Multimeric fusion single‐chain variable fragments as potential novel high‐capacity ligands

In basic and applied biotechnology, design of affinity ligands has become essential for high‐capacity applications such as affinity‐based downstream processes for therapeutic molecules. Here, we established a proof‐of‐concept for the use of multimeric fusion single‐chain variable fragment (scFvs) as high‐capacity ligands in affinity adsorbents. Mono‐ and di/tri‐scFvs separated by Pro‐rich negatively charged linkers were designed, produced, and immobilized to 6% cross‐linked agarose beads. Frontal binding experiments with a target protein of 50 kDa resulted in up to 20 mg·mL−1 and 82% in dynamic binding capacity and utilization yield, respectively, at 100% breakthrough. The utilization of the binding sites was impacted by the ligand format and ligand density, rather than limitation in pore size of adsorbent as previously suggested. Overall, we demonstrated that multimeric fusion scFvs can successfully be developed and used as high‐capacity ligands in affinity adsorbents, enabling lean process design and alignment with process specifications.

In basic and applied biotechnology, design of affinity ligands has become essential for high-capacity applications such as affinity-based downstream processes for therapeutic molecules. Here, we established a proof-ofconcept for the use of multimeric fusion single-chain variable fragment (scFvs) as high-capacity ligands in affinity adsorbents. Mono-and di/tri-scFvs separated by Pro-rich negatively charged linkers were designed, produced, and immobilized to 6% cross-linked agarose beads. Frontal binding experiments with a target protein of 50 kDa resulted in up to 20 mgÁmL À1 and 82% in dynamic binding capacity and utilization yield, respectively, at 100% breakthrough. The utilization of the binding sites was impacted by the ligand format and ligand density, rather than limitation in pore size of adsorbent as previously suggested. Overall, we demonstrated that multimeric fusion scFvs can successfully be developed and used as high-capacity ligands in affinity adsorbents, enabling lean process design and alignment with process specifications.
Affinity chromatography is a commonly used technique for purification of specific target proteins. It has shown to be ideal for direct capture of low-expressing proteins from crude starting materials [1]. However, it can be challenging to find suitable affinity adsorbents for certain proteins as the commercial market is mostly limited to affinity adsorbents for purification of affinity-tagged target proteins [2-4] and monoclonal antibodies (mAbs) [5][6][7][8][9][10]. Nevertheless, if there is an antibody available against a target protein of interest, an immunoaffinity adsorbent can be developed by immobilization of the antibody to a chemically activated resin, such as cross-linked agarose, cellulose, or synthetic polymers [11].
Important characteristics to consider when developing affinity adsorbents are chemical stability, resin lifetime, and binding capacity. In particular, the binding capacity has a great impact on manufacturing costs and productivity [12]. At the manufacturing scale, affinity adsorbents with low binding capacity can require largesized columns with up to 2 m in diameter for increasing productivity. The drawbacks of these large columns include (a) additional costs due to large volumes of adsorbent, buffers, and consumables needed; (b) design of a facility to accommodate the space needed for the unit operation; and (c) pressure drop and unpredictable fluid distribution caused by issues in the column packing, and compression of affinity adsorbent [13].
The current trend in the preparation of immunoaffinity adsorbents has been to immobilize smaller fragments of antibodies to get higher dynamic binding capacities (DBC), such as the single-chain variable fragment (scFv) [14] and the heavy variable domains of camelid antibodies (nanobodies) [15]. The commercial immunoaffinity adsorbents KappaSelect, LambdaSelect, and VIIISelect are three examples of nanobody-based affinity adsorbents that provide a high binding capacity, typically between 8 and 20 mgÁmL À1 [10,16]. Nevertheless, maximum binding capacity of these adsorbents can only be achieved via monolayer adsorption as they accommodate one antigenic binding site per ligand only. The concept of a multimeric affinity ligand may be an alternative to increase capacity further and it has shown to be successful for the bacterial-derived protein A-based hexameric ligand in MabSelect [17]. However, a thorough literature search has returned no findings on the development of recombinant multimeric fusion antibody fragment-based affinity ligands, despite the development of multimeric antibody-based formats, such as diabodies and divalent scFvs, as emerging potential drugs and probes in the biopharmaceutical industry [18,19].
The aim of this study was to establish a proofof-concept for the use of multimeric fusion scFvs as high-capacity affinity ligands. The approach consisted of the following steps: (a) development of high-expressing dimeric (di-) and trimeric (tri-) scFv ligands by assessment of peptide linkers differing in rigidity and length, (b) site-directed immobilization of the multimeric scFv ligands via thiol groups to chemically activated cross-linked agarose beads, and (c) characterization of the developed affinity adsorbents by frontal affinity chromatography and adsorption isotherms. We report on the successful development of multimeric scFvs separated by negatively charged, rigid peptide linkers. We demonstrate that both monomeric (mono-) and multimeric scFv-based affinity adsorbents can result in close to fully accessible binding sites and substantially higher binding capacity than previously reported for immunoaffinity adsorbents [14].

Design of multimeric fusion scFvs
In order to develop a multimeric fusion protein, a suitable peptide linker has to be chosen. Peptide linkers can either be flexible, rigid, or cleavable depending on the desired characteristic [20]. In the case of multimeric fusion scFvs, a linker that can impose spatial separation is crucial for preserving independent folding and functionality. For this purpose, rigid linkers, such as helical structures [e.g., (EAAAK) n , where n is the copy number] or Pro-rich sequences [e.g., (XP) n , where X is any amino acid], have shown to be the desired choice [20,21]. Particularly, Pro-rich sequences have shown to impose stiff and extended conformations and their length can be adjusted by the copy number [22]. Hence, three different Pro-rich linkers differing in length and rigidity were chosen to be investigated in terms of expression level and spatial separation: (EP) k , (ESP) m , and (ESEP) n , where (k = 6 and 8), (m = 4 and 6), and (n = 3 and 4) (Fig. 1). Glu and Ser were selected as they had previously been shown to increase protein solubility [23].
Produced di-scFvs with linker length of 16-18 resulted in substantially higher recovery relative to that of 12 residues (Fig. S1). Therefore, the longer linker length was selected. The effect of the linker rigidity on the expression level was not as pronounced as the effect of the linker length. However, di-scFvs with the most rigid linker, (EP) 8 , seemed to result in slightly higher recovery relative to di-scFvs with (ESP) 6 and (ESEP) 4 linkers. Based on these findings, the (EP) 8 was selected for further development of a tri-scFv. Produced mono-, di-, and tri-scFv ligands were of high purity (Fig. 2), and their identities were verified by LC-MS analyses (data not shown).
To investigate whether the linkers could impose spatial separation between the scFvs, the temperature of the midpoint of thermal unfolding (T m ) was determined for the highest expressed ligands. Identical proteins that are separated by a linker should theoretically exhibit one unfolding transition at the same T m . All di-and tri-scFvs exhibited one unfolding transition at a temperature close to the T m of mono-scFv (Table S1 and Fig. S2). This indicates that all of the three rigid linkers could separate the scFvs. This is essential for preserving the accessibility of the antigenic binding sites of the ligands.

Characteristics of immunoaffinity adsorbents
Mono-, di-, and tri-scFvs were successfully immobilized, both individually and in combination, via thioldirected chemistry to 6% cross-linked agarose beads (sulfhydryl-reactive resin; Divbio Science, Ulvenhout, Netherlands). In total, nine different immunoaffinity resins were developed and characterized in terms of immobilization yield, ligand density, binding capacity, and utilization yield. The ligand densities and immobilization yields ranged from 0.08 to 0.40 µmolÁmL À1 , and 80 to 97%, respectively (Table 1 and Fig. 3). No apparent effect of the linker rigidity was observed on the binding capacity as the obtained DBC 100% was approximately equal at the same ligand density (15-17 mgÁmL À1 at 0.24-0.26 µmolÁmL À1 , see Table 1). Site-directed mono-scFv resulted in a DBC 100% and utilization yield of 17 mgÁmL À1 and 82%, respectively, at a ligand density of 0.40 µmolÁmL À1 . This is a further improvement compared to the previously reported random-oriented mono-scFv with DBC 100% and utilization yield of 10 mgÁmL À1 and 67%, respectively, at a ligand density of 0.30 µmolÁmL À1 [14].
Increased density of di-scFv from 0.08 to 0.24 µmolÁmL À1 resulted in increased DBC 100% from 6.8 to 17 mgÁmL À1 albeit decreased utilization yield by approximately 16%. Tri-scFv resulted in binding capacity and utilization yield of 20 mgÁmL À1 and 82%, respectively, thereby on par with the mono-scFv. Highest utilization yield at 50% breakthrough was obtained for mono-and tri-scFvs (63-68%) ( Fig. 3A and Table 1). Furthermore, maximum utilization of the capacity can be achieved by recycling of FT in a continuous capture mode [24].

Equilibrium binding capacity
Four of the developed affinity adsorbents were selected for batch adsorption experiments: mono-scFv (0.40 µmolÁmL À1 ), di-scFv (0.08 and 0.24 µmolÁmL À1 ), and tri-scFv (0.16 µmolÁmL À1 ). The obtained data at equilibrium were fitted to the Langmuir adsorption model (Fig. S3), which assumes monolayer adsorption. This was not expected as multilayer adsorption behavior was anticipated for the di-and tri-scFvs ligands. A model for multilayer adsorption was tested, that is, the Brunauer-Emmett-Teller model; however, similarly shaped isotherms as for the Langmuir model were obtained (data not shown) [25]. A summary of the estimations can be seen in Table 2. Overall, estimated q m seem to be reasonable as they do not exceed the theoretical capacities. Highest q m and utilization yield at equilibrium were obtained for tri-scFv (22 mgÁmL À1 ; 89%) and mono-scFv (19 mgÁmL À1 ; 92%). This suggests that the binding sites are close to fully accessible. In the case of the di-scFv ligands, slightly lower q m and utilization yields were obtained relative to the DBC 100% and utilization yields in Table 1.

Capture of target protein at low titer
A multimeric scFv affinity adsorbent, tri-scFv, was evaluated with Chinese hamster ovary (CHO) supernatant containing target protein X (< 10 µgÁmL À1 ). Although 1 L of CHO supernatant was applied to the tri-scFv column [column volume (CV) = 0.5 mL], the target protein could successfully be captured with sufficient purity (Fig. 4). This demonstrates that the multimeric adsorbent performs well at low titer.

Future studies and potential applications
As the concept of multimeric fusion scFvs showed to be developable, functional, and on par with the mono-scFv, a follow-up study should investigate the potential benefit of higher order fusion scFvs and optimize the ligand density for maximization of the DBC. In addition, ligand leaching, regeneration procedure, and resin reusability should be assessed. Furthermore, the use of high-capacity affinity ligands can be of particular interest in analytical applications such as  biosensors and ELISA. A common bottleneck in these assays is low sensitivity, thereby making detection of low-abundant molecules, for example, biomarkers, challenging. However, a way to improve the sensitivity is to increase the binding capacity in order to decrease the limit of detection [26][27][28]. Thus, a possibility to improve the sensitivity of such assays could be to use high-capacity multimeric fusion scFvs instead of conventional reagents.

Concluding remarks
Overall, we demonstrated that multimeric fusion scFvs can successfully be developed and used as high-capacity affinity ligands. The rigid linkers, (EP) 8 , (ESP) 6 , and (ESEP) 4 , successfully imposed separation, which is essential for preserving independent folding and facilitating accessible binding sites of the scFvs. Overall, site-directed immobilization of the ligands to 6% cross-linked agarose resins resulted in DBC 100% and utilization yields of up to 20 mgÁmL À1 and 82%, respectively. Altogether, when binding to a target protein of 50 kDa, ligand format, site direction, and ligand density had a substantial effect on the binding capacity, rather than a limitation in pore size as previously suggested [14]. As proof-of-concept was established for the suitability of multimeric scFvs as ligands in affinity adsorbents, future studies aim to optimize  properties such as multimeric ligand format and ligand density for maximization of binding capacity, as well as evaluate the chemical stability and resin lifetime.

Construct generation
DNA sequences encoding V H -(G 4 S) 3 -V L (scFv), scFv-linker-scFv, and scFv-linker-scFv-linker-scFv with N-terminal His 6 purification tags were designed based on the DNA sequence of a full-length anti-protein X antibody raised in mice. The linker was (EP) k , (ESP) m , and (ESEP) n , where (k = 6 and 8), (m = 4 and 6), and (n = 3 and 4). The DNA fragments were synthesized and cloned in the pTT5 vector by GeneArt (Thermo Fischer Scientific).

Protein expression and purification
Plasmid DNA production and transient protein expression in mammalian HEK293 cells (0.5-1 L scale) were performed as previously described [29]. Produced proteins were purified in a single step by immobilized metal ion affinity chromatography (IMAC) using Ni Sepharose High Performance column (16 mm 9 5 cm; CV, 10 mL) with a linear flow velocity of 75 cmÁh À1 (2.5 mLÁmin À1 ). Prior to column loading, 300 mM sodium chloride and 5 mM imidazole were added to the harvested protein supernatants (0.5-1.0 L). The running buffers were equilibration buffer (20 mM sodium phosphate, 0.5 M sodium chloride, 10 mM imidazole; pH 7.4) and elution buffer (20 mM sodium phosphate, 0.5 M sodium chloride, 0.5 M imidazole; pH 7.4). The scFv fragments were eluted by linear elution gradient (0-100% elution buffer for 10 CVs), and the collected elution fractions were pooled. Protein concentrations were estimated by NanoDrop TM 8000 Spectrophotometer (Thermo Fisher Scientific) at 280 nm (molar extinction coefficient 56160 M À1 Ácm À1 at 280 nm).

General analytical methods
The following standard analyses were performed according to standard methods described previously [30]. Qualitative protein analysis was performed by SDS/PAGE. Protein identity verification was performed by LC-MS. Temperature of midpoint of heat denaturation (T m ) was determined by nano differential scanning fluorimetry for protein samples in triplicates at 0.5 mgÁmL À1 in 50 mM sodium phosphate and 150 mM sodium chloride (pH 7.0).

Preparation of affinity adsorbents and affinity chromatography
Site-directed immobilization of the ligands via thiol groups to sulfhydryl-reactive resin, and determination of immobilization yield (h) and ligand density (q Ligand ) were performed as previously described [14]. In addition, the density of binding sites (q Binding site ) was determined according to Eq. 1.
where i is the index number of the ligand and N i is the number of antigen-binding sites of ligand i. The developed affinity adsorbents were packed into Tricorn 5/50 Columns (5 mm 9 2.5 cm; CV = 0.5 mL) and evaluated in terms of dynamic binding capacity (DBC) and utilization yield by frontal affinity chromatography with pure target protein X (0.4 mgÁmL À1 ) as previously described [14].

Capture of target protein from CHO expression
Calcium chloride was added to a final concentration of 20 mM to 1 L of CHO mammalian cell harvest containing target protein X (< 10 lgÁmL À1 ). Frontal affinity chromatography was carried out as previously described [14].

Batch adsorption experiments
Batch adsorption experiments were performed according to the following. Six different start concentrations of protein X in equilibration buffer (10 mM L-histidine, 100 mM sodium chloride, 25 mM calcium chloride; pH 6.0) were prepared at 0.5, 1.0, 2.0, 4.0, 6.0, and 8.0 mgÁmL À1 , respectively. Prior to experiments, affinity resins were washed at least three times with equilibration buffer. For each selected affinity resin, 0.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Table S1. Temperature of mid-point of heat denaturation (T m ) of mono-, di-and tri-scFvs with selected linkers.