Oligonucleotide assembly in yeast, as described here, could be very useful in the synthesis of large DNA fragments, including genes and genomes. This study, together with our previous studies (17
), demonstrates that entire genomes can be synthesized within a yeast cell beginning from a series of oligonucleotides that overlap by as few as 20 nucleotides. And, the genomes can be built following just three rounds of yeast transformation. The limits of DNA uptake and recombination have not been fully explored. It may already be possible that yeast has the capacity to assemble tens of kilobases of DNA sequence in one transformation event from 100 or more ssDNA oligonucleotides. Using longer oligonucleotides () or annealing complementary adjacent oligonucleotides prior to in vivo
assembly may be viable approaches for producing longer DNA fragments in a single step, especially if the number of pieces taken up by a yeast cell becomes a limiting factor. Indeed, it is very likely that many of the overlapping oligonucleotides become annealed outside the yeast cell during the transformation procedure. This is especially likely once the molecular crowding agent, polyethylene glycol, is added.
Although oligonucleotide synthesis is drastically better than it was 40-years-ago, this process continues to produce a fraction of unintended DNA sequence (), which will be more prevalent in longer synthetic DNA fragments. The synthesis error-rate for the assembly of oligonucleotides 1–38 u is 0.181% and 0.249% for the assembly of 1–28g (; based on data from Supplementary Tables S6 and S7
). This is comparable to previously reported data obtained with in vitro
). The use of high-fidelity oligonucleotides (Ultramers, IDT) reduced the error rate to 0.054% (; based on data from Supplementary Table S8
), which is only slightly higher than what can be achieved with error-correction methods (26
). Since the error-rate of oligonucleotide synthesis increases with size, the use of smaller high-fidelity oligonucleotides, such as 60-mers, would even further increase the percentage of correct clones obtained. The accuracy of DNA assembled by in vitro
approaches would also be improved by the use of high-fidelity oligonucleotides.
Three oligononucleotide assembly schemes were designed and described here. When choosing a scheme, oligonucleotide expenses should be considered along with DNA sequencing rates. The use of high fidelity Ultramer oligonucleotides will decrease sequencing costs. However, these oligonucleotides are significantly more expensive than their lower fidelity counterparts. DNA sequencing costs can be reduced even further by sequencing only the clones that contain a full-length synthetic fragment (as determined by PCR or restriction digestion). Oligonucleotide costs can be further reduced by making use of shorter overlaps. However, longer overlaps may be preferred in instances where unique overlaps cannot be obtained by 20 bp or less. It is expected that oligonucleotide chemistry will continue to improve and soon reliably produce error-free stretches of ssDNA at a more reasonable cost. Once these advances are in place, the one-step synthesis method described here could be used to rapidly construct long DNA sequences that perfectly match the ones that are designed.
Several approaches were considered to improve the oligonucleotide assembly method described here. Yeast spheroplasts can take 2–3 days to prepare. However, it has previously been demonstrated that competent yeast spheroplasts can be prepared and stored frozen (22
). The use of frozen yeast spheroplasts would facilitate this method by having the DNA assembly host organism ready-to-go. In the transformation of oligonucleotides 1–38 u, 1–28 g and HF1‐HF6 into both fresh and frozen spheroplasts, colonies were obtained in all cases (Supplementary Table S9
). Although the transformation efficiency is reduced, enough colonies are produced to allow the identification of the assembled DNA fragments in yeast. One drawback of the method described here is the 3–4-day wait-time for cloning and growing yeast clones that contain the assembled DNA constructs. In general, however, E. coli
is the preferred host organism for propagating these recombined DNA molecules. Thus, it would be interesting to determine whether assembled fragments could be immediately cloned in E. coli
following yeast transformation, without selection and propagation in yeast. This should be possible if the recombination efficiency is high and the assembled DNA molecules can be recovered from yeast for immediate transfer to E. coli
Here the issue of process complexity in gene synthesis is addressed. Overlapping oligonucleotides are assembled and cloned in a single step. But, a method such as this one may also be useful when traditional in vitro
assembly methods fail (likewise, in vitro
approaches could be useful if this in vivo
approach fails). The oligonucleotides used in gene synthesis are often optimized so that the overlapping sequences have similar melting temperatures (27–30
). In the gene-synthesis experiments shown here, no effort was made to optimize the oligonucleotide sequences. For example, in the assembly of oligonucleotides 1–28 g, the melting temperatures at the overlaps differed by as much as 16.8°C (Supplementary Table S10
). Furthermore, many of the oligonucleotides used in this study had complete homology to yeast genomic sequences. Assembly of synthetic DNA sequences with oligonucleotides having strong homology to the yeast genome can be explained by the preference for homologous recombination at free ends of DNA (31
). Living biological cells carry out an extraordinary amount of processes that cannot be duplicated elsewhere. Chemical synthesis in yeast from overlapping oligonucleotides is likely no exception. Interestingly, in the assembly of the six 200-mer oligonucleotides (), 44% of the DNA fragments are synthesized by yeast (960 of 2200 nucleotides are added in vivo
The work described here extends beyond gene synthesis and may be applicable to other schemes in yeast that utilize oligonucleotide transformation for gene knockouts or manipulations (14
) and linker-mediated recombination (16
). It could also be useful for creating combinatorial libraries and for selecting a particular phenotype in yeast.