Directed evolution: overcoming biology's limitations.

The field of directed, in vitro evolution uses biological machinery to develop nucleic acid and peptide oligomers that bind to molecular targets, some of which are important for therapeutic applications1,2. The advantage of this approach over traditional small molecule–based medicinal chemistry is based in part on the number of molecules that may be screened: in vitro evolution can screen upwards of 1015 different molecules for binding to a target, whereas the best combinatorial libraries of small molecules are in the more modest range of 102 to 107 molecules3,4. The utility of this approach is nicely illustrated by the drug pegaptanib sodium, which was approved by the US Food and Drug Administration (FDA) in 2004 for treatment of macular degeneration5. The initial lead for this drug was identified by in vitro evolution screening of an oligonucleotide library for binding to vascular endothelial growth factor (VEGF). The main weakness of evolution-based techniques, however, has been the susceptibility to degradation of the oligonucleotides and peptides obtained from the process. Pegaptanib sodium, for example, is a highly modified version of the original oligonucleotide in which numerous chemical groups had to be incorporated to slow degradation. Furthermore, the reliance on enzymes for in vitro evolution has largely constrained the field to using natural biopolymers, although excellent advances have been made in identifying non-natural nucleotides and amino acids that are tolerated by enzymes in the process6. A new study in this issue paves the way to use a class of non-natural polymers in directed evolution7. Re-engineering in vitro evolution to use a completely non-natural backbone would eliminate concerns about degradation. In this regard, a class of molecules called peptide nucleic acids (PNAs) would appear to be ideal. First described by Nielsen et al. in 1991 (ref. 8), the aminoethylglycyl PNA (aegPNA) has become one of the gold standards among non-natural nucleic acid mimics that bind strongly and in a highly sequence-dependent manner to complementary DNA or RNA sequences9,10. PNAs are completely resistant to degradation by nucleases and proteases11; however, the polymerases that are essential for the process of in vitro evolution also do not recognize PNAs12. In the manuscript by Brudno et al.7, the authors overcome two significant obstacles to using PNA in the process of in vitro evolution. The first key to this system was to eliminate the need for a polymerase to generate a library of PNA oligomers. In this study, the authors cleverly and carefully optimized conditions for a DNAtemplated synthesis of PNA that does not require enzymes and still has fidelity similar to that seen with polymerase replication (Fig. 1a). In this paper, DNA templates that are 40–60 nucleotides long are used to condense a mixture of 12 PNA pentamer building blocks. Once the PNA library has been made (Fig. 1b), liberating the PNAs from the DNA templates to which they are bound is a challenge, as PNA binds very strongly to complementary DNA. After extensive screening, the Herculase II DNA polymerase was found to efficiently displace PNA from the DNA template at the 40-nucleotide length (Fig. 1c). Moving forward, the authors next demonstrate that the DNA-coding template strand that was used to make the PNA oligomer can be isolated and amplified in a PCR protocol (Fig. 1d). Though these three steps (translation, displacement and amplification) constitute most of the basic requirements of in vitro evolution, the process still needs to be iterated for several rounds of selection. Subjecting this system to such a test, the authors generated a mock selection system in which one PNA pentamer was labeled with biotin. The labeled PNA was uniquely incorporated into a DNA template that was