In vivo protein stabilization based on fragment complementation and a split GFP system Stina Lindman, Armando Hernandez-Garcia1, Olga Szczepankiewicz, Birgitta Frohm, and Sara Linse2 Center for Molecular Protein Science, Biochemistry and Biophysical Chemistry, Lund University, SE-22100 Lund, Sweden Edited by Alan R. Fersht, MRC Centre for Protein Engineering, Cambridge, United Kingdom, and approved September 28, 2010 (received for review April 27, 2010) Protein stabilization was achieved through in vivo screening based on the thermodynamic linkage between protein folding and frag- ment complementation. The split GFP system was found suitable to derive protein variants with enhanced stability due to the correla- tion between effects of mutations on the stability of the intact chain and the effects of the same mutations on the affinity be- tween fragments of the chain. PGB1 mutants with higher affinity between fragments 1 to 40 and 41 to 56 were obtained by in vivo screening of a library of the 1 to 40 fragments against wild-type 41 to 56 fragments. Colonies were ranked based on the intensity of green fluorescence emerging from assembly and folding of the fused GFP fragments. The DNA from the brightest fluorescent colonies was sequenced, and intact mutant PGB1s corresponding to the top three sequences were expressed, purified, and analyzed for stability toward thermal denaturation. The protein sequence derived from the top fluorescent colony was found to yield a 12 °C increase in the thermal denaturation midpoint and a free energy of stabilization of −8.7kJ∕mol at 25 °C. The stability rank order of the three mutant proteins follows the fluorescence rank order in the split GFP system. The variants are stabilized through increased hydrophobic effect, which raises the free energy of the Fig. 1. Correlation between mutational effects on the stability of the unfolded more than the folded state; as well as substitutions, intact chain and the effects of the same mutations on the affinity between which lower the free energy of the folded more than the unfolded fragments of the chain (10–12). state; optimized van der Waals interactions; helix stabilization; improved hydrogen bonding network; and reduced electrostatic interactions in partially solvent exposed regions (4, 5). Screening repulsion in the folded state. methods with combinatorial approaches such as phage display have been applied to intact proteins and used increased protease green fluorescent protein ∣ protein stability ∣ protein stabilization method ∣ resistance as a selection criterion leading to variants with protein evolution ∣ protein reconstitution increased thermal stability (6). There are many reported cases of stabilized proteins by directed evolution (reviewed in ref. 7) he fold of a protein and its stability toward denaturation are and rational design (reviewed in ref. 8). In directed evolution, Tgoverned by the sequence of amino acids and the environment most strategies in screening for stability depend on activity mea- (solvent, salts, pH, T, crowding, etc.). The fold and stability of the surements during or after exposure to denaturing conditions (9). protein are dependent on a multitude of noncovalent interactions Here we introduce an alternative approach to protein stabili- (hydrogen bonds, electrostatic interactions, van der Waals inter- zation based on the thermodynamic linkage between fragment actions, hydrophobic effect) under the steric constraints imposed complementation and protein folding relying on in vivo screening by the covalent chain. With protein stability defined as the free under physiological conditions. Our approach is based on the energy difference between the folded and unfolded state, the correlation between mutational effects on the stability of the in- stability will increase with any factor that raises the free energy tact chain and the effects of the same mutations on the affinity of the unfolded state more than the folded state and with any – factor that lowers the free energy of the folded state more than between fragments of the chain (10 12) as illustrated in Fig. 1. the unfolded state. Screening based on fragment complementation and phage A common problem with proteins used in therapeutic, diag- display has previously been used to study the role of particular nostic, or technical applications is their limited stability toward amino acids in the specificity of folding and stabilization of different forms of handling and storage, or after administration the corresponding intact proteins (13, 14). in the body. The proteins may be sensitive to denaturation and/or proteolytic degradation. The therapeutic and technical value of the Author contributions: S. Lindman, O.S., and S. Linse designed research; S. Lindman, proteins may therefore increase if their stability can be improved. A.H.-G., O.S., B.F., and S. Linse performed research; S. Lindman, A.H.-G., B.F., and S. Linse Several approaches have been used to stabilize proteins analyzed data; and S. Lindman, A.H.-G., and S. Linse wrote the paper. through formulation or through redesign of the amino acid The authors declare no conflict of interest. sequence. Formulation with sugars, liquid crystals, polymeric na- This article is a PNAS Direct Submission. noparticles, liposomes, emulsions, or microencapsulation, or by 1Present address: Laboratory of Physical Chemistry and Colloid Science, Wageningen site-specific glycosylation, may protect the native protein toward University, P.O. Box 8038, 6700 EK Wageningen, Netherlands. unfolding and/or degradation (reviewed in refs. 1–3). Computa- 2To whom correspondence should be addressed. E-mail: [email protected]. tional approaches to improve protein stability have focused on This article contains supporting information online at www.pnas.org/lookup/suppl/ improving the packing of the hydrophobic core and on optimizing doi:10.1073/pnas.1005689107/-/DCSupplemental. 19826–19831 ∣ PNAS ∣ November 16, 2010 ∣ vol. 107 ∣ no. 46 www.pnas.org/cgi/doi/10.1073/pnas.1005689107 Downloaded by guest on September 25, 2021 The combined interactions, which stabilize the protein fold, folding. As a test case we use the 56-residue B1 domain of protein are often strong and specific enough to allow for reconstitution G (PGB1) with a β-grasp fold (27) that can reconstitute from of the protein from fragments of the amino acid chain. Hence, an fragments (28). We use a split GFP system, described in depth intact covalent chain of amino acids is not necessary to maintain elsewhere (29–31), to screen for PGB1 mutants with higher affi- the fold and function of a protein. This was found in the 1950s nity between two fragments comprising residues 1 to 40 and 41 to – for ribonuclease (15) and in the 1970s 2000s for a number of 56, respectively (Fig. 2) as visualized from brighter green fluor- – other proteins (see, for example, refs. 16 24). The reconstituted escence. After expression and purification of the corresponding complex has lower and concentration-dependent stability toward intact mutants we show that this leads to the production of mark- denaturation than the intact protein, but functions like receptor edly stabilized protein variants. The protein sequence derived or ligand binding are often retained. As illustrated in Fig. 1, there is a strong correlation between from the top fluorescent colony yields a 12 °C increase in the mid- mutational effects on the stability of the intact chain and the point of thermal denaturation and a free energy of stabilization of −8 7 ∕ effects of the same mutations on the affinity between fragments . kJ mol at 25 °C. of the chain (10–12, 25). This implies that proteins may be sta- Results bilized using a method based on protein reconstitution from frag- Nomenclature. PGB1-QDD (also called parent) denotes a variant ments and screening of a fragment library for enhanced affinity. Streptococcus sp. Preliminary studies using the split GFP system and fragment com- of PGB1 (protein G B1-domain from ) with the plementation systems of widely different affinities and stabilities three substitutions T2Q (to avoid N-terminal processing) (32), pointed to a markedly brighter green fluorescence in the higher N8D, and N37D (to avoid covalent rearrangement during deami- affinity/stability system (26). Here we built on this finding to dation) (33, 34). PGB1-40-CGFP denotes a fusion construct in develop an in vivo screening method that can be used to identify which residues 1 to 40 of PGB1-QDD are followed by a seven- reconstitution couples of improved affinities and subsequently to residue linker and then residues 158 to 238 of GFP. NGFP- produce intact protein variants with increased stability toward un- PGB41-56 denotes a fusion construct in which a hexa-histidine A BC D Fig. 2. (A) A cartoon model of the ribbon E structure of split GFP with fused PGB1 fragments. The model is produced using BIOPHYSICS AND MOLMOL (42) and the pdb files 1ema (43) and 1pgb (44). (B) Topology of PGB1 COMPUTATIONAL BIOLOGY fragments fused to GFP fragments. (C) Comparison of fluorescence intensity of Top1 and parent construct as well as nega- tive (noninduced) and positive (CBD9k) controls. (D) The sequences of the fused fragments where the underlined part re- CHEMISTRY presents the PGB1 fragments. (E) Outline of the library with 3,456 possible se- F quences. Previously determined mutants that are stabilizing (F3, I7, I16, I18, E25, F29, I39), destabilizing (L18, Q25, K29), and possible mutations arising from the nature of the genetic code are included in the library. (F) Structure of parent pro- tein where the seven library positions are indicated. The structures are produced using MOLMOL (41). Lindman et al. PNAS ∣ November 16, 2010 ∣ vol. 107 ∣ no. 46 ∣ 19827 Downloaded by guest on September 25, 2021 tag is followed by residues 1 to 157 of GFP, then an eight-residue protein (around 150 mg∕L culture) were obtained for all three linker, and then residues 41 to 56 of PGB1 (Fig. 2). variants. The proteins were judged homogeneous based on analysis by SDS/PAGE, analytical gel filtration chromatography, Coexpression of Parent Constructs.