Mirror-image Streptavidin with Speci � c Binding to “ Non-natural ” L-biotin


 The streptavidin-biotin system is known to have a very high affinity and specificity and is widely used in biochemical immunoassays and diagnostics. However, this method is affected by endogenous D-biotin in serum sample measurements (biotin interference). While several efforts using alternative high-affinity binding systems (e.g., genetically modified streptavidin and biotin derivatives) have been attempted, these efforts have all led to reduction in affinity. To solve this interference issue, the enantiomer of streptavidin was synthesized, which enabled specific binding to L-biotin. We successfully obtained a functional streptavidin molecule by peptide synthesis using D-amino acids and an in vitro refolding technique. Several characterizations, including size exclusion chromatography (SEC), circular dichroism spectra (CD), and heat denaturation experiments collectively confirmed the higher-order enantiomer of natural streptavidin had been formed with comparable stability to the natural protein. Non-natural L-biotin specific binding of this novel molecule enabled us to avoid biotin interference in affinity measurements using the Biacore system and enzyme-linked immunosorbent assay (ELISA). We propose the enantiomer of streptavidin as a potential candidate to replace the natural streptavidin-biotin system, even for in vivo use.


Introduction
Avidin and streptavidin are known to bind to natural biotin with high a nity and speci city 1-2 . Core streptavidin [3][4] , which consists of the core region of natural streptavidin, has particularly high stability against thermal and chemical denaturation, as well as protease degradation, and has therefore been used for various purposes in the life sciences eld 5 . Its most widely used application is for highly sensitive quanti cation, exploiting the molecule's high a nity for D-biotin for biological analysis. Numerous immunoassay systems using the streptavidin-biotin reaction for signal ampli cation or bound/free (BF) separation have been proposed [6][7] . However, the current limitation to the use of the streptavidin-biotin reaction is that endogenous biotin from human samples binds to streptavidin used in these biological assays, phenomena referred to as "biotin interference", which may lead to false negative or false positive results depending on the assay. The unusually high concentrations of endogenous biotin in human samples are results of excess intake of foods or supplements containing D-biotin. Regulatory authorities, including the US Food and Drug Administration (FDA), required reliable action to be taken by diagnostic companies to solve this misinterpretation of results caused by biotin interference in clinical assays 8 . Streptavidin variants with low a nity for D-biotin and high a nity for biotin derivatives have been proposed [9][10][11] . However, the a nity for these biotin derivatives have so far proved to be inferior to the a nity of natural streptavidin for D-biotin, showing insu cient applicability of these newly proposed biotin derivatives for avoiding biotin interference.
In this study, we chose another approach to solve the biotin interference issue. Chirality works as an important factor in determining biomolecular interactions in broad range of scales, from small molecules to nanoparticles 12 . The key to our approach is to create a protein composed of D-amino acids, instead of Lamino acids, creating a high-order enantiomer of the natural protein. This approach would provide us the same interaction of a set of natural partner molecules because no structural modi cations were made except for the optical isomer relationship. We expect that this approach would provide high selectivity to L-biotin, instead of D-biotin, thereby avoiding biotin interference in the biological samples. This assumption is based on the strict molecular recognition formerly con rmed by the X-ray crystallography of natural streptavidin and the D-biotin complex.
Very few studies have reported the application of D-amino acids for enzymes. Such studies involved HIV protease and DNA ligase, composed of D-amino acids, which were prepared and characterized for these functions [13][14][15] and the effects of protein chirality on the interface between larger biomolecules was evaluated. Molecular chaperones can fold L-and D-substrate proteins, indicative of the promiscuous chiral recognition formed on protein interfaces 14 . In contrast, candidates for use as therapeutic peptides were found to show the potential applicability of reciprocal chiral speci city; L-peptides acted as antagonists against the D-target protein, whilst the D-peptide strictly antagonized the natural L-target form of the protein 16 . These studies addressed the stereoisomeric relationship at the interfaces of protein/peptide complexes; however, such chiral recognition tendencies observed on large interaction surfaces should be discussed on a case-by-case basis. Promiscuous molecular recognition is common in nature, for example enzyme/substrate, antigen/antibody, and ligand/receptor systems 17 . Therefore, we should carefully evaluate the rigidity of chiral speci city in each protein and peptide system.
In our study, we expected strict chiral-speci c recognition in the streptavidin-biotin system based on the accurate molecular packing and ultra-high a nity observed for natural streptavidin and D-biotin. We synthesized core streptavidin peptide consisting of D-amino acids, in place of L-amino acids, using peptide synthesis, which gave rise to a functional form (D-core streptavidin) which displayed a native high-order structure. The high-order structure and thermal stability of the obtained D-core streptavidin were compared with those of the natural core streptavidin. Furthermore, the functions of D-core streptavidin, in particular the L-biotin binding a nity and suppression of D-biotin interference in the immunoassay for biological samples, were evaluated. Finally, the applicability of this novel interaction using the D-core streptavidin and L-biotin are discussed.

Results And Discussion
Synthesis. The core streptavidin sequence is known to be 127 amino acid residues, as shown in Table 1. The core streptavidin peptide composed of D-amino acids was synthesized using GlyTech Inc. peptide synthesis service. The entire core streptavidin sequence was divided into ve segments and each segment was prepared by Fmoc solid-phase synthesis with D-amino acids and glycine. As for immunoassay evaluation, the core streptavidin sequence was divided into 2 segments to accelerate its synthetic process. Full-length D-amino acid peptides were obtained by linking the segments using the general ligation method. The sequence corresponded to residues 37-163 in the amino acid sequence of natural streptavidin. The Electrospray Ionization Mass Spectroscopy (ESI-MS) spectrum was shown in Figure S1. Table 1 Refolding recovery of each core-streptavidin AEAGITGTWYNQLGSTFIVTAGADGALTGTYESAVGNAESRYVLTGRYDSAPATDGSGTALGWTVAWKNNYRNAHSATTWSGQYVGGAEARINTQWLLTSGTTEANAWKS We also prepared L-biotin and its derivatives. L-biotin was obtained by performing optical resolution from racemic biotin synthesized according to a previous report 18 . Furthermore, 2,5-dioxopyrrolidin-1-yl 6-(5-((3aR,4R,6aS)-2-oxohexahydro-1H-thieno [3,4-d]imidazol-4-yl)pentanamido)hexanoate (L-biotin-AC5-OSu) was synthesized from L-biotin as a material for L-biotin modi cation to the ε-amino group of proteins. A synthetic route of L-biotin-AC5-OSu is shown in Figure 1. The 1 H NMR spectrum of L-biotin-AC5-OSu was sown in Figure S2.
Refolding of D-core streptavidin and natural core streptavidin. Tetrameric high-order structures of D-core streptavidin and the natural core streptavidin were obtained by well-known dilution refolding 19 with some modi cations (Table 2). Representatively, D-core streptavidin peptide (28.8 mg; synthesized) and natural core streptavidin powder (10.1 mg; Roche diagnostics) were solubilized and denatured in 6 M guanidine hydrochloride at a protein concentration of approximately 20 mg/ml for 45 min at 85°C, and then diluted 6-fold with 80 mM Tris-HCl buffer to 1 M guanidine hydrochloride and kept for 30 minutes at room temperature. Small amounts of insoluble aggregates were observed during this step. Diluted solutions were then concentrated 6-fold and applied to the second 100-fold dilution with 80 mM Tris-HCl buffer, pH 8.0. Diluted solutions were then concentrated 200-fold by centrifugation at 4°C. Obtained supernatants were kept at 4°C for 12 hours to remove non-native by-products as insoluble aggregates. Finally, protein mixtures were centrifuged at 4°C to yield natively refolded core streptavidin proteins with no aggregates. Approximately 28.8 mg and 10.1 mg of refolded D-core streptavidin and natural core streptavidin were obtained with recovery yields of 25% and 50%, respectively. The suppression of the recovery yield for D-core streptavidin (25%) may be due to less pure refolding materials compared to the purchased recombinant protein, the natural core streptavidin (Roche Diagnostics). Each streptavidin sample was subjected to buffer exchange to 20 mM phosphate buffer (pH 8.0) and then used in the following experiments. The representative protein recoveries during denaturation and dilution refolding are summarized in Table 2. Table 2 Core streptavidin amino acid sequence Step Dilution fold (from Step 1) Buffer condition Temperature Monitoring of tetrameric structure formation. The tertiary structures of refolded D-core streptavidin and the refolded natural core streptavidin were con rmed by analytical SEC (Figure 2a). Both refolded streptavidins in 20 mM phosphate buffer showed symmetrical peak shapes with retention times comparable to that of the active streptavidin reagent ( Figure S3). No other peaks with longer retention times were observed for either streptavidin, therefore only the tetramer was present. Peak areas of both streptavidins accounted for protein concentrations calculated by UV absorbance, meaning there were no aggregates or unfolded forms that could not be eluted from the SEC column. After storage under refrigerated temperature or repetitive freeze-thaw, both streptavidins showed unchanged SEC peak shapes (data not shown). These observations collectively suggest that our two-step refolding was successful in the formation of a natively folded tetramer structure.
Tertiary structures of refolded core streptavidin proteins. The higher-order structures of refolded D-core streptavidin and its natural enantiomer, L-core streptavidin, in phosphate buffer were compared using circular dichroism spectra (CD spectra). CD spectra of natural core streptavidin have been reported 20 with a negative ellipticity peak at 216 nm (derived from the β-sheet structures) and a positive ellipticity peak at 232 nm (presumably derived from the β-turn structures). In our studies, CD spectra (wavelength, 190 nm-260 nm) of refolded D-core streptavidin and the natural, L-core streptavidin are shown in Figure 2b.
These spectra showed characteristic ellipticities at 216 and 232 nm as seen in literature 20 ), but with axial symmetry. It is well known that the CD spectra of compounds with optical isomerism show axial symmetry, especially in the case of low molecular weight chemicals with simple structures. However, it is not trivial whether a mirror image of the CD spectra is always the case with protein pairs composed of D-and L-amino acids, respectively. In proteins and peptides, for example, additional CD spectra at around 230 to 250 nm, which are derived from the side chains of aromatic amino acids, might overlap with the CD spectra of the main chains 21 . These situations might break the CD axial symmetries of L-and D-amino acid protein pairs. For example, DNA ligases composed of L-and D-amino acids, could not show accurate mirror images of CD spectra (axial symmetries) at around 230 to 250 nm 15 , where the in uence of the aromatic amino acid side chains might occur. In contrast, in our study, obvious enantiomeric characteristics were observed between the CD spectra of the Dcore streptavidin and the natural core streptavidin. The positive and negative ellipticities were clearly reversed at 216 nm and 232 nm, suggesting that numerous β-turn structures in D-core streptavidin have intense negative ellipticities at approximately 232 nm. Even if ellipticities derived from the aromatic side chains of D-core streptavidin overlap with ellipticities from its main chains, these effects would be minor. As for natural core streptavidin, CD spectra would be reversed compared to D-core streptavidin. As a result, we considered that mirror-image, higher-order structures of these two core streptavidin molecules showed substantial mirror image CD spectra. In addition, axial symmetries were observed even in the near-ultraviolet region of 250 nm-300 nm ( Figure S4). These results strongly suggest a mirror image of the secondary and tertiary structures of the D-and natural core streptavidin molecules. Thermal stabilities of refolded streptavidin proteins.
Thermal stability studies were performed using differential scanning calorimetry (DSC). Heat denaturation curves and heat denaturation midpoint temperature (T m ) of D-core streptavidin and natural core-streptavidin are shown in Figure 2c. These two heat denaturation curves seemed to be quite similar and were estimated to have substantially comparable Tm values. The similar thermal stabilities observed here support the mirror image structures of the D-core and the natural core streptavidin molecules shown by CD spectra.
A nity and speci city of refolded streptavidin proteins to biotin by kinetic analysis. A nity analyses by Biacore were performed for refolded core streptavidin molecules to determine their a nities with L-and D-biotin. We used D-and L-biotin-labeled bovine serum albumin (BSA) as ligands and D-core and natural core streptavidin as analytes. BSA was labelled by a commercially available D-biotin-AC5-OSu and above L-biotin-AC5-OSu. We rst evaluated the appropriate BSA labelling conditions with biotin to obtain almost 1:1 binding between one streptavidin tetramer and one biotin on the surface of BSA. It was nally found that about 1.7 biotin molecules should react with one BSA molecule to avoid the avidity effect on binding. Sensorgrams of D-core streptavidin with L-/D-biotin are shown in gure 2d. While D-core streptavidin expectedly bound to L-biotin with su ciently high a nity, no D-biotin binding was seen on the sensorgram.
The equilibrium dissociation constant, KD, of core streptavidin to the ligand, L-biotin, determined by tting curve analysis was 6.29 × 10 −14 M. On the other hand, the KD of the natural core streptavidin reagent (11520679, Roche Diagnostics) for D-biotin was 6.85 × 10 −14 M, measured under the same conditions. In this study, the binding a nity observed between D-core streptavidin and L-biotin was comparable to natural core streptavidin and D-biotin. Importantly, the fact that D-core streptavidin did not show any binding a nity for D-biotin supported our hypothesis that stringent molecular recognition between streptavidin and biotin can exploit chiral speci city. Namely, D-core streptavidin proved to have a mirror image property in biotin binding. Evaluation of biotin interference on D-core streptavidin in immunoassay.
The streptavidin-biotin reaction is a commonly used system in immunoassays [6][7] . We set up a streptavidin-biotin sandwich chemical luminescent enzyme immunoassay (CLEIA) system for thyroid stimulating hormone (TSH) to evaluate our D-core streptavidin capabilities. L-biotin labeled anti-TSH antibodies were prepared to capture TSH as the target antigen. D-core streptavidin reagents, as binding partners to biotin labeled anti-TSH antibodies, were immobilized in a 96-well plate for the immunosorbent assay. ALP-labeled anti-TSH antibodies were used for the sandwich detection. Various concentrations of D-biotin were added to TSH antigen specimens at a nal concentration of 1 nM to 100 nM to observe the interference on its valid chemical luminescent signal. The architecture of the assay system is shown in Figure 3. The results of signal interference are shown in Figure 4. In the natural core streptavidin (reagent, 11520679, Roche Diagnostics)/D-biotin system, the signal decreased depending on the amount of D-biotin added, showing similar signal interferences seen in clinical diagnostics. However, no signi cant changes (signal decreases) in the D-core streptavidin/L-biotin system were observed, regardless of the amount of D-biotin added (± <5%). We found that a strict molecular recognition system in streptavidin and biotin enabled us to avoid biotin interference through chiral speci city.

Conclusion
The physicochemical properties shown by the SEC and CD spectra suggested that D-core streptavidin had a tetrameric and natively folded structure and was the enantiomer of a natural core streptavidin. In addition, the thermal stability shown by the DSF analysis supported this conclusion. We observed a very high binding a nity and speci city to L-biotin for D-core streptavidin. We also showed our D-core streptavidin together with L-biotin would be a solution to the issue of biotin interference, awaited in clinical diagnostics to be solved. We expect that our D-core streptavidin would have broader applicability, ranging from in vitro biochemical as-says to in vivo diagnostics, without endogenous biotin interference.

Synthesis of L-biotin derivatives
General. Unless otherwise noted, all reactions were performed in heat gun dried glassware, sealed with a rubber septum under an atmosphere of nitrogen or argon.
Materials were obtained from suppliers and used without further puri cation. Flash column chromatography was carried out using Cica Slica Gel 60 N (spherical/40-50 µm).
Reactions and chromatography fractions were analyzed employing precoated silica gel 60 F 254 plates (Merck). Compounds were visualized using an ultraviolet lamp (254 nm) and/or by staining with p-anisaldehyde (in EtOH), potassium permanganate (in water containing NaOH and K 2 CO 3 ) or ammonium molybdate (in 10% H 2 SO 4 ).

Far ultra-violet CD spectrum analysis
The measurement samples were prepared by dilution to 300 µg/mL in 20 mM phosphate buffer. CD spectra (Far ultra-violet region: 190-260 nm) were acquired at 25°C using a CD spectrometer (JASCO, J-1500). The path length was set to 1 mm. The data pitch was set to 0.1 mm. The scanning speed was set to 100 nm/min, and the spectra were averaged from three scans. The spectral baseline was recorded using a 20 mM phosphate buffer. Each data point was baseline-subtracted.

SPR analysis
Preparation of biotin labeled albumin. 50 mg/ml of bovine serum albumin (2.8 ml; Proliant Biologicals, 68700) was added to 10 mg/mL of L-biotin-AC5-OSu in N, N-dimethylformamide solution (163 µL) and the mixture was incubated at 35°C for 1 h. The mixture was buffer-exchanged to 0.1 M phosphate buffer (pH 7.5) with PD-10 column (Cytiva). D-biotin labeled albumin was produced under the same conditions using D-biotin-AC5-OSu (Dojin Chemical) Measurement. The Surface plasmon resonance was measured using a Biacore T200 instrument (Cytiva). Each prepared biotin labeled bovine serum albumin was immobilized on each ow cell of a CM5 sensor chip by the amine coupling method to a nal targeted resonance unit of 400. Non-biotin labeled bovine serum albumin (Proliant Biologicals, 68700) was immobilized on a reference ow cell under the same condition. Next, ve concentrations (0.05, 0.15, 0.45, 1.35, and 4.05 nM) of D-core streptavidin in HBS-EP+ (Cytiva) buffer were passed over the chip using a single cycle method. The sensorgram data were analyzed using Biacore T200 evaluation software version 3.  Evaluation of the effect on the measurement system in the presence of biotin.(1) In a 1.5 mL tube, as a sample, HISCL TSH calibrator C3 (Sysmex) 400 µL and HISCL TSH R1 reagent (Sysmex) was added to 400 µL, and the reaction was carried out at 37°C The reaction was carried out under the condition of 37°C and 600 rpm for 30 minutes. To evaluate the in uence of D-biotin, 0 ng/mL, 1 ng/mL, 10 ng/mL, 1 ng/mL, 10 ng/mL, 100 ng/mL, or 100 ng/mL of D-biotin was added to the samples and reacted.
(2) The streptavidin-immobilized plate was washed three times with HISCL wash solution, and 60 µL of the reacted solution was added to each well (1.5-mL. The plates were washed three times with HISCL wash solution (Sysmex), and 60 µL of the reacted solution was added to each well. The reaction was carried out at 37°C and 600 rpm for 20 min. Biotin-labeled TSH antibody prepared above was added, and the reaction was carried out at 37°C and 600 rpm for 30 min.
(3) The plate was washed ve times with HISCL wash solution, and 50 µL of HISCL R4 Reagent (Sysmex) and 100 µL HISCL R5 Reagent (Sysmex) were added, and the reaction was carried out at 37°C and 600 rpm for 3 min. After the reaction, luminescence was detected using a microplate reader (In nite 2000 PRO, TECAN). The streptavidin-conjugated plate and biotin-labeled TSH antibody after blocking were also measured in the same way.

Statistics
All values in Figure 3 are expressed as mean from duplicated analysis.

Declarations
Data availability The data that support this study are available from the corresponding author upon reasonable request.

Competing interests
The authors declare no competing interests. Figure 1 Synthesis of L-biotin-AC5-OSu