally and ontologically ancestral and is found in sharks and nearly all jawed vertebrates, with the coelacanth being the only known exception [66]. IgW (ω) is also extant in sharks and orthologous to IgD of other vertebrates, and there is good evidence for a form of class switch recombination between IgM and IgW [67]. Lastly, IgNAR is a cartilaginous fish lineage-specific isotype that does not associate with light chains [68]. Unlike humans or mice which have two light chain classes, in cartilaginous fish IgM and IgW have four IgL with which to heterodimerize: λ, κ, σ, and σ-cart [69]. Sharks have the α, β, γ and δ T cell receptor chains found in other vertebrates [70], and employ the doubly rearranging NARTCR form of TCRd that includes a variable domain very similar to that of the IgL-less Ig-NAR antibody [71]. As a unique antigen receptor, we will focus on the IgL-less IgNAR immunoglobulin isotype, as it has generated much interest beyond comparative immunologists.

Genetics of IgNAR repertoire generation

Although humoral and cellular lymphocyte lineages with diverse antigen receptor repertoires exist in the older, jawless lamprey and hagfish [6], V(D)J recombination of immunoglobulin and T cell receptor variable domain genes is a shared characteristic of the jawed vertebrates (reviewed in [72]).

Shark TCR are in the typical translocon arrangement [73], but their Ig loci exist in a multiple cluster organization throughout the genome [74] making class switch and successful haplotype exclusion [75] especially intriguing. Some of the Ig loci of cartilaginous fish are partially (VDJ) or completely (VDJ) germline joined [76], that is, the animal inherited the locus in a rearranged state from its parents. The fusion of V, D, and J elements in the germline (as opposed to a somatically developing lymphocyte) is a product of gonadal RAG expression in sharks [77]. Whether these preloaded paratopes bind antigens of particularly common or mortal pathogens of elasmobranchs still needs to be determined, but one germline-joined IgM is preferentially used in young sharks [78]. Although shark IgH arise from simpler loci with fewer elements, sequence differences between clusters and junctional diversification by N and P additions produce a repertoire as diverse as that of other vertebrates [79,80].

The variable domains of IgNAR share as much homology with other TCR as Ig (thus the moniker new antigen receptor). It has been suggested that IgNAR may have arisen evolutionarily via the invasion of an IgW cluster by a V gene of NARTCR [81]. More clusters of Ig often exist than IgNAR, yet upon antigen exposure IgNAR is heavily mutated [82]. Low IgNAR levels in young sharks, the slow rise of serum levels of IgNAR in the first year of shark development, and the clear affinity maturation of the molecule have suggested that this isotype serves a role in the shark immune battery analogous to mammalian IgG and that it may be T dependent [83,84]. A memory response characterized by specific antigen production (with secondary response kinetics after the primary response has resolved upon boost without adjuvant) is clearly capable with the IgNAR isotype in nurse shark [85]. SH in sharks may proceed through novel mechanisms as tandem non-template mutations of 2–5 contiguous bases are common [86,87].

Structure of IgNAR

Surface IgNAR has one amino-terminal V domain and either three or five constant domains. The first and third constant domains homodimerize and lend to the antibody stability, and the resolution of the constant region structure showed stabilizing motifs that also stabilize mammalian antibodies when engineered into them [88]. IgNAR can exist as a monomer but at least in the spiny dogfish has been found to multimerize [89]. The variable domains of IgNAR are unusual in that CDR H2 can form a belt around the side of the domain for a structure similar to C1-type immunoglobulin superfamily domains [90]. This peripheral orientation of CDR H2 to the antigen binding site is probably not always the case though, as IgNAR CDR H2 can have selected hypermutations and can be important for antigen binding [91^•].

The single V domain antibody general quaternary structure has independently evolved at least twice in vertebrate natural history, in cartilaginous fish and camelids [92^•]. There are genetic differences in these structural and analogic convergences, however. IgNAR is encoded at dedicated loci, whereas camelids use an IgG variant that has evolved to encode structural modifications to eschew the IgL [38]. It has been suggested that comparative immunology studies in more diverse clades will discover more vertebrate groups that utilize the single V antibody strategy [93].

Whereas the mobility and tissue-penetrance of canonical hetero-tetrameric antibodies are constrained by their large size, this is less of a problem for the 12 kDa IgNAR homodimer and less still for the IgNAR V domain by itself. This IgNARV is to date the smallest antigen binding domain known in the animal kingdom [94]. An increased frequency of polar and charged amino acids at the solvent exposed regions corresponding to the traditional V_H–V_L interface makes IgNAR particularly soluble.

Libraries and engineering

The relative stability of the IgNAR IgH homodimer and the somewhat simpler genetics encoding the single variable domain have fueled applied research to adapt this molecule to biotechnology and immunotherapy. Much is known of the optimal immunization protocols necessary for affinity matured humoral IgNAR responses in several elasmobranch species [95]. Phage display technologies have been easily adapted to analyze IgNAR antibodies from immunized sharks [96]. These technologies have been used to create recombinant shark antibodies to many pathogens of human health relevance, including Ebola [97]. The immunotherapeutics arms of some pharmaceutical companies are now exploiting the unique properties of shark IgNAR to make target-specific single V domain antibodies [98]. Despite only (at most) four hypervariable regions to bind antigen (and CDR H2 sometimes does not contribute to the paratope) IgNAR can bind antigen with very high affinities [90,96]. Diverse intradomain disulfide binding patterns divide IgNAR V domains into four types [99,100], only one of which (type III) has not yet been shown to give rise to high affinity binders.

Many different species of shark have now been used to develop libraries and screened with different display technologies, including phage display [96,101], ribosome display [102] and yeast surface display [103]. These libraries can be created from naïve sharks, immunized sharks, or synthetic vNAR libraries, but immunized sharks have been the most successful source till date, possibly due to the large phylogenetic distance between shark and man, reducing the chance of target-specific tolerance of conserved antigens [94].

The amino acid identity between IgNAR and human IgHV domains can be as low as 25%, thus the immunogenic potential is great [96], and progress has recently been made towards IgNARV humanization. A shark vNAR domain against human serum albumin (HAS) was engineered by converting more than half of the framework amino acids to those of the human germline IgL kappa variable sequence DPK9 [91^•]. This humanized molecule lost very little affinity over the product of the parental construct, and the humanized IgNARV displayed negligible immunogenicity in dendritic cell assays [104].

The smaller paratope of the shark IgNARV has contributed to its appeal for creating therapeutic binders for target antigen conformations that have been impervious to development of effective blocking or neutralizing antibodies by traditional paratopes. Discriminating binders of cholera toxin were identified from spiny and smooth dogfish naïve libraries, and these monoclonals were found to be more heat resistant than traditional mammalian cholera toxin reagents [105]. A shark IgNAR was also shown to be an effective ‘intrabody’ against the hepatitis B virus precore protein in the endoplasmic reticulum [106]. A list of notable IgNARV binders to relevant targets is shown in Supplementary Table 1 (adapted from [94,104]).

Cow

While sharks, camels, and chickens provide unique examples of alternate antibody structures both in terms of subunit composition (sharks, camels) and diversification mechanisms (chicken), recent discoveries in cow antibody structure and genetics reveal an unusual paradigm for creating both genetic and structural diversity. It has been known for quite some time, largely through the work of Kaushik and colleagues [107–113], that cow antibodies can have unusually long CDR H3 regions that can reach lengths of over sixty amino acids long. In fact, it appears that cows actually generate an ultralong CDR H3 repertoire alongside a shorter repertoire; but even this shorter repertoire is quite long compared to mouse or human antibodies, with cow ‘short’ H3 sequences often being over forty amino acids in length. In a departure from mouse and human antibodies on both a genetic and structural level, cows appear to use SH to create both amino acid content as well as disulfide bonded loop pattern diversity within the ‘stalk’ and ‘knob’ minidomains of their ultralong CDR H3.

Genetics of cow ultralong CDR H3 repertoire generation

Cows have limited combinatorial diversity potential with only 12 V_H regions that comprise one highly homologous heavy chain family (compared to seven in humans) which is related to the human V_H4 family [114–116,117^•]. Cows perform V(D)J recombination similarly to other species, however the ultralong subset of cow antibodies appears to preferentially use a single V_H (termed V_HBUL) and an ultralong D_H2 [5^••]. While the cow genome project has been published [118], the difficulty in assembly and annotation of the heavy chain locus leaves open the possibility that alternative V_H or D_H regions could still be discovered. Even if this were the case, deep sequencing analysis reveals that any undiscovered V_H or D_H regions would be highly similar to V_HBUL and D_H2 [5^••]. Unlike mice or humans, several species like cows and sheep are unusual in activating SH during development of the naïve primary repertoire, with AID induced mutations compensating for the limited V(D)J combinatorial diversity [117^•,119,120]. An unusual feature of cow germline D_H regions is that they encode multiple cysteines. In the ultralong D_H2 region a repeat of Gly-Tyr-Gly or Gly-Tyr-Ser sequences use the relatively uncommon codons GGT (for Gly) and AGT (for Ser) and TAT (for Tyr), each of which can be mutated to cysteine with one nucleotide change [5^••]. Furthermore, there are numerous RGYW SH ‘hotspots’ throughout the D_H region, making it potentially highly mutagenic. Thus, this cow D_H region appears to be primed to mutate to cysteine through AID induced SH. The D_H2 encodes four cysteines, which could potentially form two disulfide bonds. Cow antibody sequences, however, contain multiple cysteines that do not appear to align with those in D_H2. This lack of clear conservation to the germline was resolved when deep sequencing analysis revealed mutations both to and from cysteine in antibody sequences undergoing SH [5^••]. Alignment of CDR H3 sequences with the germline D_H2 revealed almost complete conservation of the first cysteine with progressively less conservation of the three remaining germline cysteines moving towards the C-terminus. Additionally, multiple ‘new’ cysteines that were generated somatically could be identified. Importantly these results suggest a novel mechanism for structural diversity generation; mutating to and from cysteines can alter the disulfide patterns in CDR H3, resulting in wholesale changes in loop structures and compositions [5^••] (Figure 3b).

Structure of cow antibodies

Wang et al. solved the structures of two bovine antibody Fab fragments which contained ultralong CDR H3s [5^••]. These two fragments, BLV1H12 and BLV5B8, were originally cloned from B-cells transformed with Bovine Leukemia Virus and their cognate antigens are not known [110]. While both had ultralong CDR H3s, their sequences in the CDR H3 regions were highly dissimilar. Remarkably, however, both structures had two unusual motifs never seen before in antibody CDR H3 structures: a β-ribbon ‘stalk’ that protruded far from the traditional antibody paratope, and a distal disulfide bonded ‘knob’ which rested upon the stalk (Figure 4b and c). This ‘stalk and knob’ feature is reminiscent of a mushroom protruding out from the surface of the antibody (Figure 4b). Both CDR H3s had six cysteines, however their positions were not conserved. When the first cysteine was ‘fixed’ and aligned with the germline D_H2 region, three of the cysteines could be aligned positionally, however only the first two were conserved with the D_H2 germline. The structures clearly revealed different disulfide bonding patterns in the knobs of BLV1H12 and BLV5B8 where the former had a 1–4, 2–6, 3–5 pattern and the latter had 1–3, 2–4, 5–6 (Figure 4c). Despite the unique ‘stalk and knob’ features, the sequences, disulfide bonding pattern, surface shape and charge, and loop lengths within the knob were highly dissimilar between the two antibody fragments. Structurally, the remaining five CDRs and variable region framework regions were nearly superimposable and had little sequence variation. Thus the diversity of these cow antibodies appears to all reside within the ultralong CDR H3 region. Deep sequencing also revealed that most sequences contained an even number of cysteines at diverse positions, further strengthening the hypothesis that disulfide bonds, and their associated loops, are an important component of the structural repertoire [5^••]. The unique ability to use a single germline genetic construct as a template to mutationally derive different disulfide patterns, and thus minifolds, is a novel way to create structural diversity in an antigen binding repertoire (Figure 3b).

Engineering cow antibodies

Compared to other species discussed above, little work has been done on discovery and engineering bovine antibodies. Phage display has been successful using cow antibody repertoires [121], however it is unclear whether the multiple disulfide bonds found in the knob of ultralong CDR H3s can be successfully selected using prokaryotic based systems which may lack the machinery to efficiently secrete and fold these complex molecules. An antigen specific ultralong CDR H3 antibody targeting the NS2-3 protein of Bovine Diarrheal Virus was expressed and its binding properties analyzed by site directed mutagenesis [5^••]. All of the binding was found to reside within the ultralong CDR H3 ‘knob’ domain, with the other CDRs being irrelevant for antigen binding. Given the unusual paratope of ultralong CDR H3s, engineering efforts in replacing the ‘knob’ with known bioactive peptides have been undertaken. Interestingly, the knob could be replaced by G-CSF (granulocyte colony stimulating factor) [122] and erythropoietin [123], with biological function of the engineered protein being maintained. Liu et al. replaced the knob domain of BLV1H12 with a β-hairpin peptide known to bind CXCR4, producing a molecule with high affinity and activity [124]. This ‘knob replacement’ engineering could endow small peptides with long half-life and effector function of traditional antibodies, and may be a way to improve dosing regimens for some recombinant therapeutic proteins. Humanization of the bovine scaffold has not yet been reported, however the homology with human VH4 suggests this may be feasible, albeit with the caveat that the other cow CDRs besides H3 may play an important structural and stability role through their contacts with the ultralong ‘stalk’. The structure of the ultralong CDR H3, whose knob is of similar size and shape to other small disulfide bonded peptides like knottins, chemokines, toxins, protease inhibitors, and defensins, certainly suggests that unique epitopes could be bound by these unusual antibodies.

Conclusions

Different species have taken unique approaches to create diverse antigen receptor repertoires. At the genetic level, while V(D)J recombination is a key molecular mechanism amongst all jawed vertebrates, its role in the actual creation of diversity ranges from providing a template for gene conversion with a single V(D)J event in chickens, to being a major source of combinatorial diversity in mice and humans through unique V(D)J events in each B-cell. Similarly, the AID enzyme is primarily utilized to create the secondary repertoire by SH in mice and humans, but appears to play a major role in creating the primary repertoire in cows and perhaps other species [117^•].

From a structural standpoint, mice and humans use a heterotetrameric IgG to bind antigens, with diversity occurring in all six CDRs, and combinatorial and junctional diversity focused on the CDR H3 and L3. These paratopes typically form a flat or undulating binding surface for contacting antigen. By contrast, camels and sharks have convergently developed heavy chain antibodies devoid of light chains. In an additional twist, the cow ultralong CDR H3 repertoire appears to contain all of its diversity within CDR H3, and none in the other five CDRs which provide a structural role for the CDR H3 ‘stalk’. Although cows use a light chain, the ultralong repertoire is limited in using a single light chain V_L which pairs with the single V_HBUL V-region. Chickens appear to have unique CDR L1 structures, and similarly to camelids, have disulfide bonds within CDR H3 or between H3 and H1 or H2. While the limited structural work done thus far shows major differences in antibodies from different species, further study on these different species should shed light on just how different the paratope repertoires are from one another as well as human and mouse antibodies.

Why significantly different genetic and structural strategies have evolved in different species is an interesting question. It appears as though both shark and camelids have evolved heavy chain-only antibodies in a convergent manner. Similarly, the strategy of using gene conversion appears to also have convergently evolved in both chickens and rabbits. Thus, both the genetic mechanism of gene conversion and the structural paradigm of heavy chain only antibodies must have certain evolutionary advantages, however what these may be are currently unclear. The structural differences between antibodies of different species suggest that different antigen structures (or structural classes) may have applied evolutionary pressure on individual species to develop alternative paratopes to enable an effective immune response. It has been suggested that the smaller Fv binding unit of camelid HCAbs may enable binding to unique concave epitopes, and camelid antibodies with protruding CDR H3s have been raised against difficult GPCR and enzyme active site epitopes (Figure 4a). The ‘knob’ regions of ultralong CDR H3 cow antibodies, which are even smaller than a nanobody, might be able to similarly ‘reach’ into such concave epitopes. Of note, both camelids and cows are ruminants which have substantial antigen load in their rumen stomach compartment that is made up of high titers of both prokaryotic and eukaryotic microorganisms [125–128]. Whether these microorganisms provided evolutionary pressure to develop novel antibody repertoires in cows and camels will require significant further study.

Given the enormous importance of antibodies in research and medicine as reagents, diagnostics, and therapeutics, the potential use of the immune system of alternative species could provide unique molecules that may have biochemical properties not present in antibodies derived from classic species like humans or mice. The paratopes provided by unique L1s, disulfide bonded or ultralong CDR H3s, or antibody Fv regions devoid of light chains certainly will allow different modes of binding to antigens, or even enable binding to certain epitope structures. Additionally, distant species may allow a more robust immune response to human antigens, due to their evolutionary distance and potential ability to overcome tolerance. Given these unique advantages, future efforts in discovery of antibodies and study of alternative species’ immune systems is sure to continue to uncover novel mechanisms for antibody diversity generation as well as further variations on the structural theme of how to design an antibody repertoire. With the recent progress in deep sequencing antibody repertoires and whole genomes, detection of novel features of antibody repertoires in diverse species could progress rapidly in the near future [129,130].