Significance and context
One way of finding protein or DNA sequences that bind a target molecule is to screen large libraries. Each DNA sequence in the library, or its expressed protein, is offered to the target, and those that bind tightly are selected and identified. Protein designers search libraries synthesized from random DNA sequences for new, non-biological binders; biochemists search libraries of genomic sequences for biological binders. Two standard library screening methods are the yeast two-hybrid and phage display techniques. Joung et al. describe a new screening method which is related to the two-hybrid scheme, but uses bacteria as the host. The method may prove a useful alternative to traditional screening approaches.
Joung et al. test their technique by screening a library of three-finger zinc-finger proteins (derived from Zif268) which are randomized at six positions on one finger. The targets used to screen this library are sequences from the nuclear receptor element (NRE), TATA box or p53 promoters. The authors report the sequences of zinc-finger variants that bind the targets. Some are specific: they bind only one target. As results are available from a comparable phage display experiment which used the same targets, Joung et al. can compare their binders to those identified from phage display. Many of the zinc finger variants that bind the TATA and p53 sequences appear to be similar in the two cases, but those that bind the NRE are different. As a control, Joung et al. used the NRE binders identified from phage display as binders in the bacterial two-hybrid protocol. Results indicated that the phage display sequences did not bind their targets in bacteria, suggesting that the bacterial screen may be more stringent.
The technique is an extension of a bacterial two-hybrid scheme previously reported by the same authors in 1997. In this implementation, targets are DNA sequences and test candidates from the library are proteins. A test candidate from the library is identified as a strong binder to the target if it can promote transcription of a reporter gene in vivo by causing interactions between a pair of mediator proteins. Joung et al. use sequences from the NRE, TATA box or p53 promoters as target DNA. The library contains variants of zinc-finger proteins and the reporter gene is the bacterial gene hisB, which encodes an enzyme of the histidine biosynthetic pathway. Their mediator proteins are Gal4 and Gal11P from yeast. It all comes together as follows. Joung et al. make one set of plasmids containing the hisB gene under the control of a promoter containing one of the targets (NRE, TATA or p53). The authors make another set of plasmids containing GAL4 fused to the gene for the RNA polymerase alpha subunit from Escherichia coli. They transfect both plasmids into a strain of E. coli that lacks hisB. Next they make oligonucleotides of the zinc-finger test sequences, ligate each to the gene for the mediator protein Gal11P, and introduce them into phage to form the library. Finally, the authors transfect the library phage into the deficient E. coli and grow them in medium lacking histidine. A successful 'hit' happens in the following sequence. E. coli expresses the zinc-finger test protein linked to Gal11P, and Gal4 linked to the polymerase subunit. If the test protein binds to the target promoter, Gal11P recruits Gal4 to the promoter. The polymerase linked to Gal4 causes transcription of hisB and the enzyme is synthesized. E. coli is then able to synthesize histidine from the medium and grow stably.
Joung et al. conclude that their method can pick out DNA-binding proteins from a randomized library. These proteins may bind more strongly and more specifically than those identified by phage display, and the procedure to detect them may be faster than in a yeast two-hybrid screen because E. coli has a higher transformation efficiency and growth rate.
This paper describes a complex and creative new technique for library screening. What it does not provide is a quantitative comparison against previous methods. For that, Joung et al. will need to measure their identified proteins' binding affinities to the target. The authors also do not provide evidence for the 'coverage' of their library - how many unique sequences they screened in E. coli.