Structure and specificity of several triclocarban-binding single domain camelid antibody fragments

Abstract

The variable VHH domains of camelid single chain antibodies have been useful in numerous biotechnology applications due to their simplicity, biophysical properties and abilities to bind to their cognate antigens with high affinities and specificity. Their interactions with proteins have been well-studied but considerably less work has been done to characterize their ability to bind haptens. A high resolution structural study of three nanobodies (T4, T9 and T10) which have been shown to bind triclocarban (TCC, 3-(4-chlorophenyl)-1-(3,4-dichlorophenyl)urea) with near-nanomolar affinity shows that binding occurs in a tunnel largely formed by CDR1 rather than a surface or lateral binding mode seen in other nanobody-hapten interactions. Additional significant interactions are formed with a non-hypervariable loop, sometimes dubbed “CDR4”. A comparison of apo and holo forms of T9 and T10 show that the binding site undergoes little conformational change upon binding of TCC. Structures of three nanobody-TCC complexes demonstrated there was not a standard binding mode. T4 and T9 have a high degree of sequence identity and bind the hapten in a nearly identical manner while the more divergent T10 binds TCC in a slightly displaced orientation with the urea moiety rotated approximately 180º along the long axis of the molecule. In addition to methotrexate, this is the second report of haptens binding in a tunnel formed by CDR1, suggesting that compounds with similar hydrophobicity and shape could be recognized by nanobodies in analogous fashion. Structure-guided mutations failed to improve binding affinity for T4 and T9 underscoring the high degree of natural optimization.

1. INTRODUCTION

In addition to the heterotetrameric antibodies commonly found in most animals, camelids have a special class of antibody that are devoid of light chain [1]. In these heavy chain only antibodies (HcAb) the antigen binding site is found entirely in the variable domain of the heavy chain (VHH), a circumstance that makes them particularly attractive for research, diagnostic or therapeutic applications [2–4]. VHHs, also known as nanobodies, have evolved as independent self-folding domains which facilitates their high-yield production by bacterial fermentation as soluble recombinant binders. This also confers exceptional physicochemical properties to nanobodies, including high thermal stability, tolerance to solvents and extended shelf-life [5, 6]. Owing to the reversibility of their thermal unfolding process, nanobodies can often be heated to temperatures which would irreversibly denature conventional antibodies and regain their function upon cooling [7]. Similarly, the chemical stability of VHHs enables their use in harsh conditions as they are able to maintain their activity in the presence of high concentrations of organic solvents [5, 8, 9].

Despite the lower complexity of their antigen binding site, camelid immunization often produces affinity-matured HcAbs with dissociation constants that are similar to those obtained for conventional antibodies [10]. The small size of their paratope and a larger than average CDR3 often facilitates the recognition of deeply hidden epitopes such as those found in enzyme active sites. Consequently, a large fraction of the HcAbs elicited against enzymes have inhibitory activity [11, 12]. VHHs are not limited to the recognition of concave epitopes and there seems to be few limits to the chemical nature of their specificity. Small molecules (haptens) may be an exception. Haptens are ignored by the immune system unless they are conjugated to carrier proteins. Conventional antibodies raised under these conditions characteristically bind haptens in pockets formed at the interface of the light and heavy variable domains [13]. HcAbs have no light chains and this could be the reason why the antibody response against haptens is dominated by conventional IgGs with a smaller proportion derived from HcAbs, typically ~25% [14]. A number of high affinity nanobodies against small compounds have been reported but failures in their generation are common (reviewed by Bever et al.) [15]. A better knowledge of the structural basis of VHH recognition could help to enable a more rational design of hapten conjugation chemistry and selection strategies but structural information regarding the interaction of nanobodies with small molecules is minimal. Only three such structures have been determined including VHH complexes with two azo dyes (Reactive Red 1 and 6) and the chemotherapy drug methotrexate [13, 16, 17]. These haptens are bound in considerably different manners. The two dyes bind largely on the surface of the protein adopting a “lateral” binding mode [16, 17] while methotrexate, is deeply embedded in the protein in a “tunnel” that is formed by the interior of the CDR1 loop involving interactions with the framework and a variable loop, sometimes referred as CDR4 [13].

Triclocarban (TCC, 3-(4-chlorophenyl)-1-(3,4-dichlorophenyl)urea, structure shown in Table 1) is a 316 Da antibacterial-antifungal agent widely used in household and personal care products. The widespread use of this compound and its bioaccumulation have raised public concern regarding its effects on human health and wildlife [18] and recently, the FDA banned its use in personal care products based on its potential role as endocrine disruptor [19–21]. In a previous work, we reported the isolation of high affinity and extremely thermostable nanobodies against TCC. These nanobodies bind TCC with nanomolar affinity and enabled the development of TCC immunoassays with detection limits of a few ng/mL (nM range). Notably, they lack cross-reactivity with other polychlorinated biphenyl compounds including triclosan [22].

Table 1.

Cross-reactivity (CR, defined as IC₅₀(TCC)/IC₅₀(cross reacting compound) × 100) of TCC structurally related compounds.

Compound	Structure	CR (%) T9	CR (%) T10	Remarks
TCC		100a	100	Antimicrobial
2’-OH TCC		16.9a	15	Metabolite of TCC
Sulfate of 2’-OH TCC		≤0.1a	1.4	Metabolite of TCC
(3-Trifluoromethyl-4,4′-dichlorocarbanilide)		6.97a	15	Antimicrobial
Triclosan		≤0.1a	0.9	Antimicrobial
Dinuron		≤0.1a	≤1	Herbicide
sEHi, #1555		≤0.1a	0.9	sEH inhibitor
sEHi, #1709		≤0.1a	1.5	sEH inhibitor
Carbanilide		1.78a	8.8	Analog
4,4’-Dichlorocarbanilide		13.3a	100	Impurity during TCC synthesis
3,3’,4,4’-Tetrachloro-carbanilide		2.4a	ND	Impurity during TCC synthesis

^aindicates data published by Tabares-da Rosa et al.; ND: not determined; sEH: soluble epoxide hydrolase

In this study, we used three VHHs, designated T4, T9 and T10 against TCC as models to increase the understanding of the interaction of single domain antibodies with small molecules. More specifically, we aimed to explore the structural diversity that may exist between nanobodies raised towards the same hapten, and whether this led to alternate modes of binding. The question of whether these independently generated proteins have similar specificities versus TCC derivatives is also addressed.

2. MATERIALS AND METHODS

2.1. Isolation of nanobodies against TCC

Nanobodies against TCC were isolated and characterized in a previous work [22]. Briefly, two adult male llamas were immunized using intramuscular injection 5 times with 600 µg of TCC coupled to thyroglobulin. One month after the last immunization, blood was collected and RNA was extracted from peripheral blood lymphocytes (10⁷ cells) and used to prepare cDNA. The variable domains were amplified by PCR using specific primers [23, 24] and subsequently cloned into pComb3x (a kind gift from Carlos Barbas Jr III) to generate the phage library in ER2738 E. coli cells (Lucigen Corporation). The resulting library, with an estimated size of 3.4 × 10⁷ independent clones, was panned on TCC coupled to BSA and the bound phage was recovered by competitive elution with decreasing concentrations of TCC. After 3 rounds of panning, selection of high affinity anti-TCC nanobodies was achieved through competitive screening using plates coated with limiting amounts of TCC-BSA in the presence or absence of serial dilutions of TCC. T4, T9 and T10 have been deposited into the Antibody Registry with Antibody IDs of AB_2721894, AB_2721896 and AB_2721895 respectively.

2.2. Nanobody expression and purification

For crystallization experiments except for the holo form of T9, the VHH genes were amplified using the forward primer 5’-GTTACTCGCGGCCCAGGCGGCCATG-3‘ and the reverse primer: 5’-CCACGATTCTGGCCGGCCTGGCCCTATTAATGGTGATGGTGATGGTGTGAGGAGACRGTGACCTGGGTCC-3‘ (Sfi I sites are underlined), cloned between the two SfiI sites of the vector pINQ-BtH6 [25] and sequenced. Mutants versions of T9 and T10 were synthesized by General Biosystems, as well as wild type genes T9 and T10 used for competitive ELISA and cross reactivity assays. These were cloned in the pET 28a(+) expression vector, flanked by the coding sequences of the ompA signal peptide at the 5’ end, and the His₆ tag and hemagglutinin epitope coding sequences.

All VHHs fused to a C-terminal His₆ tag were expressed in transformed BL21(DE3) cells (Invitrogen), grown at 37º C up to an OD 0.6 AU and then induced with 10 µM IPTG during 16 hours at 15° C. Cells were harvested and lysed by osmotic shock. The clarified cell extract was loaded on a Ni-NTA column (GE Healthcare) operated in a BIO-RAD purification system. The column was washed 3 times with 10 mM, 50 mM and 100 mM of imidazole in phosphate saline buffer (PBS) and nanobodies were eluted with 500 mM imidazole.

The holo form of T9 was generated by PCR amplifying from T9 template DNA using the forward primer: 5’-CTAGTGCCATATGGCCGAGGTGCAGCTGGTG and the reverse primer: 5’- ACTGCCCGGGTGAGGAGACAGTGACCTGGGT (NdeI and SmaI sites are underlined). The insert was ligated into the NdeI and SmaI sites in a pTYB2 vectors (New England Biolabs). Expression of the gene followed by cleavage of the intein/chitin-binding domain purification tag left T9 with an additional Pro-Gly on its C-terminus. All purified nanobodies were dialyzed against 10 mM Hepes, pH 7.4 and concentrated using Amicon Ultra-15 Centrifugal Filter units (Merck Millipore) to concentrations shown in Table 2.

Table 2.

Crystallization conditions

Protein	Concentration (mg/mL)	Well buffer
holo T4	25	30% w/v PEG monomethyl ether, 200 mM ammonium sulfate, 100 mM sodium acetate, pH 4.5
apo T9	48	25% w/v PEG MME 3350, 0.2 M ammonium sulfate, 0.1 M Hepes pH 7.5.
holo T9	10	10% PEG 3000, 200 mM zinc acetate, 100 mM sodium acetate, pH 4.5.
holo T10	24	1.3 M ammonium sulfate, 0.1 M citric acid pH 3.0
apo T10	24	1.1 M malonic acid, 0.15 M ammonium citrate, 72 mM succinic acid, 180 mM DL-malic acid, 240 mM sodium acetate, 0.3 M sodium formate, 0.1 M ammonium tartrate dibasic.

2.3. ELISA protocol

Checkerboard assays were performed for each nanobody to determine the optimal amounts of coating antigen and VHH concentration. High-binding ELISA plates (Greiner, Monroe, NC) were coated with 100 µl/well of optimal dilution of TCC-BSA in phosphate-buffered saline (PBS), 4 h at room temperature. After blocking with 1% bovine serum albumin in PBS for 30 min at 37 °C, plates were washed with PBS containing 0.05% Tween-20 (PBS-T). Subsequently, 50 µl/well of optimal amount of nanobody in 0.2% BSA-PBS were added and incubated with 50 µl/well of serial dilutions of TCC (dilution range: 0–1 ug/mL) or TCC related compounds for cross reactivity assay, 1 h at room temperature. After washing with PBS-T, the plates were incubated with 100 µl/well of anti-hemagglutinin antibody conjugate to peroxidase (Sigma-Aldrich, Saint Louis, MO, USA), 1 h at room temperature. After extensive washing, the peroxidase activity was developed by adding 100 μl of peroxidase substrate (0.4 ml of 6 mg 3,3’,5,5’-tetramethylbenzidine in 1 ml of DMSO, 0.1 ml of 1% H₂O₂ in water in a total of 25 ml of 0.1 M acetate buffer, pH 5.5) and incubated for 10 min at room temperature. The enzyme reaction was stopped by the addition of 50 μl of 2N H₂SO₄, and the absorbance was read at 450 nm. Compounds used for cross-reactivity experiments: TCC metabolites (2’-OH TCC and sulfate of 2’-OH TCC), carbanilide, 3-trifluoromethyl-4,4’-dichlorocarbanilide, triclosan, dinuron, 4,4’-dichlorocarbanilide and soluble epoxyhydrolase inhibitors (sEHi, #1709 and sEHi, #1555), were kindly provided by Professor Bruce D. Hammock (Entomology Department, UC Davis, CA, USA).

2.4. Crystallization and structure determination

The holo forms of the nanobodies were generated by adding TCC in DMSO directly to the protein in a 50-fold molar excess, incubating overnight and removing the precipitated TCC by centrifugation. The clarified solutions of the VHH-TCC complexes were then used for crystallization trials. Crystallization trials were performed using a Mosquito crystallization robot (TTP Labtech) at 25º C. Conditions for crystal growth were optimized and high-quality crystals were obtained using hanging drop vapor diffusion, suspending equal volumes of the protein or TCC-protein complex and crystallization buffer. Crystallization conditions for nanobodies are summarized in Table 2. Crystals were mounted for freezing in 20% ethylene glycol, 80% well solution in all cases. Statistics associated with data collection and refinement are given in Table 1.

The data collected for the holo form T9 were phased via molecular replacement using PHASER with a camelized human VH domain (PDB: 1OL0) as a search object. For the apo form of the nanobody, phasing was carried out using a subunit of the T9 holo structure with the TCC removed. For T4 holo form and T10 holo form, phasing was carried out using molecule A from the holo T9 structure (PDB 5VM0) with the non-protein atoms removed as a search object. Finally, a preliminary structure of the holo form of T10 was used in PHASER as a search object for molecular replacement for the T10 apo form. Final refined coordinates and structure factors have been deposited into the Protein Data Bank with accession numbers as detailed in Table 3.

TABLE 3.

Data collection and refinement statistics. Values in parentheses are for high resolution shells

	T4/TCC	T9 apo	T9/TCC	T10 apo	T10/TCC
PDB entry	5VL2	5VLV	5VM0	5VM4	5VM6
Beamline	APS 24-ID-E	ALS 8.3.1	SSRL BL 9-2	SSRL BL 9-2	SSRL BL 7-1
Spacegroup	P21	P4322	C2	C2	P21
Unit cell	a = 98.8 Å b = 55.3 Å c = 103.4 Å β = 111.7°	a = b = 48.8 Å c = 119.3 Å	a = 101.1 Å b = 52.5 Å c = 57.5 Å β = 113.7°	a = 161.6 Å b = 93.3 Å c = 110.6 Å β = 104.1°	a = 36.8 Å b = 36.4 Å c = 44.6 Å β = 111.2º
Resolution (Å)	1.9 (1.94-1.90)	1.35 (1.39-1.35)	1.8 (1.84-1.80)	1.9 (1.95-1.90)	1.5 (1.54-1.50)
Observed / unique reflections	223967 / 80496	223543 / 32653	87888 / 25168	350228 / 120758	45083 / 17461
Rmerge (%)	8.1 (97.4)	3.2 (65.8)	6.1 (31.8)	7.7 (28.1)	7.1 (40.0)
Completeness (%)	98.0 (96.4)	100 (99.9)	97.0 (86.9)	96.6 (90.8)	98.2 (95.0)
I / σ (I)	11.6 (2.2)	24.5 (2.4)	14.7 (5.2)	10.0 (4.3)	10.4 (2.7)
Rcryst (%)	21.9	16.2	20.4	19.7	17.2
Rfree (%)	26.3	18.9	24.3	24.1	19.3
rmsd bond length (Å)	0.013	0.009	0.016	0.013	0.011
rmsd bond angle (°)	1.65	1.39	1.71	1.55	1.59
Average overall B-factor for each molecule (Å2)	A=20.8; B=21.9; C=22.9; D=20.7; E=26.1; F=35.0; G=34.1; H=34.5;	25.0	A=19.3; B=17.9	A=16.9; B=14.0; C=17.0; D=14.8; E=14.2; F=14.5; G=14.1; H=17.5; I=16.7; J=17.4; K=17.3; L=16.5;	8.9

3. RESULTS AND DISCUSSION

3.1. Structural studies

Crystal structures of apo (T9, T10) and holo (T4, T9, T10) forms of the triclocarban binding nanobodies were determined at relatively high resolution (Table 3) and found to adopt the expected immunoglobulin fold (Figure 1). All models have an uninterrupted peptide chain although some residues are missing at the N- and C-termini. Nevertheless, several structures have visible C-terminal histidines derived from the His₆ tag, particularly in the case of apo T10 which had clear density for five of the six histidines in the tag. Unambiguous density was observed for the binding sites in all cases as well as the TCC ligand (Figure 2). As shown in Table 3, asymmetric units in holo T4, holo T9 and apo T10 contained two or more protein molecules. Pairwise comparisons between the C_αs revealed that there were few differences within the holo T9 pair of molecules (r.m.s.d. = 0.20 Å) and within the apo T10 molecules (r.m.s.d.s ranged between 0.10 Å and 0.24 Å). The r.m.s. deviations in the range 0.24 Å and 0.71 Å between holo T4 molecules indicated more variability. In this case, molecule E was consistently different from the others with a relatively large deviation in the position of the N-terminal residues. Where comparisons are drawn between different structures, the model with the lowest overall B-factor (as shown in Table 3) was chosen to be representative.

Figure 1.

The structure of A) T9 (blue) and B) T10 (green) bound to TCC (colored by atom) in similar orientations. The 84% sequence identity between T4 (not shown) and T9 allow TCC to bind in a nearly identical manner. T10 binds the hapten in a different orientation with the urea moiety rotated approximately 180º along the long axis of the molecule. CDR1, CDR2 and CDR3 (defined in Figure 4) are colored red, orange and yellow respectively. Regions of the non-hypervariable loop (sometimes referred to as CDR4) which interact with TCC are shown in pink. C) An overlap of T4 (purple), T9 (blue) and T10 (green) holo structures shows the high similarity between T4 and T9 and structural divergence of T10 within the binding site as well as the different modes of binding for TCC. Carbon atoms in the TCC are colored identically to the main chain trace.

Figure 2.

Details of TCC binding to T9 (which is substantially similar to T4) and T10. Side chains directly interacting and selected main chain atoms are displayed. In both cases, unbiased F_o-F_c density (grey) corresponding to the TCC is contoured at 3σ. Sites of mutation are shown in red text. a) TCC (grey carbons) binding to T9 (blue main chain and white carbons). Hydrogen bonds that determine the orientation of the urea bridge are shown in yellow dashes. b) TCC (grey carbons) interactions with T10 side chains (green main chain and white carbons) in an orientation similar to Figure 2a. Also shown are hydrogen bonding (yellow dashes) to the urea is responsible for the different mode of binding. Note that electron density for Ser100 indicated two well-defined alternate side chain conformations.

In all three cases of the holo structures, the TCC binds in a tunnel primarily formed by CDR1 and to a lesser extent, CDR3 with the di-chloro ring directed out of the nanobody binding site (Figure 1). This is in agreement with the fact that the chlorine in the 4’ position of the di-chloro ring of TCC was the position used to conjugate the hapten to the carrier protein for immunization, exposing the mono-chloro ring to the solvent. Another sequence which contributes to formation of the binding pocket is found in a non-hypervariable region of amino acids between CDR2 and CDR3 (framework 3 (FR3)) composed of residues 75–82 in T4 and T9 and 80–82 in the divergent T10. This region has also been shown to be critical in the interaction of methotrexate with a VHH domain where an affinity drop of 1000-fold was observed when some of these residues were absent [13]. Significant binding interactions are also made with residues near the N-terminus in all cases. There are no contacts with CDR2 in T4 and T9 and only a weak contact with the side chain of Trp55 on CDR2 in T10.

The binding of TCC in a “tunnel” is generally similar to that seen between methotrexate and a VHH domain [13]. It is different from the “lateral” or surface mode of binding observed between a VHH domain and the azo-dyes reactive red 1 and 6 [16, 17]. A comparison of these structures shows that the architecture of the “tunnel”, composed largely of residues in CDR1, is conserved between nanobodies binding methotrexate and TCC (Figure 3). In comparison, reactive red 1 (MW 612) and reactive red 6 (MW 628 Da), which are bound in a lateral fashion, are larger, more compact and charged than TCC (MW 315 Da) and methotrexate (MW 454 Da). In addition, the latter are both elongated and highly hydrophobic compounds. Based on this, it appears that ligand shape, size and solubility are all properties which are likely to influence whether a nanobody utilizes one binding mode or the other.

Figure 3.

A superpositioning of the T9-TCC complex (dark blue main chain, grey carbons), a methotrexate (MTX) binding nanobody (PDB accession 3QXV, light blue main chain and carbons) and the RR1-binding nanobody (PDB accession 1I3U, magenta main chain and carbons). CDRs are colored in T9 according to the scheme used in Figure 1. The smaller, elongated TCC and MTX adopt a common “tunnel” mode of binding while the larger RR1 is bound using a surface or “lateral” binding site.

Changes induced by TCC binding are generally minor, in particular when the binding sites are compared. The r.m.s. deviations between the C_αs of the apo and holo forms of T9 is 0.47 Å indicating that TCC binding induces little change in the nanobody structure. On the other hand, a larger change of 0.83 Å is seen between the apo and holo T10 structures. The largest main chain shifts in T10 are seen distal to the binding site suggesting that crystal packing interactions may be having a larger influence than the binding of the ligand. There is, however, an ~8 Å movement that is seen in the side chain of Arg28 which is clearly displaced by the outer, dichloro ring of TCC. Interactions with TCC also induces an ordering of Val4 which is not visible in the electron density of the apo form of the protein. Almost all of the interactions between TCC and the nanobodies are hydrophobic due to the nature of the hapten so the comparison of observed ordered water molecules in apo and holo structures is interesting since it would suggest an entropic contribution to binding. In the case of T9, there is only one ordered water molecule in the binding tunnel (identical in both asymmetric units) which is displaced by TCC. Upon TCC binding to T10, there is actually an ordered water molecule which is introduced into the binding site (discussed below), (Figure 2B).

Interactions between TCC and T4 and T9 are nearly identical owing to the high sequence conservation of the residues that interact with the ligand in both proteins as well as the overall 84% sequence identity between the proteins (Figure 2A). Fourteen of the 17 positions that contact the TCC in both proteins are identical and relatively conservative substitutions are noted at the two of the three other positions (Figure 4). Position 79 is an alanine in the case of T4 and lysine in T9 and the interaction with TCC is confined to main chain atoms in both cases. Since TCC is a largely hydrophobic molecule, a large fraction of the interactions involve hydrophobic side chains and specificity is therefore largely based on the shape of the binding site. Notably, the binding pockets of T4 and T9 are mainly constructed of aliphatic amino acid side chains. There is no aromatic stacking with the TCC rings. The main chain amide nitrogens of Tyr31 in T9 or His31 in T4 are engaged in hydrogen bonding donating to the urea oxygen, and the Thr100 side chain acts as a hydrogen bond acceptor for both nitrogens of the urea bridge (Figure 2A).

Figure 4.

Sequence alignment of T4, T9 and T10 with complementarity determining regions (CDRs) designated. Also shown is the non-hypervariable loop (NHVL) region which makes further important interactions with TCC as seen in Figures 1A and 1B. Numbering is given for the T4 sequence. Residues involved in contacts with the TCC molecule are highlighted according to contact areas with the nanobody: 1–10 Ų buried is yellow, 10–20 Ų buried is orange, >20 Ų buried red.

The TCC occupies the same general volume in T10 as it does in T4 and T9 with the dichloro ring system directed outwards toward the solvent (Figure 1). The orientation of the urea however, is rotated approximately 180º along the long axis of the molecule relative to that observed in the holo structures of T4 and T9 (Figure 1C). The orientation of the outer, dichloro ring remains similar. Along with the shape of the binding site determined by a number of hydrophobic residues (Figure 2B) this difference is largely due to the variations in hydrogen bonding interactions associated with the urea moiety. In T10 the urea nitrogens exhibit hydrogen bonding interactions with main chain carbonyls of residues 28 and 29 although the geometry is poor. The TCC urea oxygen interacts with a sequestered water molecule which in turn is hydrogen bonded to the side chain of Ser100 which occupies two conformations (Figure 2B). The orientation of the urea relative to T4 and T9 is therefore dictated by main chain interactions at position 31 in the case of T4 and T9 and positions 28 and 29 for T10.

A comparison of the T4, T9 and T10 structures sheds light on the relative strength of binding. Previously-determined dissociation constants for T4 and T9 suggest two modes of binding with K_D’s of 3.77 and 0.98 nM (for T4) and 1.60 and 1.10 nM (for T9) indicating that they bind with similar affinity [22]. No dissociation constant is available for T10 but a half maximal inhibitory concentration (IC50) for TCC has been determined for all three nanobodies: 20, 19 and 78 nM for T4, T9 and T10, respectively, suggesting a lower affinity for T10. In all cases, the hydrogen bond donors and acceptors on the ligand were completely satisfied with complementing groups on the protein. PISA was used to calculate the average interface surfaces for each complex (T4: 333 Ų, T9: 322 Ų and T10: 291 Ų) and shows some correlation with the observed binding constants [26]. More significant differences are seen when potential pairwise van der Waals interactions with the TCC (<4.0 Å) are enumerated: T4, 76; T9, 77 and T10, 57. The ligand interface surfaces calculated for the TCC binding nanobodies fall on the low side of the range of 346 Ų to 428 Ų found in structures complexed with reactive red dyes and methotrexate (PDB accession numbers 1I3U, 1QD0, 3QXT and 3ZXV), some of which have poorer binding constants. On the other hand, the percentage of TCC surface buried (T4: 68%, T9: 66% and T10 60%) compares favorably with the above referenced nanobodies which have a range of 33 to 63%. Improvements in nanobody binding affinity have been associated with increased buried surface in the case of methotrexate binding where the contacts with the non-hypervariable loop were improved [13].

3.2. Cross reactivity

Cross reactivity for various TCC derivatives and several antimicrobials with chemical structures similar to TCC were determined for T10. Two structurally-related, soluble epoxide hydrolase inhibitors with a central urea linker were also examined. These were compared to a number of compounds which had previously been examined for T9 using a competitive ELISA to determine ratios of the IC50s between TCC and each cross-reacting compound (Table 1) [22]. A ten- to more than a thousand-fold effect is found for compounds with changes or removal of ring substituents and alterations of the urea bridge. Despite the two different modes of binding with the two nanobodies, there was remarkably little difference when comparing these parameters. Assuming that these derivatives bind in a manner largely similar to TCC, steric conflicts and loss of favorable hydrophobic contacts are the most likely explanations for the cases in which ring substituents are added and removed respectively. The addition of a relatively small hydroxyl at the 2’ position is not severely disruptive and could be accommodated by a slight shift in the N-terminus and/or CDR1 due to potential interactions with the side chains of residues at positions 4, 26 and 30. Placement of a larger sulfate group at this position abolishes binding. Substitution of an ether oxygen for the urea bridge as found in the antibacterial/antifungal triclosan abolishes binding as does the lack of a complete urea in dinuron. Aside from changes in the substitution of the rings which alter the shape of the molecule, the hydrogen bonding interactors on the protein will no longer be compatible with the bridge. The most notable difference in binding is found with 4,4’-dichlorocarbanilide which is missing the chloro substituent at the 3 position relative to TCC which is more disruptive in the case of T9 than T10. This substituent makes 6 van der Waals interactions with T9 (Arg101, Ala102, Pro114 and Ile115) whereas in the case of T10, it only engages in 3 with the Tyr120 side chain (Figure 2). There is volume available to allow some degree of conformational flexibility in the Tyr120 side chain. Moreover, the interactions engaging the 3-chloro in T9 are either from the main chain or are relatively constrained side chains.

3.3. Attempts to improve TCC affinities

As structures show, T10 binds TCC using a largely different set of interactions resulting in a lower apparent affinity than T4 or T9. To investigate the possibility of improving TCC binding, three T10 point mutants were generated based upon interactions observed in the T10 holo structure (Figure 2B). The first two attempted to create π-stacking interactions between the protein and the dichlorophenyl group in TCC. Leu31 was replaced by two aromatic amino acids to produce a T10-Leu³¹Phe and a T10-Leu³¹Tyr mutant. A third mutation involving Ser100 was made to form a direct interaction with the protein. Ser100 exists in two clearly defined different conformations and each engages in a hydrogen bonding interaction with a sequestered water molecule which in turn interacts with the TCC urea oxygen. Replacement of Ser100 for Asn could permit the formation of a direct hydrogen bond between TCC and the protein, simultaneously displacing the ordered water molecule.

Mutant and wild type T10 were expressed and purified in a manner similar to the wild-type protein and a competitive ELISA was performed to analyze affinity of nanobodies against TCC. None of the mutants improved the affinity of T10 for TCC (Figure 5B). The IC50 for T10-WT was 22.6 ±2.0 ng/mL. Only estimated values could be determined for T10Leu³¹Tyr (232 ng/mL) and for T10-Leu³¹Phe (315 ng/mL) because inhibition curves were incomplete as a result of limitations in the solubility of TCC. TCC binding to T10-Ser¹⁰⁰Asn was severely disrupted and no IC50 estimation was possible.

Figure 5:

Competitive ELISA with T9 and T10 mutants. Representative curves showing the inhibition of the nanobody binding to TCC-BSA with increasing concentrations of TCC. A) T9-WT (black squares) T9-Thr¹⁰⁰Asp (asterisk). B) T10-WT (black squares), T10-Leu³¹Tyr (open circles), T10-Leu³¹Phe (black triangles), T10-Ser¹⁰⁰Asn (asterisk). Each point represents the average of three replicates.

Similar results were obtained when trying to improve T9 affinity with the mutant T9-Thr¹⁰⁰Asp. In this case, the threonine side chain acts as a hydrogen bond acceptor for both nitrogens of the urea bridge. We substituted this threonine for aspartate in order to have a carboxylate which is capable of functioning as a bidentate hydrogen bond acceptor from the urea nitrogens. Nevertheless, IC50 for T9-Thr¹⁰⁰Asp was 33.0 ± 3.0 ng/mL, approximately four times larger than T9-WT (8.7 ± 2.5 ng/mL) (Figure 5A). Although the mutations in both T9 and T10 were not isosteric, modeling using the holo structures suggested that the extra steric volume might be accommodated by loss of a water molecule (in the case Ser100 in T10) or adjustments to the main chain and/or side chain conformations. Without structures of the mutants, we speculate that this flexibility was not possible. Subsequent structures of the apo forms of T9 and T10 revealed that there was little flexing of the protein upon TCC binding. This lack of flexibility is also suggested by the cross-reactivity data that shows that even minor changes in the structure of TCC affect affinity.

4. CONCLUSIONS

With only few small molecule-VHH structures available, the structural information reported here represents an important contribution in understanding the interaction of single domain antibodies with haptens. The three holo structures present here provide additional evidence showing that nanobodies bind haptens in variable manners. In all instances, the hapten is embedded in a tunnel within the VHH in a fashion similar to that found for methotrexate. In both cases, the buried surfaces are large and hydrophobic, comparable to those observed in conventional antibody-hapten complexes. This accounts for the fact that the dissociation constants of the anti-TCC and methotrexate nanobodies are both in the low nanomolar range. Approximately half of the interactions observed in the VHH-TCC complexes occur with main chain atoms implying that sequence diversity may not be necessarily translated directly into a similar diversification of the binding cavity. These constraints impose a restriction to the actual number of haptens that can be recognized in this “tunnel-like” fashion which may be one of the reasons explaining the high frequency of failures in the generation of anti-hapten nanobodies. Finally, in the case of both T9 and T10, the lack of improvement in any of the mutants is a testament to the power of the affinity maturation process that occurred during the immunization of the animal resulting in highly optimized nanobody-hapten interactions.