Tailoring in vitro evolution for protein affinity or stability

Abstract

We describe a rapid and general technology working entirely in vitro to evolve either the affinity or the stability of ligand-binding proteins, depending on the chosen selection pressure. Tailored in vitro selection strategies based on ribosome display were combined with in vitro diversification by DNA shuffling to evolve either the off-rate or thermodynamic stability of single-chain Fv antibody fragments (scFvs). To demonstrate the potential of this method, we chose to optimize two proteins already possessing favorable properties. A scFv with an initial affinity of 1.1 nM (k(off) at 4 degrees C of 10(-4) s(-1)) was improved 30-fold by the use of off-rate selections over a period of several days. As a second example, a generic selection strategy for improved stability exploited the property of ribosome display that the conditions can be altered under which the folding of the displayed protein occurs. We used decreasing redox potentials in the selection step to select for molecules stable in the absence of disulfide bonds. They could be functionally expressed in the reducing cytoplasm, and, when allowed to form disulfides again, their stability had increased to 54 kJ/mol from an initial value of 24 kJ/mol. Sequencing revealed that the evolved mutant proteins had used different strategies of residue changes to adapt to the selection pressure. Therefore, by a combination of randomization and appropriate selection strategies, an in vitro evolution of protein properties in a predictable direction is possible.

Figure 1

Analysis of off-rate selections: semiquantitative inhibition RIA of selected c12 mutants (for details see Experimental Protocol). Kinetic and affinity parameters of purified c12 and three mutants were determined experimentally as described in the text. Because of the similar RIA profile of all mutants three c12 mutants of this set along with the parental protein were further analyzed. c12 wild-type: k_off = 0.0028 s⁻¹, k_on = 2.5 × 10⁶ M⁻¹ s⁻¹, K_{D, calc} = 1.1 nM, K_{D, titr} = 1.5 nM; c12 B5–6: k_off = 0.00014 s⁻¹, k_on = 3.4 × 10⁶ M⁻¹ s⁻¹, K_{D, calc} = 0.04 nM, K_D,titr = 0.1 nM; c12 B5–2: k_off = 0.00020 s⁻¹; c12 B5–9: k_off = 0.00018 s⁻¹. Measurements were performed at 25°C as described (19). K_{D, calc} is the ratio obtained from the dissociation rate (k_off) and the association rate (k_on), and K_{D, titr} is the equilibrium dissociation constant directly determined by titration. The error between duplicate measurements for k_off and k_on of all measured mutants is about ±3%; the error for K_{D, titr} is about ±8%.

Figure 2

Sequences of the selected clones. (A and B) Aligned sequences of affinity-matured c12 mutants (A) and of stability-matured hag mutants (B). Residue numbering and CDR localization are according to Kabat et al. (33).

Figure 3

Localization of the mutated residues in a three-dimensional model of c12 with docked fluorescein (theoretical model obtained by homology modeling, based on Protein Data Bank (PDB) entries 1iai, 1nca, 1igc for V_L; 1for, 1plg for V_H; and 1mim for CDR-H3) (A) and in the anti-influenza hemagglutinin antibody Fab 17/9 [experimental structure, PDB entry 1ifh (34)] (B). V_H is shown in blue; V_L is in purple. The strongly selected mutations anti-hag L83 Leu to Gln, c12 L94 His to Tyr, and c12 H101 Asp to Gly, Ser, or Ala and its salt-bridge partner Arg H94 are indicated in red. The positions of the remaining mutated residues are indicated by orange if the affected side chains are buried in the domain core and by green if they are located on the surface of the molecule.

Figure 4

Analysis of stability-matured hag mutants after in vivo and in vitro expression. Absolute RIA signal after reducing and oxidizing in vitro translation (A) and demonstration of specific binding by inhibition with 1 μM soluble hag peptide signals from in vitro translation under oxidizing conditions (ox) or reducing conditions (red) are shown. + or −sol. peptide indicates the presence or absence of 1 μM hag peptide as a competitor. By subtraction of the radioactivity of samples that were inhibited from the radioactivity of the uninhibited samples, the inhibitable portion of the RIA signal was obtained as a measure for specific binding activity. The resulting ratios of relative binding are calculated by dividing the inhibitable portion of the RIA signal in the presence of 10 mM DTT by the inhibitable RIA signal in the absence of DTT. They are as follows: anti-hag scFv wild-type: 0.2%; h5–5: 10.9%; h5–12: 1.6%; h6–4: 40.8%; h6–6: 11.1%; h6–11: 13.1%; h6–12: 26.4%; h6–22: 84.5%; h6–28: 20.9%; h6–33: 17.9%; h6–35: 17.0%. Urea denaturation at 10°C (B) of periplasmically produced hag wild-type and mutant proteins. ▴, Anti-hag wild-type scFv; ○, mutant h6–11; □, mutant h6–4; ⧫, mutant h6–22. (C) Urea equilibrium renaturation under reducing conditions at 20°C of periplasmically produced hag wild-type and mutant proteins. (D) Normalized crude cell lysate ELISA after expression of hag wild-type and mutant proteins at 25°C in the cytoplasm of E. coli. Bars represent the binding signals of active protein in the absence (−) and presence (+) of competing soluble hag-peptide (1 μM), to indicate specific binding.