Directed evolution of anti-HER2 DARPins by SNAP display reveals stability/function trade-offs in the selection process

Abstract

In vitro display technologies have proved to be powerful tools for obtaining high-affinity protein binders. We recently described SNAP display, an entirely in vitro DNA display system that uses the SNAP-tag to link protein with its encoding DNA in water-in-oil emulsions. Here, we apply SNAP display for the affinity maturation of a designed ankyrin repeat proteins (DARPin) that binds to the extracellular domain of HER2 previously isolated by ribosome display. After four SNAP display selection cycles, proteins that bound specifically to HER2 in vitro, with dissociation constants in the low- to sub-nanomolar range, were isolated. In vitro affinities of the panel of evolved DARPins directly correlated with the fluorescence intensities of evolved DARPins bound to HER2 on a breast cancer cell line. A stability trade-off is observed as the most improved DARPins have decreased thermostability, when compared with the parent DARPin used as a starting point for affinity maturation. Dissection of the framework mutations of the highest affinity variant, DARPin F1, shows that functionally destabilising and compensatory mutations accumulated throughout the four rounds of evolution.

Keywords: DARPin; SNAP display; alternative scaffold; antibody; directed evolution; in vitro compartmentalisation; trade-off.

Fig. 1

SNAP display selection scheme. (1) Water-in-oil emulsion droplets compartmentalise DNA library members separately, such that one gene (or none) is Poisson-distributed in droplets. In vitro expression is performed from single linear DNA templates. (2) The SNAP-tag forms a covalent thioether bond between the protein-coding DNA (bearing a covalently-linked SNAP-substrate, benzylguanine, BG) and the corresponding expressed protein (Keppler et al., 2003). The stability of the covalent genotype–phenotype linkage enables panning selections (3) to be performed under a wider range of conditions without the risk of disassembly of genotype and phenotype. Panning conditions are tuned to achieve the required level of stringency for the isolation of binders (4). The DNA recovered from selections is amplified by PCR (5). After each round of selection, additional mutations are easily introduced during PCR by low fidelity polymerases. Reproduced with permission from Houlihan et al. (2014).

Fig. 2

Primary screen of DARPins selected against HER2 using SNAP display. Four rounds of affinity selections were performed and 90 DARPins from the initial library, second and fourth round were tested for binding in a crude extract ELISA. Binding of DARPins to HER2 was detected using an HRP-conjugated antibody. The orange line indicates the ELISA signal of DARPin G3-HAVD in each screen. Over 90% of the clones screened gave a signal >5-fold over background while ∼30% of the selected DARPins showed greater binding signals compared with the parent DARPin G3-HAVD.

Fig. 3

(A) The frequency of mutations measured in the sequenced DARPin mutants is plotted onto the structure of DARPin H10-2-G3 (shown as putty cartoon): frequently mutated positions are colour-coded and rendered with backbone thickness proportional to mutation rate. Fully conserved positions are shown in dark blue. The position that accumulated ∼42% of sequence diversity, H52, is shown in red. The positions most frequently mutated are in proximity of the binding interface. Overall, positions mutated in greater than 10% of the sequenced population are all located in framework positions in the designed N2C DARPin library. Mutations do, however, map also on framework positions away from the binding interface. The interacting domain of HER2 is represented as a semi-transparent sand-coloured surface. (B) The published X-ray structure of DARPin H10-2-G3 (green) in complex with HER2 (blue) (Jost et al., 2013) was used to visualise the likely locations of the mutations obtained in DARPin F1. DARPin H10-2-G3 was affinity matured from the same parent clone (DARPin G3 HAVD) as the DARPins evolved by SNAP display. Mutations that contribute to binding are coloured brown while neutral mutations are coloured yellow (see Table III). Figures were prepared with Pymol based on PDB 4HRN (Jost et al., 2013).

Fig. 4

(A) ELISA analysis of selected DARPins. The top 10 DARPins (10 μg/ml of purified DARPin) were analysed for binding to immobilised HER2. All selected DARPins from SNAP display outputs gave a greater binding signal than DARPin G3-HAVD. All evolved DARPins were also analysed for their specificity in binding. No binding was observed to Streptavidin (which was used to capture biotinylated HER2 during selections) or lysozyme indicating the selected DARPins do not non-specifically bind to other proteins. Competition for binding of selected DARPins to HER2 with DARPin H10-2-G3 (10-fold excess DARPin H10-2-G3 over selected DARPins) showed that each selected DARPin competed for the same epitope as DARPin H10-2-G3. (B) Correlation between the observed fluorescence values of cell populations and the reciprocal of their binding affinity shows a correlation between the two properties. Most notably, Trastuzumab appears to have a significantly lower affinity on cells, probably as a result of the different binding mode and epitope the antibody binds to on HER2.

Fig. 5

Thermal stability analysis of the selected DARPins. (A) Melting temperatures of the selected DARPins were measured by differential scanning fluorimetry (plotted based on increase in affinity going from right to left). Measurements were performed in Tris buffer and repeated twice for all mutants. (B) Correlation between melting temperature and affinity of selected DARPins. An order of magnitude in binding strength leads to a ∼8°C loss in stability (correlation coefficient r = 0.43; fit not shown).