On the basis of alignment results, the VL framework structure of scFv (H4) was constructed with the consensus of an anti-human vascular endothelial growth factor Fab (PBD code: 1za3) with a resolution of 3.35 Å [6]. The VH (H4) domain was constructed by referring to the consensus structure of the anti-sodium citrate symporter CitS Fab (PBD code: 2v7n) with a resolution of 1.92 Å [7]. Both the H4 domains share more than 90% sequence identity with the corresponding domains of the antibodies. We took the previously resolved structures into account [7–10] and predicted that scFv (H4) has a tubiform structure. The complementary-determining regions (CDRs) formed loops at the tip of each domain and were supported by the structurally conserved framework region [11,12]. Two potential locations for the interdomain bond were identified on the basis of domains proximity, evidenced in the vertical view of the model (indicated by the dashed box in Figure 1a). The shortest distance was measured between framework H2 and framework L4, but also framework L2 and framework H4 were sufficiently close for a putative disulfide bridge. Some amino acid residue pairs were designed to introduce the interdomain disulfide bond, and the distance between the Cα atoms of these pairs was evaluated (Table 1). Previous research showed that the Cα-Cα distance of the disulfide bond in cysteine residues in known proteins ranges from 4.2 Å to 6.6 Å [8,13]. H44-L100, H46-L98, and H103-L43 were the possible sites with an optimal Cα-Cα distance [14]. However, disulfide bonds of H46-L98 and H103-L43 were very close to the CDR L3 (from L89 to L97) and CDR H3 (from H95 to H102) [15], respectively. H44-L100 was selected since it possessed the shortest distance between the two sulfide atoms and was sufficiently distant from the CDRs to prevent possible interference with the antibody binding capacity. In scFv (H4), the substitution of Gly with Cys at the site of H44 introduced a slight structure stress, neutralized by the Gly at H42. Similarly, the substitution of Gln with Cys at L100 would not significantly affect the domain folding due to the neighboring Gly residues at L99 and L101. Hence the introduction of the two cysteines produced a scFv (H44-L100) mutant with a general conformation similar to the cognate scFv (H4).

According to molecular docking, the interdomain disulfide bond of H44-L100 is between structurally conserved framework positions, which are distant from the antigen binding sites (Figure 1b). According to the software, the affinity of scFv (H4) and the scFv (H44-L100) mutant to AFB₁ was 1.6 μM and 1.4 μM, respectively.

The expression of scFv in bacteria often encounters the formation of inclusion bodies, which can occur due to the reducing environment in the cytoplasm, which does not allow protein folding stabilization through disulfide bonds [16]. In our experiment, scFv (H4) and its derivates were secreted into the periplasm in a soluble form. Functional H4 was expressed at a concentration of 36 mg/L of the culture medium, while the amount of functional scFv (H44-L100) mutant was 8.85 mg/L of the culture medium. No free sulfhydryl group was found, indicating that the all the six sulfhydryl groups existed in disulfide bond form. The functionality of scFv (H44-L00) suggested correct formation of the native intradomain disulfide bonds in VH and VL, which was necessary for the functional conformation of scFv. Hence, we deduced that the sulfhydryl groups of the mutation cys were involved the interdomain disulfide bond.

Aggregation of scFv might result in the loss of functionality and interference in the study of the corresponding monomer [4]. Aggregation was concentration-dependent, and oligomerization of scFv (H4) and the scFv (H44-L100) mutant at various concentrations was assessed by size exclusion chromatography. For scFv (H4), the monomer fraction accounted for approximately 30% of the total protein at 10 μM (Figure 2a). H44-L100 existed primarily as monomeric proteins at 10 μM and aggregated forms were not detectable (Figure 2b). As shown by the results of the unfolding study (Figure 2), the aggregation of antibody fragments was negligible when the concentration of scFv was less than 1 mM.

The wavelength at the maximum emission of the fluorescence spectrum (λ_max) was used in the unfolding study, because the emission maximum depends solely on the environment of the tryptophan residues detected and is not affected by the protein concentration during unfolding [3,17,18]. The half-denaturation concentration of GdnHCl (GdnHCl₅₀) was used to describe the stability of the protein in the unfolding experiment. A comparison of the denaturizing profiles of isolated domains indicated that VH (H4) with a GdnHCl₅₀ of 1.9 M was more stable than VL (H4) with a GdnHCl₅₀ of 1.5 M; their overlapped unfolding transition resulted in a broader unfolding transition of scFv (H4). A GdnHCl₅₀ of 2.4 M for scFv (H4) suggested co-operative unfolding of two domains in scFv (Figure 3), which was higher than those of the isolated domains. The stability of scFv (H4) was increased by the introduction of an interdomain disulfide bond, and the scFv (H44-L100) mutant had a GdnHCl₅₀ of 4.2 M.

Generally, previous studies on the effect of interdomain disulfide bonds focused on stability, while their effect on antibody affinity was not well documented. The reactivity of scFv (H4) and the scFv (H44-L100) mutant was determined by direct and competition ELISA (Figure 4). In non-competitive ELISA studies, scFv (H4) and scFv (H44-L100) were able to bind to AFB₁-BSA (50 μM) coated on microtiter plates in a concentration-dependant manner (Figure 4a) and they showed similar binding behaviors.

Comparable ELISA results were obtained with both scFv (H4) and the scFv (H44-L100) mutant. Pre-incubation of scFvs (10 μM) with free AFB₁ resulted in the inhibition of scFvs binding to the microtiter plates coated with AFB₁-BSA, and the binding profiles of scFv (H4) and scFv (H44-L100) mutant to free AFB₁ were statistically identical (Figure 4b). The IC₅₀ values of both scFv were approximately 0.4 ng/mL.

The recombinant antibody clone scFv (H4) consisting of the anti-AFB₁ scFv was selected from the Tomlinson J Library (Geneservice, Cambridge, U.K.) [6]. E. coli BL21 (DE3) and pET22b vector were from Novagen (Beijing, China). All the enzymes used in this study were obtained from New England Biolabs (Beijing, China). AFB₁ and aflatoxin B₁-bovine serum albumin (AFB₁-BSA) conjugates were purchased from Sigma (Shanghai, China). Other chemicals were supplied by Sangon (Shanghai, China).

The model of H4 was performed using WAM^® algorithm [19,20]. The dead-end eliminations method was used for side-chain building, and root mean square deviation was used to screen the final model. Molecular docking was processed with a PatchDock^® server and refined with a FireDock sever. The docking solution was evaluated using the Logplot software [21]. The docking solution was visualized using the program 3DMol^® viewer (Informax Inc.); distance measurements were carried out with the same software package.

The H4 gene was amplified by polymerase chain reaction (PCR) with the primers LMB and pHEN (Table 2), and the PCR product was double digested with the restriction enzymes NcoI and NotI. The enzymatically digested scFv gene fragment was linked with pET22b (+) and then transformed into E. coli BL21 (DE3). Restriction enzyme digestion and DNA sequencing (Sangon, Shanghai) confirmed colonies bearing the pET22b/H4 construct.

The scFv (H44-L100) mutant was constructed with the introduction of cysteine residues to H44 and L100 by PCR with clone H4 as the template. PCR A was run with primers H44For and L100Back, while PCR B was run with LMB and H44Back. The 2 target segments were linked by PCR without primers for 5 cycles, and then amplified in another 35 cycles after addition of the primers of LMB and L100Back. The resulting segment was digested with NcoI and NotI and cloned to pET22b (+), as described above.

The expression and purification of scFv in E. coli BL21 (DE3) was performed according to the methods described in the pET^® system manual. Briefly, a single clone of BL21 (DE3)/pET22b/H4 was selected and grown to mid-logarithmic phase (OD600 = 0.8), after which IPTG was added to a final concentration of 1 mM. The cells were allowed to grow at 20 °C for another 20 h. The cells were harvested by centrifuging at 3300 × g for 15 min. The resulting pellet was resuspended in 10 mL of phosphate-buffered saline (PBS) and ultrasonicated in an ice bath. The soluble scFv was purified using Ni-chelating affinity chromatography (GE Healthcare, Shanghai).

The sulfhydryl content in scFv (H4) and the scFv (H44-L100) mutant was determined by mixing 70 μL of scFv (10 mg/mL) with 1 mL of freshly prepared 2-nitro-5-thiosulfobenzoate (NSTB) test solution. The absorbance at 412 nm was determined using the NSTB solution as a reference [22].

Gel filtration of scFv (H4) and the scFv (H44-L100) mutant were performed with the AKTA^® system using a column of Superdex^® 75 (Tricorn, Amersham Bioscience). All measurements were carried out in PBS containing 0.005% Tween 20. One milliliter of scFv fragments (10 μmol/L) was injected in the column. All the purified scFv fragments were analyzed after incubation at 37 °C for 20 h. The incubated samples were centrifuged at 14,000 × g for 5 min before loading to remove the insoluble aggregates.

All fluorescence measurements were performed with a fluorescence spectrophotometer (Hitachi 650-60) at 25 °C. The protein was excited at 280 nm and the emission spectrum was recorded from 300 nm to 380 nm. For the equilibrium denaturation measurements, the protein sample (0.5 μM) was incubated at 16 °C for 20 h in PBS containing different amounts of GdnHCl.

Maxisorp^® plates were coated with 25 μL of AFB₁-BSA (5 μg/mL) at 37 °C for 2 h or with 25 μL of BSA (5 μg/mL) as negative control [18]. The coated wells were washed with PBS twice, blocked with 2% BSA at 37 °C for 2 h, and washed again with PBS. Fifty milliliters of the antibodies amplified from each round of selection were mixed with 100 μL of 2% BSA and incubated at 37 °C for 1 h, after 4 washes with PBS containing 0.1% Tween 20 (PBST). Protein A/HRP antibody in 2% BSA (1:5000 dilution) was added and incubated at 37 °C for 1 h. The plate was washed again with PBST 4 times. The signal was visualized with 100 μL of 3,3',5,5'-tetramethylbenzidine (TMB, 100 μg/mL) in acetate buffer (pH 5.5). The reaction was stopped by adding 50 μL of 1 M H₂SO₄, and the optical densities at 450 nm (OD450) and 650 nm (OD650) were recorded with a MK3^® microplate reader (Thermo, Shanghai, China). Individual clones from the titer plates of the third selection round were randomly chosen, amplified in phage form, and used in monoclonal phage enzyme-linked immunosorbent assay (ELISA).

In the competitive ELISA, Maxisorp^® plates were coated with AFB₁-BSA and blocked as described above. AFB₁ standard solutions with a range of concentrations of 0.05–50 ng/mL were added to each well. Then, 50 μL of soluble scFv was added to each well. The plates were incubated at 37 °C for 1 h and washed with PBST. Protein A/HRP was incubated at 37 °C for 1 h, aspirated, and then 100 μg/mL TMB was added, and the absorbance was recorded at 450 and 650 nm. The half-maximal inhibitory concentration (IC₅₀) of free AFB₁ was determined.