Engineering bispecificity into a single albumin-binding domain

Abstract

Bispecific antibodies as well as non-immunoglobulin based bispecific affinity proteins are considered to have a very high potential in future biotherapeutic applications. In this study, we report on a novel approach for generation of extremely small bispecific proteins comprised of only a single structural domain. Binding to tumor necrosis factor-α (TNF-α) was engineered into an albumin-binding domain while still retaining the original affinity for albumin, resulting in a bispecific protein composed of merely 46 amino acids. By diversification of the non albumin-binding side of the three-helix bundle domain, followed by display of the resulting library on phage particles, bispecific single-domain proteins were isolated using selections with TNF-α as target. Moreover, based on the obtained sequences from the phage selection, a second-generation library was designed in order to further increase the affinity of the bispecific candidates. Staphylococcal surface display was employed for the affinity maturation, enabling efficient isolation of improved binders as well as multiparameter-based sortings with both TNF-α and albumin as targets in the same selection cycle. Isolated variants were sequenced and the binding to albumin and TNF-α was analyzed. This analysis revealed an affinity for TNF-α below 5 nM for the strongest binders. From the multiparameter sorting that simultaneously targeted TNF-α and albumin, several bispecific candidates were isolated with high affinity to both antigens, suggesting that cell display in combination with fluorescence activated cell sorting is a suitable technology for engineering of bispecificity. To our knowledge, the new binders represent the smallest engineered bispecific proteins reported so far. Possibilities and challenges as well as potential future applications of this novel strategy are discussed.

Figure 1. Structure of the three-helical albumin-binding domain (ABD) from streptococcal protein G (PDB: 1GJT).

Helices are numbered as 1, 2 and 3 and indicated in the figure. A. Indicated in blue are amino acids on helix 2 and 3 (S18, Y20, Y21, K22, N23, L24, K29 and E32) that have been suggested to participate in the interaction to albumin (defined as resulting in at least a 2-fold decrease in affinity when mutated to alanine [16]). B. Indicated in green are amino acids on helix 1 and 3 that were targeted for randomization in the library. The library was designed in order to randomize a surface that was non-overlapping with the positions that are interacting with albumin. C. Structure of ABD observed from above with the albumin-interacting surface on helix 2 and 3 indicated in blue and the randomized surface on helix 1 and 3 indicated in green.

Figure 2. Selection strategy for phage display enrichment of TNF-α-binding ABD-variants and sequences of bispecific first-generation molecules.

A. Four cycles of selection, split into four parallel tracks, were performed. The percentage of Tween 20 and the number of washes were increased in each round of selection. Neutravidin coated paramagnetic beads (N) were used to capture phage-target complexes in cycles 1-3, and in cycle 4 streptavidin coated beads (S) were used. B. Amino acid sequences of the bispecific variants ABD_TNF1 and ABD_TNF2 and the number of times those clones were observed. Non-randomized ABD is shown as a reference and randomized positions are indicated with bullets.

Figure 3. Staphylococcal display vector and schematic representation of the cell-surface displayed ABD-library.

A. Staphylococcal display vector for display of heterologous proteins on the surface of S. carnosus. Abbreviations: Z2, dimeric Z-domain gene; Bla, beta-lactamase encoding gene; Cmlr, chloramphenicol acetyl transferase-encoding gene; OriE and OriS, origin of replication for E. coli and staphylococci, respectively; PP, propeptide from S. hyicus lipase; PLip, promoter from S. hyicus lipase. S, signal sequence from S. hyicus lipase; XM, cell wall anchoring region from staphylococcal protein A. B. Schematic representation of the recombinant ABD-library in fusion to the dimeric Z-domain displayed on the staphylococcal cell surface. The dimeric Z-domain functions as a reporter tag for monitoring of the surface expression level of individual cells through binding to labeled IgG. The target-binding signal can thereby be normalized with the expression level in order to minimize expression bias during sorting. Labeled proteins in the first and secondary incubations for the dual selection are indicated in the figure.

Figure 4. Randomized positions in the affinity maturation library (Sc∶ABD_TNFlib).

A. Sequence in randomized positions of the first generation binders ABD_TNF1, ABD_TNF2 and Sc∶ABD_TNFlib. B. Positions targeted for randomization indicated by light circles. Three positions randomized in the initial library were omitted in the design, indicated by dark circles. The randomizations are listed beside the position. ARA  =  Arg, Lys. BBG  =  Ala, Arg, Gly, Leu, Pro, Ser, Trp, Val. GAK  =  Asp, Glu. NYC  =  Ala, Ile, Leu, Phe, Pro, Ser, Thr, Val. NNK  =  all amino acids and one amber stop. The theoretical size of the library is 6.6·10⁶.

Figure 5. Density plots showing the results from flow-cytometric sortings of Sc∶ABD_TNFlib.

A. The density plots are showing the staphylococcal library before flow-cytometric sorting round 1, 2 and 3, respectively, with regions used for gating outlined in each plot. The target protein and the concentration used in each sorting are indicated on top of respective density plot. FL-4 channel fluorescence intensity corresponding to surface expression level (monitored via IgG-binding) on the x-axis and FL-1 channel fluorescence corresponding to TNF-α- or HSA-binding on the y-axis. B. Density plots showing the analysis of bispecificity of Sc∶ABD_TNFlib in cycle 2 and cycle 3 (incubated with 100 nM TNF-α and 300 nM HSA). FL-1 channel fluorescence intensity corresponding to HSA-binding on the x-axis and FL-2 channel fluorescence corresponding to TNF-α-binding on the y-axis.

Figure 6. Amino acid sequences, origin and binding affinities of clones from the third round of FACS.

A. The sequence of non-randomized ABD is shown on top with the eight positions subjected to randomization in Sc∶ABD_TNFlib indicated. Boxes denote the localization of the three helices in the structure of the parental ABD-molecule. The candidates are named and grouped according to selection strategy. Variants from selections targeting only TNF-α (tracks I-IV, ABD_T001-013) and from the dual-selection (tracks V-VI, ABD_HT014-023) are shown. The columns to the right indicate the number of times each clone was identified by DNA-sequencing and the selection track(s) from where the sequence originated. B. Representative sensorgrams from SPR-analysis of immobilized ABD_T001 and ABD_HT014 binding to TNF-α and HSA. TNF-α was injected at concentrations ranging from around 6–100 nM and HSA from 10–180 nM, data are double referenced by subtraction of simultaneous responses from interspots and a buffer injection.