Nuclear magnetic resonance spectroscopy solution structures of proteins S-824 and S-836. Top: Ribbon diagrams of four-helix bundle proteins. Bottom: Space-filling models of corresponding proteins with non-polar amino acids sequestered in the core (yellow) and polar amino acids on the exterior (red or blue). Figure adapted from Wei et al . (2003b) and Go et al . (2008). 

Contexts in source publication

Context 1

... : binary code/directed evolution/peroxidase/protein design/synthetic biology Designing proteins de novo provides a unique opportunity to explore sequences, structures and functions that are chem- ically reasonable, but which have not been sampled by biological evolution. In the initial steps toward such explorations, we designed combinatorial libraries of a -helical sequences, determined the three-dimensional structures of several proteins from these libraries, and probed their potential for binding and catalytic function (Kamtekar et al ., 1993; Rojas et al ., 1997; Moffet et al ., 2000; Wei et al ., 2003a,b; Hecht et al ., 2004; Go et al ., 2008; Patel et al ., 2009; Fisher et al ., 2011; Arai et al ., 2012; Cherny et al ., 2012). Combinatorial libraries of a -helical proteins were constructed using the binary code strategy for protein design. A canonical alpha helix has 3.6 residues per turn. Accordingly, a non-polar amino acid is placed every 3 or 4 positions in the primary sequence, generating the following pattern: W † WW †† WW † WW †† W , where W and † represent arbitrarily chosen polar and non-polar residues, respectively (Kamtekar et al ., 1993; Hecht et al ., 2004). This pattern encodes an amphipathic helix, where all non-polar residues align on one face of the helix and all polar residues align on the opposite face. When four such helices are linked, the hydro- phobic effect drives them to pack against one another, thereby forming a four-helix bundle with non-polar residues pointing toward the protein core, and polar residues exposed to solvent. Several proteins from previously constructed libraries have been shown to form stable well-ordered structures (Wei et al ., 2003a). Furthermore, determination of the structures of two proteins (S-824 and S-836) provided evidence that these binary- patterned proteins folded into four-helix bundles as designed (Wei et al ., 2003b; Go et al ., 2008) (Fig. 1). Although the proteins in our binary-patterned libraries were not explicitly designed for any particular activities, characterization of several hundred arbitrarily chosen sequences showed that many of them bound the heme cofactor, and several of these novel heme proteins catalyzed peroxidase activity at levels substantially above background (Patel et al ., 2009). The observation of cofactor binding and catalytic activity in proteins from a library of a -helical bundles that had neither been designed nor selected ( in vivo or in vitro ) for function suggested that this superfamily of de novo proteins could be viewed as a novel ‘feedstock’ for evolution. Thus, by starting with a binary-patterned protein with relatively low-level activity, one can mimic natural selection by intro- ducing random mutations and screening for variants with improved activity. This process is reminiscent of numerous directed evolution experiments that have been conducted on various natural proteins (Joo et al ., 1999; Farinas et al ., 2001; Glieder et al ., 2002; Gould and Tawfik, 2005). However, there is a significant difference. Previous experiments in laboratory-based evolution started with sequences that were themselves products of billions of years of natural selection. In contrast, laboratory evolution of our binary- patterned sequences enables the exploration of evolutionary trajectories emanating from naive sequences that are not biased by natural selection for biological function in living organisms. Starting with the previously characterized binary-patterned proteins, S-824 and S-836, we demonstrate that mutagenesis of a de novo -designed protein, followed by screening of a relatively small number (hundreds or thousands) of variants can yield novel sequences with improved peroxidase activity. Error-prone polymerase chain reaction (PCR) using nucleotide analogs was used to introduce random mutations into genes encoding the novel proteins (Zaccolo et al ., 1996). The nucleotide analogs d-PTP (2 0 -deoxy-P-nucleoside-5 0 triphosphate) and 8-oxo-dGTP (8-oxo-2 0 -deoxyguanosine- 5 0 -triphosphate) were obtained from TriLink BioTechnologies. The products were gel purified (ZymoClean) and amplified using error-free PCR. The final PCR products were purified, concentrated (ZymoClean) and double-digested with NdeI and BamHI. Digestion mixtures were run on a low-melting 2% agarose gel, and the band at 300 bp was excised and purified. The mutation rate was three to four base mutations per sequence of 306 bp. Mutagenized gene sequences were cloned into the pET-3a vector. Previously, a stuffer fragment of lambda DNA had been cloned into the vector to ensure that only double-cut vector would be isolated following double di- gestion and gel purification. The purified double-digested vector was ligated with the library insert at 16 8 C for 12 h, desalted and electroporated into ElectroMAX competent cells (200 m l) at 2.3 kV (Transporator Plus) using a 2-mm gap cuvette. Transformation efficiency was 10 8 colony forming units/ m g DNA. After overnight growth, the colonies were scraped and pooled, and libraries of plasmid DNA were isolated for subsequent experiments. BL21(DE3) Escherichia coli cells were transformed via elec- troporation with plasmid library DNA, and the cells were grown overnight at 37 8 C on LB/amp (100 m g/ml) plates. The colonies were picked and placed in 96-deep well plates containing auto-induction media (600 m l) and grown overnight at 37 8 C on a plate shaker. Auto-induction media was prepared by combining glycerol (0.4% v/v), glucose (0.05% w/v) and a -lactose (0.2% w/v) in LB/amp (Studier, 2005). Auto-induction media facilitates overexpression of the target protein without the need to monitor cell density or add isopro- pyl b -D-1-thiogalactopyranoside (IPTG). Following protein expression, aliquots of 100 m l were saved and the remaining cells were harvested by centrifugation. The cells were dis- rupted by addition of BugBuster lysis solution (200 m l; Novagen) and activity buffer (300 m l, 50 mM Tris – HCl, pH 8). The lysed cells were centrifuged and the supernatant was assayed for peroxidase activity. Samples for the assay were prepared by mixing protein supernatant (50 m l), hemin chloride (5 m M final concentration; Sigma), ABTS (1 mg/ml final concentration; Sigma), H 2 O 2 (0.006% final concentration) in activity buffer (50 mM Tris –HCl, pH 8) for a final volume of 200 m l. After addition of H 2 O 2 , product formation was detected at 650 nm using a plate reader. BL21(DE3) cells were transformed and grown overnight on LB/amp (100 m g/ml) plates as described above. The colonies were replica plated using nitrocellulose filter paper (Millipore). The original cells were grown at 37 8 C until the colonies reformed. Duplicate colonies on the nitrocellulose filter paper were placed on LB/amp (100 m g/ml) IPTG (200 m g/ml) plates for 4 h at room temperature to induce protein expression. The cells were ...

Context 2

... purified double-digested vector was ligated with the library insert at 16 8 C for 12 h, desalted and electroporated into ElectroMAX competent cells (200 m l) at 2.3 kV (Transporator Plus) using a 2-mm gap cuvette. Transformation efficiency was 10 8 colony forming units/ m g DNA. After overnight growth, the colonies were scraped and pooled, and libraries of plasmid DNA were isolated for subsequent experiments. BL21(DE3) Escherichia coli cells were transformed via elec- troporation with plasmid library DNA, and the cells were grown overnight at 37 8 C on LB/amp (100 m g/ml) plates. The colonies were picked and placed in 96-deep well plates containing auto-induction media (600 m l) and grown overnight at 37 8 C on a plate shaker. Auto-induction media was prepared by combining glycerol (0.4% v/v), glucose (0.05% w/v) and a -lactose (0.2% w/v) in LB/amp (Studier, 2005). Auto-induction media facilitates overexpression of the target protein without the need to monitor cell density or add isopro- pyl b -D-1-thiogalactopyranoside (IPTG). Following protein expression, aliquots of 100 m l were saved and the remaining cells were harvested by centrifugation. The cells were dis- rupted by addition of BugBuster lysis solution (200 m l; Novagen) and activity buffer (300 m l, 50 mM Tris – HCl, pH 8). The lysed cells were centrifuged and the supernatant was assayed for peroxidase activity. Samples for the assay were prepared by mixing protein supernatant (50 m l), hemin chloride (5 m M final concentration; Sigma), ABTS (1 mg/ml final concentration; Sigma), H 2 O 2 (0.006% final concentration) in activity buffer (50 mM Tris –HCl, pH 8) for a final volume of 200 m l. After addition of H 2 O 2 , product formation was detected at 650 nm using a plate reader. BL21(DE3) cells were transformed and grown overnight on LB/amp (100 m g/ml) plates as described above. The colonies were replica plated using nitrocellulose filter paper (Millipore). The original cells were grown at 37 8 C until the colonies reformed. Duplicate colonies on the nitrocellulose filter paper were placed on LB/amp (100 m g/ml) IPTG (200 m g/ml) plates for 4 h at room temperature to induce protein expression. The cells were then lysed by placing the nitrocellulose paper on top of filter paper soaked in BugBuster lysis solution (Novagen). The nitrocellulose paper was then moved to filter paper soaked in hemin chloride solution (5 m M; Sigma) to allow proteins in the lysate to bind heme. Finally, the nitrocellulose paper was moved to a Petri dish containing ABTS (2 mg/ml, Sigma) and H 2 O 2 (0.06%) in activity buffer (50 mM sodium phosphate, pH 6). Colonies with peroxidase activity turned green and were marked. Replica colonies from the original plate were used to recon- firm activity, to make cell stocks and for plasmid preparations. A single colony was used to inoculate 10 ml of LB/amp (100 mg/l final concentration) which was grown overnight at 37 8 C. This liquid culture was used to inoculate 1 l of LB/ amp (100 mg/l final concentration) at a 1 : 1000 ratio. The 1 l culture was grown at 37 8 C until reaching an OD600 of 0.6 and IPTG (200 mg/l final concentration) was added to induce expression. Protein was released from cells using the freeze-thaw method (Johnson and Hecht, 1994). The protein was resuspended in MgCl 2 (100 mM; 10 ml per 1 l of culture) to extract the protein from the perforated cells, and cellular debris was removed by centrifugation. The supernatant was then acidified (1 M sodium acetate buffer, pH 4; 1 ml per 1 l of culture) and cellular contaminants were removed by acid precipitation and centrifugation at 6000 Â g for 10 min. The resulting supernatant was loaded onto a Poros HS cation exchange column and purified using a gradi- ent of NaCl. The protein was concentrated and exchanged into activity buffer using Centricon Plus-20 filters (Millipore). Samples for the peroxidase assay were prepared by mixing protein (20 m M) with hemin chloride (5 m M) and ABTS (1 mg/ml) in activity buffer (50 mM Tris – HCl, pH 8). A hemin chloride stock solution (1 mM) was prepared in dimethyl sulfoxide, and an ABTS stock solution (20 mg/ml) was prepared in activity buffer. As the heme concentration is much lower than protein concentration, we assume the maximal concentration of heme – protein complex is 5 m M. On addition of H 2 O 2 (0.086 m M – 11 m M), time points were recorded for 30 min at 650 nm. Kinetic data were recorded using the Varioskan Flash Spectral Scanning Multimode Reader and SkanIt software (Thermo Fisher Scientific, Waltham, MA). Peroxidase rate constants were analyzed using Michaelis – Menten kinetics: 1/ V 1⁄4 K M /( k cat [E 0 ][S]) þ 1/( k [E ]). The de novo- designed four-helix bundles, S-824 and S-836, were targeted for directed evolution. These sequences were chosen as test cases because both proteins had been characterized extensively, and their structures were solved previously by Nuclear magnetic resonance spectroscopy, as shown in Fig. 1 (Wei et al ., 2003a,b; Go et al ., 2008). Furthermore, as both proteins are very stable, it seemed likely that their scaf- folds would tolerate mutations. Mutations in both sequences were created by error-prone PCR using nucleotide analogs (Zaccolo et al ., 1996). Mutant libraries of both sequences were initially screened for peroxidase activity in 96-well plates. These screens provide clear results, but are difficult to scale up without robotic instrumentation. An example of a 96-well plate assay for peroxidase activity in cell lysates for a library of S-836 mutants is shown in Fig. 2. The range in the intensity of color indicates a range of activity in the mutant library. Not surprisingly, many mutant sequences display lower levels of activity than the parental S-836 sequence. However, several mutants appear to be more active than S-836. The sequences responsible for the darkest color were chosen for further rounds of mutagenesis and screening. To facilitate sampling of larger libraries, we also devel- oped a colony-based screen. For this screen, the cells were grown, expressed, lysed and assayed for activity in colonies growing on filter paper above agar Petri dishes. When the colonies are exposed to substrate, those that change color first are picked for further analysis. In our hands, the 96-well format was convenient for screening several hundred sequences, whereas the colony-based assay allowed screening of thousands of clones. Figure 3 shows a control experiment for a colony screen. Each plate contains colonies expressing a single known sequence. The negative control (left) shows colorless colonies expressing sequence WA9, which has no peroxidase activity. In contrast, the positive control (right) shows colored colonies for cells expressing the active sequence, WA63. Thus, colonies expressing sequences with different levels of peroxidase activity can be distinguished readily. Initially, we used 96-well plates to screen lysates for enhanced peroxidase activity amidst libraries of mutants of proteins S-824 and S-836. Figure 4a shows the results for a single 96-well plate of mutants of S-824. The activity of each mutant is plotted relative to the parental sequence. Not surprisingly, most mutations are deleterious, and most sequences are less active than the parental sequence (ratios , 1 in Fig. 4a). However, in this initial screen, four proteins ( 4% of the library) showed activity above the parent by 3 standard deviations (a value of 1.3). Next, we screened the S-836 library in four 96-well plates (Fig. 4b). In this case, 20% of the library showed activity at least 3 standard deviations above the parent. As a larger percentage of the S-836 library (compared with the S-824 library) showed improved activity, we chose this library for further studies. To examine a larger collection of sequences, we screened the library of S-836 mutants using the colony screen described above. Approximately 2 Â 10 3 colonies were screened. Those that turned green first were picked and reassayed in liquid media. The results of these assays are shown in Fig. 5. All of the picked clones showed greater activity than the parent. Next, we took the top hits from the 96-well plate screen (clones 1-2B11, 1-2G6) and from the colony screen (clones 1-C3, 1-C8) and mutagenized these to create a second- generation library of mutants. This library was produced in the same way as the first-generation library, using error-prone PCR and nucleotide analogs. The second-generation library of mutants was then screened in four 96-well plates, and the results are shown in Fig. 6. Two percent of the mutants showed activity significantly above the parent S-836. The top hits from the first-generation 96-well screen (1-2B11, 1-2G6), the first-generation colony screen (1-C3, 1-C8) and the second-generation 96-well screen (2-2H12, 2-3F1, 2-4C1) were then reassayed for peroxidase activity in cell lysates. This assay confirmed that all of the putative hits showed activity above the parental S-836 protein. These results are shown in Fig. 7 and the corresponding sequences are shown in Fig. 8. As the activity observed in a cell lysate depends on both the intrinsic activity of a protein and the level of its expression, we estimated expression levels by gel electrophoresis. As shown in Fig. 9, all sequences expressed at similar levels. To verify that the mutant sequences are indeed more active than the parental S-836 protein, we chose two proteins for purification and biochemical characterization. Sequences 1-2B11 and 2-2H12 were chosen because they displayed relatively high levels of activity in lysates (Fig. 7), and were straightforward to purify. Protein 1-2B11 had been isolated in the 96-well screen of the first-generation library, and protein 2-H12 was the top hit in the 96-well screen of the second-generation library. Sequence 2-H12 is a variant of 1-2B11 (Fig. 8). The kinetic profiles of the purified mutant proteins are compared with ...