Efficient generation of single domain antibodies with high affinities and enhanced thermal stabilities

Single domain antibodies (sdAbs), made of natural single variable regions of camelid or cartilaginous fish antibodies, or unpaired variable regions of mouse or human IgGs, are some of the more promising biologic modalities. However, such conventional sdAbs have difficulties of either using unwieldy animals for immunization or having high affinity deficiencies. Herein, we offer a versatile method to generate rabbit variable domain of heavy chain (rVH) derived sdAbs with high affinities (K D values of single digit nM or less) and enhanced thermal stabilities (equal to or even higher than those of camelid derived sdAbs). It was found that a variety of rVH binders, including those with high affinities, were efficiently acquired using an rVH-displaying phage library produced at a low temperature of 16 °C. By a simple method to introduce an additional disulfide bond, their unfolding temperatures were increased by more than 20 °C without severe loss of binding affinity. Differential scanning calorimetry analysis suggested that this highly efficient thermal stabilization was mainly attributed to the entropic contribution and unique thermodynamic character of the rVHs.

phage. The temperature at which phages are produced in Escherichia coli (E. coli) is presumed to be the key to achieving this. Previously, an attempt was made to acquire rVH binders using an rVH-displaying phage library produced at 25 °C 16 , which is lower than the conventional temperature for rabbit scFv-displaying phages (30 °C or 37 °C) [17][18][19] , resulting in obtainment of only weak binders. Thus, in this study, the temperature was further lowered to investigate its impact on the rVH display level and resulting acquisition of rVH binders.
In order to utilize rVHs for various applications, high thermal stability would be one of the most desired properties. There are many papers that shows improvement of physicochemical properties by phage display method. In such approaches, mutations are sometimes involved at or near the CDRs 20,21 . To avoid the risk of affinity change by such mutations, we considered the enhancement of thermal stabilities of rVHs by introducing covalent bond into the fixed position of deep inside the framework. Other groups reported introduction of an additional (artificial) disulfide bond at the residues 54 and 78 of camelid VHH (IMGT numbering) [22][23][24] . Most of the VHHs were successfully stabilized by the disulfide bond with relatively small negative effect on the binding affinity to the antigen, while some VHHs lost their antigen binding abilities 24 . Therefore, we thought it intriguing to investigate the impact of the disulfide bond on the thermal stabilities and affinities of rVHs, and to learn what kind of factors control them.
In this study, we attempted to acquire a wide variety of rVHs against tumor antigens HER2 and HER3 through phage production at temperatures lower than 25 °C. The obtained rVHs were then characterized by antigen binding affinity and thermal stability. Finally, the new disulfide bond was introduced into obtained rVHs and its impact on the thermal stabilities and affinities was examined. We provide a new platform to generate potent and highly stable rVH derived sdAbs for various applications, including therapeutic uses.

Results
Phage production and acquisition of antigen specific rVHs. Our rVH binder acquisition process is summarized in Supplementary Figure S1. Briefly, we immunized rabbits with HER2 and HER3 antigens and confirmed the increased antibody titers of these antigens in the immunized rabbits ( Supplementary Fig. S2). We used a total of 1.9 × 10 8 spleen and lymph node cells to amplify VH genes from these rabbits. rVH genes were amplified with designed primers (Supplementary Table S1) and inserted into phagemid vector. A total of 8.0 × 10 8 transformants were obtained and it was expected that they would cover a wide variety of input rVH genes. For rVH-displayed phage production, we first cultivated the obtained transformants at several temperatures to investigate which temperature maximizes rVH display level. As shown in Fig. 1a, cultivation at 16 °C gave the strongest . The rVH-gIIIp fusion protein was detected by anti-E-tag antibody and the amount of detected rVH-gIIIp fusion protein was correlated to the display level of rVHs. Fusion protein, which lacks an rVH portion, was also detected as a below band of the intact fusion protein. (b) Phage recovery rates after panning against HER2 or HER3. Gray bars indicate phage recovery rates (ratio of output to input phage) after each round of panning and black bars indicate phage recovery rates after the third round of panning without antigen.
band intensity corresponding to rVH-gIIIp (gIII coat protein of M13 bacteriophage) fusion proteins, which are indicative of rVHs being displayed on the phage (Fig. 1a, upper band). The band become weaker as the cultivation temperature increased to 20 or 22 °C and was not detected at 25 °C. Thus, we adopted 16 °C as the cultivation temperature to produce an rVH-displaying phage library. The lower bands shown in Fig. 1a were considered to be impurities of rVH-gIIIp fusion proteins lacking the rVH region because their molecular weights (about 15 kDa) were consistent with the difference of molecular weight between the upper and lower bands. This consideration was supported by the fact that such an intense band was not observed for the control phage VCSM13, which is composed of only native gIIIp ( Supplementary Fig. S3).
Starting from the prepared rVH phage library, phage recovery rates (numbers of output phage/input phage) increased step-wisely after every round of panning with each antigen (Fig. 1b). In the third round, remarkable differences in the phage recovery rates were confirmed between the panning with and without antigens. These results indicated that antigen binding rVHs were concentrated from a vast number of library clones. After three rounds of panning, about 300 output clones were screened for their binding abilities to antigens by Enzyme-Linked ImmunoSorbent Assay (ELISA), and 55 and 125 hit clones were obtained for HER2 and HER3 ( Supplementary Fig. S1), respectively. These hit clones were subjected to sequence analysis and assessment of concentration-dependent binding to antigens ( Supplementary Fig. S4). Non-specific binders were eliminated by counter screening for bovine serum albumin (BSA). Finally, eight rVHs were obtained respectively for both HER2 and HER3 (whose names start with H2 and H3 as Fig. 2a). Introduction of additional disulfide bond to rVHs. Aiming for thermal stability enhancements of the rVHs, we considered the introduction of an additional disulfide bond between residues 54 and 78 (Gly/Ala and Ile, blue-colored in Fig. 2a and b) based on the previous successful results in VHHs [22][23][24] . Because Cys is a hydrophobic amino acid, the residues of Cys mutations should be buried in a structure so as not to alter the hydrophobicity of the structure's surface. In fact, accessible surface areas (ASAs) of residues mutated to Cys were less than 20% in the previous studies that reported the introduction of artificial disulfide bonds 22 . In order to estimate the ASA of residues for Cys mutation in the rVHs, we constructed model structures of five obtained rVHs using rabbit antibody structures that were available from a protein data bank (PDB) as a template (Supplementary Table S3). The estimated ASAs of the model structures were less than 10% (Supplementary Table S4), indicating that these residues were buried deep inside in the model structures. In order to investigate if the disulfide bond between C54 and C78 (C54-C78) would invoke structural alteration, we calculated the root mean square deviation (RMSD) between each model structure of rVHs with and without C54-C78. The Cα RMSDs of the five rVHs showed values similar to that of VHH compared to the α-subunit of human chorionic gonadotropin (VHH hCG , PBD ID: 1HCV). Because VHH hCG was thermally stabilized by C54-C78 without loss of binding activity 22 , the introduction of C54-C78 was unlikely to have a negative impact on the rVHs' binding affinities to antigens. Based on these considerations, in this study we attempted to introduce an additional disulfide bond into the wild type rVHs by Cys mutation of residues 54 and 78.
The rVHs whose residues 54 and 78 were mutated to Cys (C54-C78 mutant) were prepared by the same method as wild type rVHs and those purities were confirmed by SDS-PAGE analysis ( Supplementary Fig. S5). It was not clearly observed that purification yields were improved by disulfide bond introduction under our preparation conditions using E. coli. In the case of H2-1-1, which was prepared using mammalian cell, purification yield of its C54-C78 mutant was increased to 3-or 7-fold compared with wild type. The formation of C54-C78 was experimentally confirmed by MS analysis after chymotrypsin digestion (Supplementary Table S4). Chymotrypsin digestion of mutant rVHs produced peptide fragments linked with C54-C78 or C23-C104 as expected. Neither undesired peptide fragments linked with other combinations of disulfide bond nor those with free Cys were detected. These results indicated that both of the two disulfide bonds were correctly formed as we had designed.
Physico-chemical properties of C54-C78 mutant rVHs. The antigen binding affinities of C54-C78 mutant rVHs were evaluated by SPR. Their K D values, except for H3-9, were within several-fold of their respective wild type counterparts ( Fig. 3a and Table 1). The affinities of rVHs in Fig. 3a (H2-2-2 and its C54-C78 mutant) were also evaluated with BioLayer Interferometry method (BLI method). Using BLI method, which is completely different biosensor system from SPR, we could obtain similar fold change in K D values as SPR between H2-2-2 and its mutant (6.3 fold for BLI, 6.8 fold for SPR, Supplementary Fig. S7). The thermal stabilities of C54-C78 mutants were evaluated ( Table 1). The T peak values of all mutants were more than 20 °C higher than those of corresponding wild type rVHs, and remarkably, such T peak increases were much larger than those of previous results by VHHs [22][23][24] . rVHs can be highly thermally stabilized by C54-C78 introduction without severe loss of binding affinities.
In order to obtain information about the major factors that contributed to the high thermal stabilization mentioned above, the thermodynamic parameters and T m , at which the Gibbs free energy change (ΔG) becomes zero, were determined for H2-2-2 and its C54-C78 mutant by DSC ( Fig. 3b and Table 2). In comparison, those of VHH hCG were estimated based on the published data 22 (Table 2). With the obtained thermodynamic parameters and T m , ΔG values at each temperature (ΔG(T)s) were illustrated in Fig. 3c. The introduction of C54-C78 increased ΔG of mutant of H2-2-2 and VHH hGC to the same extent at T m of wild type (T m W ). As for T m , the T m increase (ΔT m ) of H2-2-2 was more than 20 °C (from 35.0 °C to 58.1 °C), while that of VHH hCG was 10 °C (from 46.0 °C to 56.0 °C). Comparing ΔG(T) curves of mutants and wild types, C54-C78 introduction did not have large impact on the shape of the ΔG(T) curve itself. Regarding the enthalpy and entropy change accompanied by the thermal unfolding (ΔH and ΔS), the mutant of H2-2-2 showed the values of 162.2 kJ/mol and 489.7 J/mol/K, respectively. These values were less than half those of mutant of VHH hCG (369.2 kJ/mol and 1121.7 J/mol/K, respectively). In terms of the influence of the C54-C78 introduction on the ΔH and ΔS, compared to their wild types both ΔH and ΔS were slightly decreased for the mutant of H2-2-2 while were increased for the mutant of VHH hCG . The change in the heat capacity (ΔC P ) of the mutant of H2-2-2 was 1.8 kJ/mol/K, which was one-third that of the mutant of VHH hCG . The C54-C78 introduction led to the slight increase in ΔC P of both mutants of H2-2-2 and VHH hCG compared to their wild types. Figure 3d shows the differences in the change of enthalpy, entropic term, and free energy (ΔΔH, TΔΔS and ΔΔG) between mutant and wild type of H2-2-2 at various temperatures. ΔΔH was negative but largely compensated by TΔΔS, resulting in a positive ΔΔG at all indicated temperatures.

Discussion
In this study, rVHs were shown to have the potential for specific binding to antigens with sub-nanomolar K D values. Based on our knowledge, this is the first report to obtain such high affinity binders composed of an unpaired variable region [3][4][5][6][7][8] . As correlation was observed for representative five rVHs between binding efficiencies of ELISA ( Supplementary Fig. S4) and dissociation constants of SPR (Table 1), we consider that other ELISA binders such as H2-2-1 or H3-14, which are as efficient as H2-2-2 and H3-15, could also have smaller dissociation constants as those of H2-2-2 and H3-15. Physical stability of therapeutic proteins in solution is governed mainly by the combination of conformational stability that corresponds to the free energy difference between native and denatured states and colloidal stability that reflects the dispersion state of the protein molecules 25,26 . Upon considering the acquisition of the rVHs, their reduction in conformational stabilities was presumed because they lacked partner VLs [13][14][15] . In fact, the difference in the free energy between scFv and unpaired VH was calculated to be 9.0 kJ/mol from the denaturation curve in the previous report 13 . This large decrease in the stability elicited our concern that phage production at high temperatures would cause inefficient display of rVHs on the phage. The phage produced at 25 °C, which was used in the previous study 16 , did not show a detectable level of rVH display (Fig. 1a). While, the rVH display level was enhanced by lowering temperature and maximized at 16 °C. Using a rVH-displaying phage library produced at 16 °C, we could obtain a variety of HER2 and HER3 binders. Some of rVHs were poorly produced in E. coli and thermally unstable ( Supplementary Fig. S6). This result implied that the lowered temperature contributed to rVH binder acquisition by enhancing soluble expression 27 of rVHs fused to gIIIp and/or suppressing thermal unfolding. For industrial applications as therapeutic agents, a simple and universal method is strongly needed to enhance the thermal stability of rVHs. This study showed that the introduction of C54-C78 increased unfolding temperatures of rVHs, and surprisingly their shifts of unfolding temperature were much larger than those of VHHs 22-24 (24.0 °C for rVHs vs 9.0 °C for VHHs on average, Fig. 4). The thermal stabilities of mutant rVHs were comparable to those of mutant VHHs. We considered our approach is applicable not only to our representative rVHs but also any other rVHs because almost all of the rVH frameworks (80-90%) have high sequence similarity due to adopting only one germ-line gene segment 9, 11, 12 and the beneficial mutations in the framework could be shared among the rVHs. To be employed for various applications, expression yield in E. coli might be one of the important factors for sdAbs. Improvement of purification yield was not clearly observed by disulfide bond introduction under our preparation conditions, however, in the case of H2-1-1, purification yield increased by 3-or 7-fold due to disulfide bond introduction when it was prepared using mammalian cell. This result supports the correlation between thermal stabilities and purification yields as suggested in Supplementary Figure S6. Besides the conformational stability, protein yield could be influenced by various factors including mRNA transcription and translation efficiency, folding efficiency, solubility, and so on. Therefore, by optimizing preparation conditions including modification of expression vectors and E. coli strains, introduction of disulfide bonds could increase purification yield using E. coli expression system.
We next examined the thermodynamic parameters to reveal the origin of the larger T m increase in the C54-C78 mutant of H2-2-2 compared to the mutant of VHH hCG . The ΔG increase at each T m W (ΔΔG(T m W )) accompanied by the mutation was 9.
Here ΔG at T m W is expressed as ΔG°, while ΔG becomes zero at T m W , ΔH is indicated as Equations 2 and 3 from Equation 1.  Fig. S8). The smaller ΔC P apparently resulted in the smaller ΔG temperature dependence, and ΔC P of the mutant of H2-2-2 resulted in a ΔT m of 23 °C, which is higher than the ΔT m (19 °C) calculated using ΔC P of the mutant of VHH hCG . These estimations indicated that the small ΔS and ΔC P values of the mutant of H2-2-2 could be causes of its higher ΔT m .
As for the ΔG increase of the mutant of H2-2-2, Fig. 3d suggested the contribution of a large negative ΔΔS. In order to have further thermodynamic insight into the effect of C54-C78 introduction to rVHs, we compared experimental ΔΔS of H2-2-2 with its theoretical entropy change of unfolded state by C54-C78 formation (ΔS calc ). In the classical chain-entropy model, an enhancement of stability by the disulfide bond formation is primarily considered to be attributed to ΔS calc [28][29][30][31] . The ΔS calc can be calculated using an equation 28 :  In this study, we obtained rVHs without their partner rVLs by displaying only rVHs on the phage. Isolation of VH from Fv accompanies the exposure of the VL-interacting surface, which is generally composed of hydrophobic amino acids, to solvent. The exposure of the VL-interacting surface could cause intermolecular interactions and thereby decrease colloidal stability, leading to undesirable non-specific oligomerizations and eventually aggregations 26 . The features of VL-interacting surfaces of rVHs obtained in this study were next investigated and compared with those of rVHs that were previously obtained as Fv [32][33][34][35] . The amino acid residues located on this surface (cyan-colored in Fig. 2b) are listed in Supplementary Figure S9a. Comparing the ASAs of our rVHs and rVHs from Fv, no significant difference was found in the ASA of non-polar groups (ASA non-pol , Supplementary  Fig. S9b). On the other hand, our rVHs showed a clear tendency of larger ASA polar groups (ASA pol ) than rVHs in Fv. Surprisingly, this significant difference was mainly attributed to only one amino acid at the residue 120. Most of the obtained rVHs had Gln at residue 120 while all rVHs in Fv had Pro. Compared with the Pro at residue 120, Gln had a smaller ASA non-pol (30 Å 2 -40 Å 2 smaller) and larger ASA pol (80 Å 2 -100 Å 2 larger) ( Supplementary  Fig. S9c), providing a higher hydrophilic surface area at the VL-interacting surface of our rVHs. These results might originate from the elimination of rVHs with hydrophobic VL-interacting surfaces during the panning step using hydrophobic magnetic beads and microtubes. Considering that the VHHs, which are a natural single domain variable region of camelids, also adopt Gln [22][23][24] , Gln at the residue 120 of sdAbs might generally be advantageous from the point of colloidal stability. Further investigation of colloidal stabilities, three-dimensional structures of rVHs, and their molecular states in highly concentrated solution 36 will clarify the general rules of VH stabilization.
In conclusion, rVHs proved to have sufficiently high affinities that could not be achieved by VH from mice and human IgGs. The low thermal stability concern of rVHs was eliminated by introducing an additional disulfide bond. Thus rVHs are a promising new source of sdAbs, and their higher availability than conventional sdAbs would enable more frequent usage of sdAbs in various applications as therapeutic use.

Materials and Methods
Rabbit immunization. Three Japanese white rabbits (Inoue-shouten, Takasaki-Shi, Gunma, Japan) were immunized with a mixture of 33 μg of recombinant human ErbB-2/HER2 protein (ACROBiosystems, Newark, DE, USA) and recombinant human ErbB-3/HER3 protein (ACROBiosystems) in combination with Freund's Complete Adjuvant. Seven days after the first immunization, rabbits were re-immunized with the same antigen mixture with Freund's Incomplete Adjuvant, and this process was repeated eight times every two weeks. Seven days after the final immunization, rabbits were euthanized to isolate spleen and lymph node cells. The serum titers of each antigen were checked by Enzyme-Linked ImmunoSorbent Assay (ELISA) using HRP conjugated goat anti-Rabbit antibody (Immuno-Biological Laboratories Co., Ltd., IBL, Fujioka-Shi, Gunma, Japan) at seven days after the fourth, sixth and last immunization, respectively. All experiments with animals were approved by the Institutional Animal Care and Use Committee of Daiichi Sankyo and carried out in strict accordance with the IBL guidelines for animal experiments, which complies with the laws concerning animal protection and management.
Preparation of rVH-displaying phage library. From a total 1.9 × 10 8 spleen and lymph node cells of immunized rabbits, mRNAs were extracted using Dynabeads mRNA DIRECT Kit (Life Technologies Corporation, Grand Island, NY, USA) and reverse transcribed to cDNA with Transcriptor High Fidelity cDNA Synthesis Kit (Roche, Basel, Switzerland). Four 5′-sense and two 3′-antisense primers were designed to cover all rVH germ-line sequences (Supplementary Table S1) and used for polymerase chain reaction (PCR) to amplify rVH genes from the synthesized cDNA library. Amplified rVH genes were inserted into the Sfi I/Not I site of phagemid vector pCANTAB5E (Amersham plc, Buckinghamshire, UK) and fused following the 5′ end of the E-tag (GAPVPYPDPLEPR) and the gIII coat protein of M13 bacteriophage (gIIIp) coding sequence. E. coli TG-1 strain (Agilent Technologies, La Jolla, CA, USA) was transformed with these phagemid vectors. Obtained transformants were pooled and infected with enough amounts of helper phage VCSM13 (multiplicity of infection >100). Subsequently rVH displaying phages were produced by cultivation of these transformants at 16, 20, 22 or 25 °C overnight with 2× YT medium supplemented with 0.25 mM IPTG, 100 μg/mL ampicillin, and 50 μg/mL kanamycin. Produced phage was then precipitated from overnight cultured medium using polyethylene glycol 6,000 and dissolved with phosphate buffered saline (PBS). Comparison of rVH display levels on the phage, 1.0 × 10 10 virions of phages, produced at each temperature were subjected to western blotting (WB) using anti-E-tag antibody (Bethyl Laboratories, Montgomery, TX, USA), and appropriate secondary antibodies were used for detections. Virion numbers of purified phages were quantified using spectrophotometry with the following formula 37 . Here, A 269 and A 320 indicate UV absorption of 269 and 320 nm, respectively, and the number of nucleotide bases per virion was set to be 5000.
Panning against antigens. The rVH displaying phage library was first subjected to negative selection using Dynabeads M-280 Streptavidin (Life Technologies Corporation) without antigen for eliminating non-specific binders. All beads were previously blocked with BSA (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA) in all of the experiments. rVH-displaying phages unbound to the beads were then exposed to 50 pmol of biotinylated HER2 or HER3 at 4 °C for overnight (first round) or at room temperature for one hour (second and third rounds). Biotinylated antigens were prepared using ChromaLink ™ Biotin Antibody Labeling Kit (Solulink, Inc., San Diego, CA, USA) according to the manufacturer's instructions. Subsequently, new beads were added to recover biotinylated antigen binders. Library treated beads were washed by three different conditions as follows: PBS containing 3% (w/v) BSA and 0.05% Tween-20, PBS with 0.05% Tween-20, and PBS, respectively. Specific binders were eluted with 0.1 M Glycine-HCl (pH 2.2) and immediately neutralized with 1 M Tris-HCl (pH 8.0). After that, E. coli TG-1 strain was infected with eluted phages and cultivated on an LB agar plate supplemented with 100 μg/mL of ampicillin. Appearing colonies were used for the next round of phage production or ELISA screening. Phage recovery rates of each round of panning were calculated as a ratio of input to output titer (colony forming unit for TG-1). To confirm the concentration of antigen specific binders, panning without antigen was also conducted as a negative control at the third round.
ELISA screening of antigen binding rVHs. A randomly selected 317 output colonies from the third panning were inoculated to 2× YT medium supplemented with 100 μg/mL of ampicillin and 0.1% (w/v) glucose and cultivated at 37 °C overnight. After final concentration of 0.5 mM, IPTG was added to induce rVH expression, and cultivation started again at 16 °C. Lysozymes were then added to overnight culture and the mixture was transferred to a Nunc MaxiSorp flat-bottom 96-well plate (Thermo Fisher Scientific, Inc., Waltham, MA, USA) precoated with antigen (signal) or bovine serum albumin (noise). Bound rVHs were detected with HRP-conjugated anti-E-tag antibody (Bethyl laboratories) and >2.0 as a signal-to-noise ratio was set as the criterion for positive. Positive clones with repeatability were regarded as a hit. The osmotic shocked supernatant was mixed with the cultured medium, and rVH was affinity purified from this mixture by using Ni Sepharose excel (GE Healthcare UK Ltd., Little Chalfont, Buckinghamshire, England). Concentration dependent ELISA were conducted using affinity purified rVHs, and as for VHs for evaluations of binding affinity and thermal stability, affinity purified rVHs were further purified with gel filtration using a Superdex 75 10/300 GL with AKTA system (GE Healthcare UK Ltd). rVH of the clone, H2-1-1, was prepared using an Expi293F mammalian cell expression system (Life Technologies Corporation) according to the manufacturer's instructions. Genes encoding H2-1-1 or its C54-C78 mutant with FLAG and His tags were sub-cloned into pcDNA3.1 vector for mammalian expression and expressed H2-1-1 rVH was purified similarly to rVHs using an E. coli expression system. The purities of finally purified rVH samples were confirmed by SDS-PAGE analysis and the protein concentrations were determined from the absorbance of 280 nm with the extinction coefficients which were calculated from amino acid sequences in Fig. 2a  Regarding thermodynamic analysis, DSC was conducted using a MicroCal VP-Capillary DSC (Malvern Instruments Ltd, Worcestershire, UK.) at a heating rate of 60 °C/h. To evaluate thermal stability, the T peak value (temperature where heat capacity takes the maximal value) was determined with rVH samples at a concentration of 0.2 mg/mL in PBS using the software MicroCal Origin 7 (Malvern). To obtain detailed thermodynamic parameters, DSC analysis was conducted at 1.0 mg/mL. The ΔH values were estimated by the integration of endothermic heat accompanied by the unfolding. The T m values were determined as the temperature at which the integration of endothermic heat is equal to half the area of ΔH. The enthalpy change (ΔH(T)), entropy change (ΔS(T)) and free energy change (ΔG(T)) are indicated as Equations 7-9.

Preparation of rVHs.
From Equations 7-9, ΔS can be calculated as ΔH/T m because ΔG becomes zero at T m . In this analysis, we assumed that ΔC P was constant and determined as the difference between the baseline of folded and unfolded states in the DSC curve.
Scientific RepoRts | 7: 5794 | DOI:10.1038/s41598-017-06277-x Model structure construction and structural analysis. Model structures for obtained rVHs were constructed with the antibody structure prediction function of BioLuminate (Schrodinger, New York, NY). The PDB data, which were employed to construct the model structure, are listed in Supplementary Table S3. To argue the possibility of additional disulfide bond introduction to rVHs, model structures of rVHs with additional disulfide bonds were created using the function of Discovery Studio Ver. 4.0 (Accelrys, San Diego, CA, USA). Structural analysis of VHH hCG was conducted using available PDB data (PDB ID: 1HCV). The ASA of residues for Cys mutation and estimated RMSD of Cα of residues in the framework were calculated using Discovery Studio Ver. 4.0. To estimate hydrophobicity of VL-interacting surfaces in obtained rVHs, the ASA pol and ASA non-pol of residues located at the VL-interacting surface were calculated using Discovery Studio Ver. 4.0. The residues of rVH in the VL interacting surface were defined as the common residues, which are less than 5 Å from its VL, in the rVH of available rabbit Fv structures (PDB ID: 4HBC, 4HT1, 4JO1, 4JO4, 4O4Y). The ASAs of these rVHs, which were originally obtained as Fv [32][33][34][35] , were also calculated with the three dimensional coordinates of the VH without the VL portion, which had been derived from the crystal structures of the Fv.
Introduction of additional disulfide bonds to rVHs. For additional disulfide bond introduction, Gly or Ala and Ile at positions 54 and 78 of rVHs were both mutated to Cys by site-directed mutagenesis. Cys introduced mutants were prepared by the same methods as wild type rVHs, and disulfide formations were confirmed for H2-1-1, H3-9, and H3-15 as follows. rVHs and their mutants were treated with sequencing grade chymotrypsin (Roche, Basel, Switzerland) at 37 °C for 4 hours followed by addition of 10% formic acid, and then subjected to LC-MS analysis using a Waters Synapt G2S (Waters Corporation, Milford, MA, USA). Peptide identification was conducted with MassLynx Mass Spectrometry Software ver. 4.1 and BiopharmaLynx Software ver. 1.3 (Waters Corporation).