Dan Tawfik's Lessons for Protein Engineers about Enzymes Adapting to New Substrates.

Natural evolution has been creating new complex systems for billions of years. The process is spontaneous and requires neither intelligence nor moral purpose but is nevertheless difficult to understand. The late Dan Tawfik spent years studying enzymes as they adapted to recognize new substrates. Much of his work focused on gaining fundamental insights, so the practical utility of his experiments may not be obvious even to accomplished protein engineers. Here we focus on two questions fundamental to any directed evolution experiment. Which proteins are the best starting points for such experiments? Which trait(s) of the chosen parental protein should be evolved to achieve the desired outcome? We summarize Tawfik's contributions to our understanding of these problems, to honor his memory and encourage those unfamiliar with his ideas to read his publications.

Substrate ambiguous enzymes within the Escherichia coli proteome offer different evolutionary solutions to the same problem.

Many enzymes exhibit some catalytic promiscuity or substrate ambiguity. These weak activities do not affect the fitness of the organism under ordinary circumstances, but can serve as potential evolutionary precursors of new catalytic functions. We wondered whether different proteins with the same substrate ambiguous activity evolve differently under identical selection conditions. Patrick et al. (Patrick WM, Quandt EM, Swartzlander DB, Matsumura I. 2007. Multicopy suppression underpins metabolic evolvability. Mol Biol Evol. 24:2716-2722.) previously showed that three multicopy suppressors, gph, hisB, and ytjC, rescue ΔserB Escherichia coli cells from starvation on minimal media. We directed the evolution of variants of Gph, histidinol phosphatase (HisB), and YtjC that complemented ΔserB more efficiently, and characterized the effects of the amino acid changes, alone and in combination, upon the evolved phosphoserine phosphatase (PSP) activity. Gph and HisB are members of the HAD superfamily of hydrolases, but they adapted through different, kinetically distinguishable, biochemical mechanisms. All of the selected mutations, except N102T in YtjC, proved to be beneficial in isolation. They exhibited a pattern of antagonistic epistasis, as their effects in combination upon the kinetic parameters of the three proteins in reactions with phosphoserine were nonmultiplicative. The N102T mutation exhibited sign epistasis, as it was deleterious in isolation but beneficial in the context of other mutations. We also showed that the D57N mutation in the chromosomal copy of hisB is sufficient to suppress the ΔserB deletion. These results in combination show that proteomes can offer multiple mechanistic solutions to a molecular recognition problem.

Transformable facultative thermophile Geobacillus stearothermophilus NUB3621 as a host strain for metabolic engineering.

Metabolic engineers develop inexpensive enantioselective syntheses of high-value compounds, but their designs are sometimes confounded by the misfolding of heterologously expressed proteins. Geobacillus stearothermophilus NUB3621 is a readily transformable facultative thermophile. It could be used to express and properly fold proteins derived from its many mesophilic or thermophilic Bacillaceae relatives or to direct the evolution of thermophilic variants of mesophilic proteins. Moreover, its capacity for high-temperature growth should accelerate chemical transformation rates in accordance with the Arrhenius equation and reduce the risks of microbial contamination. Its tendency to sporulate in response to nutrient depletion lowers the costs of storage and transportation. Here, we present a draft genome sequence of G. stearothermophilus NUB3621 and describe inducible and constitutive expression plasmids that function in this organism. These tools will help us and others to exploit the natural advantages of this system for metabolic engineering applications.

Expression vectors for Acinetobacter baylyi ADP1.

Acinetobacter baylyi ADP1 is naturally competent and proficient at homologous recombination, so it can be transformed without restriction digests or ligation reactions. Expression vectors for this system, however, are not yet widely available. Here we describe the construction and characterization of inducible expression vectors that replicate as plasmids in A. baylyi or integrate into a nonessential part of its chromosome. These tools will facilitate the engineering of genes and genomes in this promising model organism.

Golden Gate assembly of BioBrick-compliant parts using Type II restriction endonucleases.

Aims: New methods of DNA recombination that capture the principal advantages of the BioBrick standard (ease of design) and Golden Gate assembly (decreased labor) are demonstrated here. Methods & materials: Both methods employ DNA methyltransferase expression vectors, available from Addgene, that protect selected sites on different plasmids from particular Type II restriction endonucleases. No other reagents are required. Results: The 4R/2M discontinuous DNA assembly is more efficient (produces more desired recombinant plasmids) and as specific (produces few undesired recombination products) as conventional subcloning. The 5RM continuous DNA assembly is approximately as efficient and specific as conventional Golden Gate assembly, even though in vivo methylation of one plasmid is incomplete. Conclusion: Both methylase-assisted methods streamline BioBrick assembly workflows without complicating the design of synthetic parts.

Methylase-assisted subcloning for high throughput BioBrick assembly.

The BioBrick standard makes possible iterated pairwise assembly of cloned parts without any depletion of unique restriction sites. Every part that conforms to the standard is compatible with every other part, thereby fostering a worldwide user community. The assembly methods, however, are labor intensive or inefficient compared to some newer ones so the standard may be falling out of favor. An easier way to assemble BioBricks is described herein. Plasmids encoding BioBrick parts are purified from Escherichia coli cells that express a foreign site-specific DNA methyltransferase, so that each is subsequently protected in vitro from the activity of a particular restriction endonuclease. Each plasmid is double-digested and all resulting restriction fragments are ligated together without gel purification. The ligation products are subsequently double-digested with another pair of restriction endonucleases so only the desired insert-recipient vector construct retains the capacity to transform E. coli. This 4R/2M BioBrick assembly protocol is more efficient and accurate than established workflows including 3A assembly. It is also much easier than gel purification to miniaturize, automate and perform more assembly reactions in parallel. As such, it should streamline DNA assembly for the existing community of BioBrick users, and possibly encourage others to join.

Statistical noise from recombinant plasmids can be abated via complementation of a ribosomal protein gene deletion.

The phenotypes conferred by recombinant plasmids upon host cells often exhibit variability between replicate populations. This statistical noise is mostly a consequence of adaptive evolution in response to fitness burdens imposed by the plasmids themselves. We developed a novel strategy, 'ribosome pegging', to exclude common unwanted mutations that benefit host cells at the expense of heterologous gene expression. Plasmids that constitutively co-expressed the fluorescent reporter tagRFP and ribosomal protein L23 (rplW) were used to transform Escherichia coli cells that lacked the essential chromosomal rplW gene. Cells within the population that expressed too little L23, or too much, were evidently inviable. Ribosome pegging obviates the need for antibiotics, thus facilitating the deployment of recombinant bacteria in uncontrolled environments. We show that ribosome-pegged E. coli carrying a plasmid that constitutively expresses L23 and an artificially evolved enzyme protects fruit flies from otherwise toxic doses of the insecticide malathion.

Antibody-induced oligomerization and activation of an engineered reporter enzyme.

Our objective is to produce a protein biosensor (or molecular switch) that is specifically activated in solution by a monoclonal antibody. Many effector-dependent enzymes have evolved in nature, but the introduction of a novel regulatory mechanism into a normally unregulated enzyme poses a difficult design problem. We used site-saturation mutagenesis and screening to generate effector-activated variants of the reporter enzyme beta-glucuronidase (GUS). The specific activity of the purified epitope-tagged GUS variant was increased by up to approximately 500-fold by the addition of an equimolar concentration of a monoclonal antibody. This molecular switch is modular in design, so it can easily be re-engineered for the detection of other peptide-specific antibodies. Such antibody-activated reporters could someday enable point-of-care serological assays for the rapid detection of infectious diseases.

Artificially evolved Synechococcus PCC6301 Rubisco variants exhibit improvements in folding and catalytic efficiency.

The photosynthetic CO2-fixing enzyme, Rubisco (ribulose-1,5-bisphosphate carboxylase/oxygenase), is responsible for most of the world's biomass, but is a slow non-specific catalyst. We seek to identify and overcome the chemical and biological constraints that limit the evolutionary potential of Rubisco in Nature. Recently, the horizontal transfer of Calvin cycle genes (rbcL, rbcS and prkA) from cyanobacteria (Synechococcus PCC6301) to gamma-proteobacteria (Escherichia coli) was emulated in the laboratory. Three unique Rubisco variants containing single (M259T) and double (M259T/A8S, M259T/F342S) amino acid substitutions in the L (large) subunit were identified after three rounds of random mutagenesis and selection in E. coli. Here we show that the M259T mutation did not increase steady-state levels of rbcL mRNA or L protein. It instead improved the yield of properly folded L subunit in E. coli 4-9-fold by decreasing its natural propensity to misfold in vivo and/or by enhancing its interaction with the GroES-GroEL chaperonins. The addition of osmolites to the growth media enhanced productive folding of the M259T L subunit relative to the wild-type L subunit, while overexpression of the trigger factor and DnaK/DnaJ/GrpE chaperones impeded Rubisco assembly. The evolved enzymes showed improvement in their kinetic properties with the M259T variant showing a 12% increase in carboxylation turnover rate (k(c)cat), a 15% improvement in its K(M) for CO2 and no change in its K(M) for ribulose-1,5-bisphosphate or its CO2/O2 selectivity. The results of the present study show that the directed evolution of the Synechococcus Rubisco in E. coli can elicit improvements in folding and catalytic efficiency.

A GAL80 Collection To Inhibit GAL4 Transgenes in Drosophila Olfactory Sensory Neurons.

Fruit flies recognize hundreds of ecologically relevant odors and respond appropriately to them. The complexity, redundancy and interconnectedness of the olfactory machinery complicate efforts to pinpoint the functional contributions of any component neuron or receptor to behavior. Some contributions can only be elucidated in flies that carry multiple mutations and transgenes, but the production of such flies is currently labor-intensive and time-consuming. Here, we describe a set of transgenic flies that express the Saccharomyces cerevisiae GAL80 in specific olfactory sensory neurons (OrX-GAL80s). The GAL80s effectively and specifically subtract the activities of GAL4-driven transgenes that impart anatomical and physiological phenotypes. OrX-GAL80s can allow researchers to efficiently activate only one or a few types of functional neurons in an otherwise nonfunctional olfactory background. Such experiments will improve our understanding of the mechanistic connections between odorant inputs and behavioral outputs at the resolution of only a few functional neurons.

Semi-automated Tip Snip cloning of restriction fragments into and out of plasmid polylinkers.

Synthetic biologists rely on semi-synthetic recombinant plasmids, but DNA synthesis is constrained by practical limits on length, accuracy, and sequence composition. Cloned DNA parts can be assembled into longer constructs via subcloning, but conventional methods are labor-intensive. One-pot recombination reactions are more convenient but harder to troubleshoot, and those that depend on PCR to create fragments with compatible ends necessitate re-sequencing. The Tip Snip protocol described here enables the subcloning of an insert from one plasmid polylinker into another without PCR or gel purification steps. Cohesive ends of unwanted restriction fragments are snipped off by additional restriction endonucleases. The resulting short fragments (snippets) are eliminated by hybridization to complementary oligonucleotides (anti-snippets) and subsequent size-selection spin-column chromatography. Unwanted linear donor vectors are ligated to double-stranded oligonucleotides (unlinkers) so that only the desired insert and recipient plasmid form circular DNA capable of transforming bacteria. This new method is compatible with high-throughput processing and automated liquid handling, and because no specialized vectors, reagents, selection schemes, or analytical techniques are required, the barriers to adoption are low.

Natural history as a predictor of protein evolvability.

Natural selection generally produces specific and efficient enzymes. In contrast, directed evolution experiments usually produce enzyme variants with broadened substrate specificity or enhanced catalytic promiscuity. Some proteins may be more evolvable than others, but few workers consider this problem when choosing starting points for laboratory evolution. Here, we review the variables associated with enzyme evolvability, namely promiscuity and mutational robustness. We present a qualitative model of adaptive evolution and recommend that protein engineers exploit their knowledge of natural history to identify evolvable wild-type proteins. Three examples of 'generalist' proteins that evolved in the laboratory into 'specialists' are described to illustrate the practical utility of this point.

Directed evolution of RuBisCO hypermorphs through genetic selection in engineered E.coli.

The Calvin Cycle is the primary conduit for the fixation of carbon dioxide into the biosphere; ribulose 1,5-bisphosphate carboxylase/oxygenase (RuBisCO) catalyzes the rate-limiting fixation step. Our goal is to direct the evolution of RuBisCO variants with improved kinetic and biophysical properties. The Calvin Cycle was partially reconstructed in Escherichia coli; the engineered strain requires the Synechococcus PCC6301 RuBisCO for growth in minimal media supplemented with a pentose. We randomly mutated the gene encoding the large subunit of RuBisCO (rbcL), co-expressed the resulting library with the small subunit (rbcS) and the Synechococcus PCC7492 phosphoribulokinase (prkA), and selected hypermorphic variants. The RuBisCO variants that evolved during three rounds of random mutagenesis and selection were over-expressed, and exhibited 5-fold improvement in specific activity relative to the wild-type enzyme. These results demonstrate a new strategy for the artificial selection of RuBisCO and other non-native metabolic enzymes.

Site-saturation mutagenesis is more efficient than DNA shuffling for the directed evolution of beta-fucosidase from beta-galactosidase.

Protein engineers use a variety of mutagenic strategies to adapt enzymes to novel substrates. Directed evolution techniques (random mutagenesis and high-throughput screening) offer a systematic approach to the management of protein complexity. This sub-discipline was galvanized by the invention of DNA shuffling, a procedure that randomly recombines point mutations in vitro. In one influential study, Escherichia coli beta-galactosidase (BGAL) variants with enhanced beta-fucosidase activity (tenfold increase in k(cat)/K(M) in reactions with the novel para-nitrophenyl-beta-d-fucopyranoside substrate; 39-fold decrease in reactivity with the "native"para-nitrophenyl-beta-d-galactopyranoside substrate) were evolved in seven rounds of DNA shuffling and screening. Here, we show that a single round of site-saturation mutagenesis and screening enabled the identification of beta-fucosidases that are significantly more active (180-fold increase in k(cat)/K(M) in reactions with the novel substrate) and specific (700,000-fold inversion of specificity) than the best variants in the previous study. Site-saturation mutagenesis thus proved faster, less resource-intensive and more effective than DNA shuffling for this particular evolutionary pathway.

HIV protease-activated molecular switches based on beta-glucuronidase and alkaline phosphatase.

Our long-term goal is to direct the evolution of novel protease variants. To this end we have engineered a new type of protease-activated reporter enzyme. Many protease-activated enzymes evolved in nature, but the introduction of novel regulatory mechanisms into normally unregulated enzymes poses a difficult design challenge. Random Elongation Mutagenesis [1] was used to fuse the p6 peptide, which is recognized and cleaved by HIV protease, and twelve random sequence amino acids to the C-termini of beta-glucuronidase (GUS) and alkaline phosphatase (AP). The resulting GUS-p6-(NNN)12 and AP-p6-(NNN)12 libraries were expressed in E. coli and screened for clones that were inactivated by the C-terminal extension (tail). The inactivated clones were co-expressed with HIV protease, and those that were re-activated were isolated. The AP and GUS activities of the most responsive clones were each >3.5-fold higher when co-expressed with HIV protease, and this activation is correlated with in vivo proteolysis. It should be possible to generalize this strategy to different reporter enzymes, different target proteases, and perhaps to other types of protein-modifying enzymes.

Pooling for improved screening of combinatorial libraries for directed evolution.

Following diversity generation in combinatorial protein engineering, a significant amount of effort is expended in screening the library for improved variants. Pooling, or combining multiple cells into the same assay well when screening, is a means to increase throughput and screen a larger portion of the library with less time and effort. We have developed and validated a Monte Carlo simulation model of pooling and used it to screen a library of beta-galactosidase mutants randomized in the active site to increase their activity toward fucosides. Here, we show that our model can successfully predict the number of highly improved mutants obtained via pooling and that pooling does increase the number of good mutants obtained. In unpooled conditions, we found a total of three mutants with higher activity toward p-nitrophenyl-beta-D-fucoside than that of the wild-type beta-galactosidase, whereas when pooling 10 cells per well we found a total of approximately 10 improved mutants. In addition, the number of "supermutants", those with the highest activity increase, was also higher when pooling was used. Pooling is a useful tool for increasing the efficiency of screening combinatorial protein engineering libraries.

A comparison of directed evolution approaches using the beta-glucuronidase model system.

Protein engineers can alter the properties of enzymes by directing their evolution in vitro. Many methods to generate molecular diversity and to identify improved clones have been developed, but experimental evolution remains as much an art as a science. We previously used DNA shuffling (sexual recombination) and a histochemical screen to direct the evolution of Escherichia coli beta-glucuronidase (GUS) variants with improved beta-galactosidase (BGAL) activity. Here, we employ the same model evolutionary system to test the efficiencies of several other techniques: recursive random mutagenesis (asexual), combinatorial cassette mutagenesis (high-frequency recombination) and a versatile high-throughput microplate screen. GUS variants with altered specificity evolved in each trial, but different combinations of mutagenesis and screening techniques effected the fixation of different beneficial mutations. The new microplate screen identified a broader set of mutations than the previously employed X-gal colony screen. Recursive random mutagenesis produced essentially asexual populations, within which beneficial mutations drove each other into extinction (clonal interference); DNA shuffling and combinatorial cassette mutagenesis led instead to the accumulation of beneficial mutations within a single allele. These results explain why recombinational approaches generally increase the efficiency of laboratory evolution.

Simulation modeling of pooling for combinatorial protein engineering.

Pooling in directed-evolution experiments will greatly increase the throughput of screening systems, but important parameters such as the number of good mutants created and the activity level increase of the good mutants will depend highly on the protein being engineered. The authors developed and validated a Monte Carlo simulation model of pooling that allows the testing of various scenarios in silico before starting experimentation. Using a simplified test system of 2 enzymes, betagalactosidase (supermutant, or greatly improved enzyme) and beta-glucuronidase (dud, or enzyme with ancestral level of activity), the model accurately predicted the number of supermutants detected in experiments within a factor of 2. Additional simulations using more complex activity distributions show the versatility of the model. Pooling is most suited to cases such as the directed evolution of new function in a protein, where the background level of activity is minimized, making it easier to detect small increases in activity level. Pooling is most successful when a sensitive assay is employed. Using the model will increase the throughput of screening procedures for directed-evolution experiments and thus lead to speedier engineering of proteins.

Whole plasmid mutagenic PCR for directed protein evolution.

Protein function can be engineered through iterated cycles of random mutagenesis and screening (directed evolution). Optimization of protein expression is essential for the development of sensitive and precise high throughput assays. Here we optimize the performance of a plasmid-borne Escherichia coli lacZ gene in two rounds of directed evolution. First, its promoter was "randomized" by whole plasmid polymerase chain reaction (PCR) and intra-molecular self-ligation. A genetically stable constitutive expression vector was isolated in an in vivo genetic selection. Second, the entire plasmid was randomly mutated in a slightly mutagenic long polymerase chain reaction. The PCR products were digested with a restriction enzyme, self-ligated by T4 DNA ligase and transformed into E. coli. The resulting library of beta-galactosidase (beta-gal) mutants consisted mostly ( approximately 80%) of hypomorphs, suggesting that the mutation rate was appropriate for directed evolution applications. We isolated and characterized 14 variants with increased activity in reactions with 5-bromo-4-chloro-3-indolyl-beta-d-galactopyranoside (X-gal). The purified protein derived from one clone exhibited a 100-fold improvement in k(cat) over its parent in reactions with para-nitrophenyl-beta-d-galactopyranoside (pNP-gal). This latter result clearly demonstrates the utility of whole plasmid mutagenic PCR for directed protein evolution.