Sharks, similar to camelids, possess unconventional heavy (H) chain antibodies, consisting of heavy chain homodimers in which each chain contains one single variable and five constant domains [1,2]. The conserved amino acid residues between the shark heavy chain antibody and those involved in forming the core of immunoglobulin and T-cell receptor variable regions, gave impetus for naming shark heavy chain antibody Immunoglobulin New Antigen Receptor (IgNAR or NAR) [2,3]. Structural analysis by electron microscopy, crystal structure, and 3D modeling revealed that there are three NAR isotypes. These isotypes are defined according to their pattern of inter-loop disulfide linkages within the variable region and the timing of their appearance during the animal's development [3-6].

Using genetic engineering a single NAR variable (V) antigen-binding domain can be expressed as a separate soluble protein often referred to as a shark single domain antibody (sdAb). Shark sdAbs contain four conserved frame regions (FRs) and two complementarity-determining regions (CDRs), making them the smallest (~12 kD) Ig-based recognition units with full capacity for antigen binding affinity and specificity. Due to their small size, they may be able to access antigen epitopes not generally recognized by recombinant conventional antibodies [7]. Although both shark and camelid sdAbs share similar structural features and functional requirements [3], shark sdAb lack a conventional CDR2, and contain two hyper-variable regions (HVs), HV2 and HV4, which may contribute to antigen binding [3,8,9]. According to the crystallographic analysis of NAR V structure, the loop of HV2, located within the FR2-CDR2 region, is located across the middle of the molecules and may influence the conformation of the CDR3; moreover, the loop of HV4, located between HV2 and CDR3, is formed proximal to CDR1 and may influence the antigen binding interactions [3].

Similar to camelid sdAbs, shark sdAbs exhibit excellent solubility for protein production, superior to many recombinant conventional antibodies, and retain conformational stability when heated or refold correctly upon cooling, [10-13]. These intrinsic properties make sdAb exceptional alternatives for diagnostic applications. Using phage display technology and PCR amplification, the repertoire of the naturally occurring NAR V from either immunized or naïve (non-immunized) animals were established and used for panning against target antigens [10,11]. High affinity binders to a specific target were obtained from immunized libraries however required a waiting period for suitable immunization to be achieved and an animal care facility [11]. On the other hand, weak binders against a wide variety of target antigens were obtained from naïve libraries; they were selected rapidly and no immunization period was required [10,14]. If higher affinity binders are desired than those obtained from the naïve library, the sdAb can be enhanced using in vitro affinity maturation [12,15].

It is believed that the diversity of naturally occurring NAR V results from multiple rearrangements of the CDR genes and somatic hypermutations in vivo [2]. Consistent with this finding, the complexity of shark NAR V usually resides in CDR1 with sequence variation within residues 28–33 and an extended CDR3, which varies in length (5–23 residues) and in amino acid composition. Routes to introduce diversity and increase the complexity of naïve libraries include: variation of CDRs via DNA shuffling [16], PCR using randomized primers, and random mutagenesis by error prone PCR [17] or dNTP analogs [18]. Although DNA shuffling, which involves the recombination of several small DNA fragments within a whole variable region, has been successfully used to create a semi-synthetic llama library with giga diversity [19], it has not been used for constructing a more diverse shark display library due to the short NAR V DNA fragments, less than 400 bp in size. Instead, PCR using randomized CDR3 sequences has been used to construct more diverse synthetic libraries, resulting in binders with higher affinity and specificity to target antigens [10,12].

Our previously described naïve NAR V library, SP, from spiny dogfish shark (Squalus acanthias) was successfully panned for the isolation of binders to cholera toxin; however our ability to obtain specific binders towards other targets was disappointing [20]. In this study, we expanded the utility of the naive display library by introducing variations in CDR3 amino acid composition and length (13, 16, or 18 residues). In this manner, a semi-synthetic NAR V display library, SPSL1, with a complexity of ~1e9 was successfully constructed and used for panning against ricin, staphylococcal enterotoxin B (SEB) and BoNT/A complex toxoid. We successfully selected binders against these three toxins and characterized the purified sdAbs using ELISA and Luminex 100 assays. Our results suggest that this new diversified NAR V display library can serve as a fruitful source of sdAbs against a variety of toxins.

Table 1 Randomized oligonucleotide primers used for amplification of randomized spNAR V CDR3 sequences.

After three rounds of panning the SPSL1, we observed mild enrichment (~3-fold) for SEB binding (Fig. 1A) and strong enrichment (> 10-fold) for ricin. The binding enrichments from panning SPSL1 for both toxins were higher than those from panning the naturally occurring NAR V display library, SP (Figs 1A and 1B). In addition, we observed a 12-fold enrichment for BoNT/A complex toxoid binding after the first SPSL1 panning (Fig. 1C). For each toxin we screened 96 clones randomly selected from SPSL1 panning round 2 or 3 using monoclonal phage ELISA as described in the methods section. From each selection we sequenced 15 clones that had a ratio of toxin binding to BSA binding of greater than 3. We obtained two unique SEB binding sequences, one ricin binding sequence and two BoNT/A complex toxoid binding sequences.

Figure 1 Polyclonal phage ELISA. Phage pools obtained from panning the new SPSL1 with the three tested toxins and panning the previously established SP with SEB and ricin were tested for binding specificity using ELISA. After three rounds of panning SPSL1, SEB (more ...)

Figure 2 Sequences of isolated toxin binding sdAb. Fifteen potential binders for each tested target were sequenced. The resulting unique sequences were decoded into amino acid sequences, which were aligned using ClustalW and compared. Two BoNT/A complex toxoid (more ...)

The two SEB binders along with the BoNT/A complex toxoid binder, P4BF7-3, had no formation of inter-loop disulfide bonds and shared a conserved W31 within CDR1, which are major features of type 3 NAR V originally defined as appearing only in juvenile nurse sharks less than one year old [5,18]. Previously, we discovered that our spiny dogfish NAR V display library derived from adult animals over one year old contained approximately 7% of type 3 NAR V fragments. This may be a unique feature of spiny dogfish shark [20]. These type 3 spiny dogfish shark NAR V differ from ones found in nurse shark, since they exhibited more diverse length and composition with or without a conserved F96 in CDR3 [5,20,22]. Amino acid composition has a great influence on conformation and stability. As the spiny dogfish shark type3 NAR V binders lack the invariant aromatic Y86 and F96 found to participate in forming more rigid CDR3 loops, they likely exhibit different CDR conformational plasticity than type3 NAR V from nurse sharks [22]. Ricin binder, P4RA7-1 is an atypical type 2 NAR V with no formation of inter-loop disulfide bonds due to the lack of Cys in the CDRs (Fig. 2C). One of the BoNT/A complex toxoid binders, P4BH8, is an atypical type 1 NAR V that lacks Cys residues in FR2 and FR4 but contains two CDR3 Cys residues, which allows the formation of an intra-loop disulfide bridge to stabilize CDR conformation. The mechanism of CDR conformation stability for these atypical NAR V should be different from their corresponding typical NAR V and is a topic for future investigation. It is likely that HV2 and HV4 may play extremely dominant roles to stabilize CDR conformation [3,8,23].

Table 2 Equilibrium dissociation constants

The isolated ricin binder, P4RA7-1, bound specifically to intact ricin and ricin A relative to irrelevant targets, CT, BoNT/A complex toxoid and SEB, suggesting it is a ricin A chain binder (Fig. 3A). Its superior specificity makes it an ideal candidate for further affinity maturation. One of the BoNT/A complex toxoid binders (P4BF7-3) was fairly specific to BoNT complex toxoid subtypes, but the other (P4BH8) showed significant non-specific binding to ricin and CT (Figs. 3B and 3C). The measurement of direct binding for the ricin binder (Fig. 3A) and CT binders isolated previously [20] (data not shown) to microspheres with BoNT complex toxoids immobilized to their surface showed no significant interaction, indicating that the sdAb do not as a class nonspecifically bind BoNT complex toxoids. Thus, the non-specific binding observed for any binder selected appears to be intrinsic to just that sdAb, however it points out the necessity to critically evaluate specificity for binders derived by this or similar methods. Nonetheless, for the BoNT binders selected, both had stronger binding to the three tested BoNT complex toxoid subtypes, E, B, and A than to irrelevant targets (Fig. 3B). Generally, they appeared to bind subtypes E and B even more than A (Figs. 3B and 3C), suggesting that toxoid complexes derived by formalin-inactivation may share some common epitopes among subtypes (Figs 3B and 3C). Since P4BF7-3 specifically cross-reacts to other complex toxoid subtypes it may be useful for generic BoNT detection, as opposed to subtype classification. The anti-BoNT sdAb were also tested for their binding to a range of BoNT subtypes using the various toxins as wells as their toxoids immobilized on microspheres. Both BoNT/A binders showed considerable cross-reactivity to other BoNT subtypes; this binding was significantly greater than binding to irrelevant targets (preliminary data not shown). To obtain specific BoNT/A toxin binders, more library screening may be necessary; however these selected binders may find use for sample enrichment purposes, a critical need for such a potent toxin.

Figure 3 Luminex direct binding assay. Microspheres with covalently attached toxins and irrelevant targets were mixed with purified sdAbs at concentrations of 0, 3.4, 10.9, 30.9, 92.6, 277.8, 833.3, and 2500 nM to test the binding affinity. Microspheres were sorted (more ...)

The SEB binders showed very different binding affinity and specificity despite their identical CDR1 and homologous CDR3 sequences (Figs. 2B, 3D and 3E). P1SD3-3 bound to SEB specifically relative to irrelevant targets, while P2SC8 exhibited more non-specific sticky interactions with irrelevant targets, CT and ricin (Figs 3D and 3E). Moreover, P1SD3-3, has ten-fold better binding affinity (kd = 10.2 nM) than P2SC8 (kd = 106.8 nM) (Table 2). A recent affinity maturation study showed that changing S61 to R61 within the HV4 in nurse shark NAR V contributed to increased binding affinity resulting from more van der Waals' contacts on the mature sdAb [8]. The change of K61 to E61 within HV4 in P1SD3-3 might exhibit a similar effect and perhaps resulted in observed higher binding affinity and specificity of P1SD3-3 compared to P2SC8 (Fig. 2B, Figs. 3D and 3E). P2SC8 and P4BF7-3 had identical sequences for most regions except CDR3 and amino acid 48 in HV2. The resulting differential binding reactivity to all tested toxins suggests that CDR3 alone may play a major role to determine the binding specificity in these two clones.

Figure 4 Thermal stability assay. Antibodies were heated at 85°C for various time points up to 60 min and cooled down prior to measuring binding affinity and specificity using luminex direct binding assays. Each antibody at a concentration of maximal binding (more ...)

Of the two, P4BF7-3 appeared to be more stable than P4BH8 (Fig. 4C), suggesting that it might have a more stable CDR conformation. The conformational stability may partially result from the formation of an intra-loop disulfide bridge within CDR3 (Fig. 2A). Interestingly, P4BH8, showed 30% more than initial binding reactivity to target, upon heating for 5 min (Fig. 4C). This phenomenon is often seen in sdAbs, possibly due to better binding of the refolded structure that results from a short period of heating and rapid cooling; this same phenomenon may also be the cause for improvements in specificity [12,13,20].

Both SEB binders exhibited better thermal stability than tested conventional antibodies (Fig. 4F). They showed similar thermal stability profiles, suggesting both share a similar CDR conformation resulting from similar sequences (Fig. 2B).

Figure 5 Luminex Sandwich assay. To measure the utility of sdAbs as tracer antibody, luminex sandwich assays were conducted. Purified antibodies for each tested target along with Ni-SA-PE were mixed with toxins captured by microspheres coated with the capture (more ...)

The sdAbs selected from the new randomized naïve shark library exhibit serviceable specificity and excellent thermal stability, but their binding affinity was weaker than the conventional antibodies tested. Often, affinity maturation is required to obtain high affinity binders. Besides the common approaches described in the background section, 3D structural modeling also provides a powerful tool to reveal more detailed information regarding the change of amino acids within the constant regions and HVs [9]. However, the typical increase in binding affinity has been only 10-fold using these approaches, which would still not produce binders comparable to the best conventional antibodies [9]. Alternatively, multiple copies of sdAbs against the same target can be linked together to generate high avidity structures [2,25,26], which can result in three orders of magnitude higher affinity than the monomeric sdAb. This may be a more attractive way to improve binding reactivity for our sdAb (Table 2).

In the future, we may need to introduce more variation in the CDR3 loop length in combination with the preferred insertion of Cys into defined CDR3 positions. Construction of a more diverse library, containing a greater percentage of conventional type 1 and 2 NAR sequences may allow us to obtain specific binders for each BoNT subtype for detection purpose. Nevertheless, our results indicated that the new semi-synthetic shark library had better CDR3 diversity and better utility than our previously established naturally occurring NAR V display library, suggesting this may be the correct path towards obtaining a limitless source of sdAbs against a variety of toxins for sensor applications.

JL designed the studies and the strategy to introduce the complexity into CDR3 and developed the semi-synthetic hyper-diversified shark SPSL1 library. She selected binders against toxins, purified sdAbs and drafted the manuscript. GA performed the binding assays and helped to draft the manuscript. ERG designed the studies, helped to draft the manuscript and guided the overall project. All authors read and approved the final manuscript.

We would like to thank Dr. Andrew Hayhurst for his expression vector, pEcan22. This work was supported by DTRA JSTO CBD Physical Science and Technology, DTRA JSTO CBD Medical Science and Technology, and the Office of Naval Research. The views expressed here are those of the authors and do not represent the opinions of the U.S. Navy, the U.S. Department of Defense, or the U.S. Government.