|Home | About | Journals | Submit | Contact Us | Français|
Present Address: Department of Chemistry & Biochemistry, Georgia Institute of Technology, 901 Atlantic Drive, Atlanta, GA 30332-0400
Present Address: Division of Biology, University of California San Diego, La Jolla, CA 92093
Virginia Cornish, 1209 NWC Columbia University, 550 W 120th Street MC 3111, New York, NY 10027
Genetic recombination is central to the generation of molecular diversity and enhancement of evolutionary fitness in living systems. Methods such as DNA shuffling that recapitulate this diversity mechanism in vitro are powerful tools for engineering biomolecules with useful new functions by directed evolution. Synthetic biology now brings demand for analogous technologies that enable the controlled recombination of beneficial mutations in living cells. Thus, here we create a Heritable Recombination system centered around a library cassette plasmid that enables inducible mutagenesis via homologous recombination and subsequent combination of beneficial mutations through sexual reproduction in Saccharomyces cerevisiae. Using repair of nonsense codons in auxotrophic markers as a model, Heritable Recombination was optimized to give mutagenesis efficiencies of up to 6% and to allow successive repair of different markers through two cycles of sexual reproduction and recombination. Finally, Heritable Recombination was employed to change the substrate specificity of a biosynthetic enzyme, with beneficial mutations in three different active site loops crossed over three continuous rounds of mutation and selection to cover a total sequence diversity of 1013. Heritable Recombination, while at an early stage of development, breaks the transformation barrier to library size and can be immediately applied to combinatorial crossing of beneficial mutations for cell engineering, adding important features to the growing arsenal of next generation molecular biology tools for synthetic biology.
Progress in synthetic biology has spurred renewed interest in techniques for in vivo mutagenesis so that cellular components and pathways can be directly engineered in the intracellular environment in which they must perform.(1–3) To that end, recently improved E. coli mutator strains,(4) the co-opting of B cells for somatic hypermutation of exogenous genes,(5) recombination-mediated library mutagenesis by transformation,(6) a phage-based system for continuous mutagenesis,(7) and a Multiplex Automated Genome Engineering technology (MAGE) have all been reported.(8) None of these mutagenesis methods, however, also allows for crossing of beneficial mutations that arise in separate cells. We sought to design a system that would enable repeated rounds of mutagenesis and crossing of beneficial mutations in vivo with no intermediate in vitro manipulation of the DNA, built from classic genetic methods for recombination and mating and sporulation in the yeast S. cerevisiae (Figure 1).
Thus, we created Heritable Recombination, a system centered on library cassette plasmids that allow for inducible mutagenesis in a target gene via recombination and horizontal dissemination of beneficial mutations through sexual reproduction in yeast (Figure 2a). To provide specificity for mutagenesis of the target gene(s), the cassette is designed to utilize homologous recombination (HR) as the mechanism for mutagenesis. The high efficiency of HR in S. cerevisiae is well documented and widely exploited in both classical genetics and genetic engineering.(9–11) Additionally, the cassette is designed such that the mutagenesis can be initiated by an inducible endonuclease that introduces a double strand break (DSB) in the cassette plasmid.(12, 13) Experiments using endonucleases to create double strand breaks in vivo for the study of HR and for chromosome mutation suggested that induction of the DSB would be rapid and efficient, but also that intact cassette plasmids would be regenerated and passed on to daughter cells, thereby allowing for repeated cycles of evolution and straightforward tracking of beneficial mutations.(14) A significant difference between Heritable Recombination and previous HR technologies like Delitto Perfetto(11) is that the DSB is engineered in the mutagenic cassette rather than in the target gene. As elaborated here, there are likely numerous advantages to the flexibility provided by making the DSB in the cassette. Because the cassettes are encoded on plasmids, cell populations carrying two different cassette libraries can be subjected to selective pressure and the winners then crossed with one another by mating and sporulation to produce daughter cells with a beneficial mutation from one library in the target plasmid and a beneficial mutation from the other library in the cassette plasmid. Thus, Heritable Recombination breaks the transformation barrier to library size by enabling combinatorial exploration of extremely large libraries by crossing multiple different cassettes. Alternatively, or in addition, it can enable virtual searches of extremely large sequence space along different pathways by varying the content and ordering of cassette library introduction (Figure 1).
The first step in building the Heritable Recombination system was to effect endonuclease-inducible recombination so that mutagenesis could be initiated by a double-stranded break (DSB) to the cassette plasmid rather than by transformation. While designed to be modular and to enable library mutagenesis, the system was validated based on repair of a nonsense codon in the trp1-R78TAA gene. The TRP1 gene is one of the most commonly used auxotrophic markers in yeast and was the focus of our initial studies on oligonucleotide mutagenesis.(6) For the catalytic endonuclease, we chose the meganuclease I-SceI (SceI), an established tool for genetic engineering in yeast(15, 16); SceI was placed under control of the tightly repressed, yet strongly inducible, GAL1 promoter. Finally, the mutagenic cassette plasmid was designed with 30 bp of homologous sequence on either side of the wild-type Arg78 codon and flanked on both sides by the 18-bp recognition sequence for SceI (Figure 2b). Previous studies have shown that 30 bp is the minimum amount of homology needed to obtain high efficiency recombination in yeast(17); yet at the same time, this length of DNA corresponds to 10 amino acids and thus should allow mutagenesis of multiple regions within a single gene. As depicted in Figure 2a, in this initial inception, the target gene trp1-R78TAA, the galactose-inducible SceI endonuclease, and the mutagenic Arg78 cassette are each encoded on their own plasmid.
Satisfyingly, we were able to readily optimize this initial design to obtain the high recombination efficiencies essential for library mutagenesis. Our definition of recombination efficiency is the ratio of the number of colonies that grow in the absence, versus the presence, of tryptophan, indicating successful repair of the trp1 nonsense codon by the mutagenic Arg78 cassette plasmid. Our initial experiments yielded growth in only 0.12% of cells, but protocol optimization led to an average maximum efficiency of 6 ± 1% after 10 h of induction with galactose (Figure 2c). Numerous control experiments confirmed the requirement for all three components of our system: no mutagenesis above background levels was observed in strains without the endonuclease plasmid, without the mutagenic cassette plasmid, or with a cassette plasmid that lacked endonuclease recognition sites. Sequencing of individual clones confirmed that the expected mutations had in fact occurred without any other detected changes in the TRP1 gene in 24/25 of the samples analyzed (Figure 2d).
Furthermore, we were able to confirm that intact cassette plasmids were successfully passed on to daughter cells following induction. Plasmids were harvested from cells that displayed the Trp+ phenotype, and PCR amplification of the cassette region produced products of the expected size (Figure 2e). Sequencing of the PCR products showed not only that the cassettes retained all of their essential parts, but also that they were not mutagenized, even after undergoing prolonged DSB induction and successful mutagenesis of the target gene. Together, these results establish that the high-copy cassette plasmid is competent for targeted mutagenesis and further dissemination to daughter cells, providing a simple way to track and cross the accumulated mutations in an evolving population of cells.(18)
Addition of the sexual reproduction step was executed in a model system in which stop codons in two different biosynthetic genes were successively repaired with no intermediate in vitro manipulation of the DNA. Specifically, we first repaired a nonsense mutation in the hisA gene from Thermotoga maritima to complement a his6− genotype and then repeated the process with the trp1 gene to complement trp1−, exploiting mating and sporulation in S. cerevisiae as the mechanism of sexual reproduction (Figure 3a).(19) Two cassette plasmids were used to achieve this sequential mutagenesis: the first encoded a wild-type Val50 residue for repair of the corresponding nonsense mutation in an engineered hisA-V50TAA gene; the second cassette encoded Arg78, as above, to repair trp1-R78TAA. Importantly, while the cassette plasmids were initially introduced into the haploid a-type donor cells by transformation, in this experiment, mutation of the hisA gene was executed simply by mating haploid a-type cassette cells with haploid α-type target cells containing the target gene and inducible endonuclease. Sporulation followed by lyticase digestion of any unsporulated diploids was used to regenerate haploid cells between rounds of selection for mating with additional cassette cells (see Supporting Information Section S.3). Cells were selected for histidine prototrophy after the first round of mutagenesis and for tryptophan prototrophy after the second round. To confirm the link between the observed phenotypes and the desired recombination events, individual plasmids were isolated and retransformed into the background strain, yielding cells that exhibited the expected growth profile (Figure 3b). Restriction mapping and sequencing of both pooled DNA and DNA from individual colonies showed that both encoded mutations had been successfully incorporated into the hisA and TRP1 genes with no other detected changes to their sequences (Figure 3c).
This two-stage experiment was also executed as a mock selection to confirm the capability of the system to effect mutagenesis from a library. For these experiments, cassette strains carrying mutagenic plasmids with the appropriate repair codon were mixed with an excess of empty cassette cells in which the plasmid contained no homology to the target gene and a single SceI site. The resulting mock libraries were mated to the target strain at the start of each round of Heritable Recombination, and cells carrying both repaired genes were successfully isolated from a 104 excess of empty cassette plasmids in the first round and a 102 excess in the second round. Thus, while validated here with just two cassettes, Heritable Recombination allows multiple rounds of library evolution to be carried out simply through repeated cycles of sporulation followed by mating with prepackaged cassette cells under conditions of selective pressure without any intermediate steps of DNA isolation and manipulation.
Finally, we challenged the utility of crossing beneficial mutations in the course of selection with Heritable Recombination by rapidly evolving the hisA biosynthetic enzyme to accept a new substrate. Wild-type hisA codes for N′-[(5′-phosphoribosyl)formimino]-5-aminoimidazole-4-carboxamide ribonucleotide (ProFAR) isomerase, a (βα)8-barrel enzyme that catalyzes an Amadori rearrangement as the 4th step in the de novo biosynthesis of histidine. Previous work by the Sterner group has shown that directed evolution followed by rational combination of beneficial mutations can be used to convert ProFAR into highly active variants of the evolutionarily related tryptophan biosynthetic enzyme phosphoribosylanthranilate isomerase (PRAI).(20, 21) Thus, we reasoned that Heritable Recombination should allow us to evolve hisA variants with combinations of beneficial mutations in a rapid and straightforward manner.
Rewardingly, three rounds of mutagenesis and selection via Heritable Recombination yielded hisA variants that could complement a trp1− auxotrophy in yeast in under a month (Figure 4a). Three separate libraries of mutagenic cassettes were designed targeting distinct loops at the C-terminal end of the hisA (βα)8-barrel (Figure 4b). Loops at this face of the protein contribute residues for substrate recognition and catalysis in nearly all known examples of (βα)8-barrel enzymes.(22) Each cassette randomized three residues using degenerate NNS codons (N = A, T, C, G; S = G, C) to provide a complexity of 3.4 × 104 variants per cassette—or 3.5 × 1013 total variants with crossing of all three cassettes.
In the first round, cells containing a target plasmid with the hisA gene were separately mated to two populations of cells carrying libraries 1 and 2. Recombination-mediated mutagenesis was followed by selection against the his6− auxotrophy to generate a pool of variants that retained hisA activity. This “neutral drift” step allowed the identification of mutations that still led to folded and functional proteins, providing a privileged pool of starting points for discovery of hisA mutants that could complement trp1− auxotrophy in yeast.(23, 24)
Surviving cells from these initial selections were induced to sporulate and then mated to each other to combine the cassette plasmids from these first two libraries. A second round of mutagenesis and neutral drift selection was then carried out to merge the winning mutations from the first round before mating the surviving cells with a population carrying library 3.
Following a third round of mutagenesis, the cells were placed in media lacking tryptophan to select for enzymes with the desired new substrate specificity. Mutated target plasmids were isolated after the final round of selection, retransformed into yeast to confirm the Trp+ phenotype, and sequenced. Significantly, by using an intermediate selection to hone the libraries so that they contained only functional mutations, we were able to combine three libraries of 3.4 × 104 variants to virtually explore a sequence space equal to the product of the three libraries, or approximately 3.5 × 1013 sequences—several orders of magnitude greater than the current transformation limit to library size in yeast. In theory the process of mutation followed by sexual reproduction could be continued with more cassette libraries to explore vast amounts of sequence diversity.
This efficient search of sequence space readily produced a variety of hisA mutants with different cassette combinations that could fully complement a trp1− genotype in yeast (Figure 4c). Sequencing of these mutants showed that Heritable Recombination was working properly: several cassettes were re-used in multiple different solutions, indicating combinatorial sharing of beneficial mutations among members of the population. Despite cassettes encoding full saturation at each targeted locus, several residues were unchanged from the wild type in all variants that were sequenced (see Supporting Information Section S.2). Several of the sequenced genes contained additional single-base mutations that were not encoded in the initial degenerate libraries but rather were located in the homology region for library 2; we are currently engaged in studies to understand the origin of these mutations as well as the overall mechanism of mutagenesis in this system.
While the design of our libraries benefited from insight into this enzyme class from published work(20, 21), the ability to cross beneficial mutations in vivo using sexual reproduction allowed us to readily identify several different cassette combinations that produced enzymes that strongly complemented the trp1− genotype. The ease with which we isolated these combinations stands in contrast to previous efforts that required rational engineering to combine beneficial mutations arising from error-prone PCR to achieve highly active enzymes.(21) We discovered seven distinct sequence solutions, and most of the mutations we found were at sites other than those previously reported. Interestingly, all of our results contained at least one new basic amino acid, supporting the hypothesis that electrostatic effects play a dominant role in recognizing the negatively charged substrate phosphoribosylanthranilate.(20, 25) The preliminary neutral drift steps show that nondeleterious mutations can potentially provide increased evolvability under changing selective pressure. We are currently exploring ways to test the evolutionary significance of this concept using Heritable Recombination.
The success of this directed evolution experiment showcases all three parts of our technology working in concert to efficiently search sequence space and distinguishes it from previously reported in vivo mutagenesis techniques. Homology-directed library cassettes target the mutations to the regions of interest in a vast excess of chromosomal DNA, thus avoiding the undesired genomic instability often observed with mutator strains that prevent continuous rounds of mutation and selection.(26) The recently reported PACE technology uses continuous phage infection of E. coli to surmount the genomic instability problem, but does not enable targeted mutations or addition or deletion of DNA and does not address recombination of beneficial mutations.(7) By using an endonuclease-induced double-strand break in a heritable cassette plasmid inside the cell to initiate mutagenesis, our method is distinct from previous efforts—both long-standing work in yeast (27–30) and the more recent MAGE technology in E. coli (8) and work from our laboratory in yeast (6)—to generate molecular diversity that rely on transformation of linear DNA fragments to achieve recombination. A technique called “gene gorging” uses DSBs generated by SceI to initiate lamda Red-mediated recombination in E. coli, but this process has not been explored for library creation and cannot be iterated over multiple rounds.(31) Several methods using in vivo recombination in E. coli to generate antibody libraries have been reported (32, 33), but these methods are often limited to antibody-derived genes and, unlike Heritable Recombination, have no mechanism for repeated diversification through sexual reproduction. DNA shuffling can mimic sexual reproduction, but it relies on PCR and therefore lacks the other benefits of our entirely in vivo process.(34) Thus, Heritable Recombination uniquely breaks the transformation barrier to library size and, by analogy to computational algorithms like dead-end elimination, enables virtual searches of extremely large libraries.(35) Most important, our strategy allows heritable cassette plasmids to be exchanged among cells via mating and sporulation and so provides a simple and efficient way to track and combine beneficial mutations. In contrast to CAGE, which relies on bacterial conjugation to combine large contiguous regions of E. coli chromosomes, mutations can be crossed organically with Heritable Recombination over many rounds of the mutation and sexual reproduction cycle.(36)
Heritable Recombination is immediately compatible with any screen or selection available in yeast. While we have demonstrated it here using classic yeast auxotrophic growth selections, we anticipate that it will be equally useful under a variety of existing assays, including analytical screens and affinity-based selections. Currently, we are applying Heritable Recombination in conjunction with yeast cell surface display for the evolution of antibodies(37) and enzymes.(38) Our long-term goal is to couple Heritable Recombination with an expanded Chemical Complementation selection to enable the routine evolution of new chemistry directly in living cells.(39)
We foresee this technology becoming an efficient production platform for the directed evolution of biomolecules. For example, a privileged scaffold, such as the (βα)8-barrel or an immunoglobulin fold, could be used as a common starting point, and libraries specific to the scaffold would be stored in yeast as frozen glycerol stocks. While demonstrated here with diversification of active-site loops in a single gene, we also plan to apply Heritable Recombination to several genes at once, ultimately targeting multiple loci in the yeast genome simultaneously. Regardless of the desired target(s), the appropriate libraries could be combined and searched through mating and in vivo mutagenesis to identify useful sequences in a manner of days. Further maturation of the activity of the target(s) could be executed simply by additional rounds of mating with pre-existing cassette cell libraries.
In a broader context, the methods presented here represent an advance in synthetic biology. The manipulation of nucleic acid sequences in vivo is a rapidly developing field with implications ranging from conditional gene expression to completely synthetic genomes. We foresee techniques based on homologous recombination—in yeast, bacteria, and mammalian cells—gaining even greater relevance as the mechanism of these processes becomes further elucidated.(40–42) We have already begun related projects using Heritable Recombination for the evolution of enzymes and binding proteins, metabolic engineering, and pathway construction.(43)
Standard molecular biology protocols were used for plasmid and strain construction. For specific details regarding strains and plasmids, individual experiments, and mock selections, please refer to the Supporting Information.
Overnight cultures of mutagenesis strains were grown in synthetic complete media with 2% (w/v) glucose (SC-Glu) lacking the necessary nutrients for plasmid maintenance. Cells were centrifuged, the supernatant was removed, and the cells were resuspended in preinduction media (SC with 2% lactate). Preinduction cultures were shaken at 30 °C for 3 hours then split evenly into two tubes and centrifuged. Supernatants were removed and one cell pellet was resuspended in SC with 2% galactose while the other was resuspended in SC-Glu. The cultures were shaken at 30 °C for the desired induction time, typically 24 hours, and then serial dilutions were plated on selective and nonselective SC-Glu plates. Colonies were counted after incubation at 30 °C for 48 hours.
Diploid cells were placed in 10 mL presporulation media (YP + 10% glucose) at an initial cell density of OD600=0.1. The cultures were shaken at 30 °C until they reached a density of 0.5–0.7, typically 7–9 hours. Cells were then collected by centrifugation and resuspended in an equal volume of sporulation media (0.5% potassium acetate) and shaken at 30 °C for 48 hours. The spores were harvested by centrifugation and resuspended in 1 mL water. Zymolyase was added (10 units), and the spores were placed in a 37 °C bath for 2 hours and then sonicated for 35 minutes. Cells were germinated by overnight growth in SC media. Freshly germinated cells were mixed in a 1:3 ratio with cells containing the new cassette (appx. 107 total cells), and this cell mixture was plated on YPD. The mating plate was incubated at 30 °C for 8 hours, and then the cells were harvested with water and diluted into SC-Glu media to select for diploids carrying all plasmids.
This work was supported in part by funds from the NIH (R01GM62867, R01GM096064) and the NSF (CHE0957569). DWR is supported by a postdoctoral fellowship from the National Institute of General Medical Sciences (F32GM089031). We would like to thank Prof. Aaron Mitchell, Dr. Robert Reid, Dr. Jon Bronson, Dr. Laura Wingler, and Dr. Nili Ostrov for helpful discussions, and Dr. Minchen Chien for help with sequencing. We would also like to acknowledge Prof. Janet Lindsley (U. Utah) for her gift of the I-SceI gene.
AUTHOR CONTRIBUTIONSDWR, PPY, VM, and VWC conceived and designed the research. DWR, PPY, and VM conducted the experiments. DWR and VWC analyzed the data and wrote the paper.
Supporting Information Available: This material is available free of charge via the Internet.
Dante W. Romanini, Department of Chemistry, Columbia University, New York, NY 10027.
Pamela Peralta-Yahya, Department of Chemistry, Columbia University, New York, NY 10027.
Vanessa Mondol, Department of Chemistry, Columbia University, New York, NY 10027.
Virginia W. Cornish, Department of Chemistry, Columbia University, New York, NY 10027.