ch4106.qxd 02/15/2000 09:01 Page 34 34 Design of novel sequence-specific DNA-binding proteins David J Segal and Carlos F Barbas III* The design and selection of DNA-binding proteins or individual domains capable of novel sequence recognition continues to make great strides. Recent studies have also highlighted the role of the non-DNA-contacting portions of the protein and the optimal assembly of the domains. For the first time, it appears that it is possible to produce proteins capable of targeting any gene with an 18 base pair recognition domain. A variety of applications are being explored, such as targeted transcriptional regulation, recombination and viral integration. These proteins will probably find diverse applications in gene therapy, functional genomics, and agriculture. Addresses The Skaggs Institute for Chemical Biology and the Department of Molecular Biology, The Scripps Research Institute, BCC-515, North Torrey Pines Road, La Jolla, CA 92037, USA *e-mail: carlos@scripps.edu Current Opinion in Chemical Biology 2000, 4:34–39 1367-5931/00/$ — see front matter © 2000 Elsevier Science Ltd. All rights reserved. Abbreviations DBD DNA-binding domain ER estrogen receptor PR progesterone receptor Introduction The design of proteins with the capacity to recognize extended nucleotide sequences with high affinity and specificity has been a goal of two decades of research; a goal that has remained elusive until the past two years. Ideally, proteins could be designed to specifically regulate the transcription of a single gene in a genome as complex as our own four billion base pair genome. These proteins would allow scientists to reach into genomes and to modulate the transcription of a gene of interest, to assign its function or to change the phenotype of the organism. Specific recognition of double-stranded DNA by proteins is most typically accomplished using an α-helix as the key recognition element. An α-helix provides a rigid scaffold from which amino-acid sidechains are presented to the accessible edges of the bases in the major grove of DNA. This presentation is modulated by ‘framework’ structural elements of the protein. For proteins in which the α-helix is the key recognition element, the task of creating DNA-binding proteins with novel sequence specificity consists of modifying the DNA-contacting residues in the α-helix, then modifying the framework to optimize the display of the α-helix. In this review, we discuss the recent successful application of these principles to create novel DNA-binding proteins, with particular emphasis on the advances achieved using Cys2-His2 zinc finger proteins. Helical domains For proteins that do not contact DNA with an organized α-helix, success in generating novel sequence-specific proteins has been extremely limited. For example, the base-specific hydrogen bonds between the restriction endonuclease EcoRV and its recognition base pairs are made through a highly cooperative set of interactions by five residues in a surface loop of the enzyme (Figure 1a; [1]). Through application of DNA shuffling, mutagenesis and screening, Pingoud and co-workers [2••] were able to extend recognition to include the base pairs immediately flanking the EcoRV site. Although this is an impressive study, it remains to be seen if more can be expected from these types of complex recognition motifs. A decade ago, the dimeric bZIP proteins were proposed as candidates for the development of novel DNA-binding proteins [3]. Each bZIP monomer contacts DNA with an α-helix that is oriented and constrained by being part of longer α-helix, which makes up the leucine zipper dimerization interface (Figure 1b; [4]). In vitro selection methods produced mutants of the bZIP protein C/EBP that could recognize sequences that differed in two of its five half-site nucleotides [5]. Interestingly, some of the mutations occurred in the hinge region between the α-helix of the DNA-binding domain (DBD) and the leucine zipper framework, suggesting that reorientation of the α-helix may have been required. Theoretically, DBDs of bZIP heterodimers, such as JunFos, could be modified to recognize novel 10-nucleotide, non-palindromic sites. Alternatively, tandem monomers of the bZIP DBDs could potentially be used to recognize extended sequences, but these typically bind poorly to DNA. In an elegant design approach, the affinity of a bZIP monomer DBD, that of GCN4, was improved 270-fold by grafting the DNA-contacting residues from the α-helix of GCN4 onto the stable and compact scaffold of the avian pancreatic polypeptide [6••]. This miniature 42-aminoacid protein, consisting of a single amphipathic α-helix that is stabilized through hydrophobic interactions with a type II polyproline helix, was able to bind a GCN4 half site with an affinity and specificity greater than that of the parental α-helix. This study again illustrates that binding can be improved by optimizing not only the DNA-contacting residues but also the display and stability of the α-helix. It remains unclear if these domains can be effectively linked as tandem multimers to recognize extended sites. Chemical linkage of two mini-proteins provided a ‘dimer’ with increased affinity but lower specificity [6••]. Nuclear receptors, such as the estrogen receptor (ER) and progesterone receptor (PR), also bind as dimers to their DNA targets. Each monomer inserts an α-helix of a ch4106.qxd 02/15/2000 09:01 Page 35 Design of novel sequence-specific DNA-binding proteins Segal and Barbas 35 Figure 1 DBDs that have been used to create proteins with novel sequence specificity. (a) EcoRV [1]. (b) bZIP of GCN4 [4]. (c) Cys4 zinc fingers of glucocorticoid receptor [8]. (d) Helix-turn-helix domains of Myb [10]. (e) Cys2–His2 zinc fingers of Zif268 [13]. Double stranded DNA is shown in orange and brown, DBD domains are shown in red, blue or green ribbon representation, and zinc ions are shown as yellow spheres. (b) (a) (d) (c) (e) Current Opinion in Chemical Biology Cys4-type zinc finger into the major grove, and for some receptors the DBD contains a strong dimerization interface (Figure 1c; [7,8]). Using in vivo survival-based selection of randomized libraries, Shapiro and co-workers [9••] found PR-DBD mutants that lost their ability to recognize the PR response element (PRE) but could bind the ER response element (ERE), which differs in two of the six half-site nucleotides. The mutants bound the ERE with a 15-fold higher affinity than wild-type ERDBD, albeit with a specificity considerably broader than wild-type ER-DBD. From domains to proteins Tandemly repeated elements, although having the potential for extended, non-palindromic recognition, introduce another level of complexity: inter-domain cooperativity. One domain may make protein–protein contacts with the next domain, affect the binding geometry of adjacent domains, or it may contact the nucleotides of another domain’s binding site. These interactions that allow for concerted recognition may be considered as evolutionary levers for optimizing the binding of a particular protein; however, for those seeking to design novel binding proteins these interactions present additional challenges. For example, Myb domains (Figure 1d) consist of two or three tandem repeats (designated R1, R2 and R3) of a helix-turn-helix motif, similar to the motif found in the λ repressor and homeodomian proteins [10]. Myb DBDs with novel specificity have been generated by combining the R2 and R3 of different species [11]; however, the combination of two tandem R3 repeats severely reduced binding affinity [12]. The authors concluded that cooperative interactions between the R2 and R3 domains were required for high affinity binding. By any measure, the greatest success in producing proteins with novel binding specificity has been achieved with the classic Cys2–His2 zinc-finger domains (Figure 1e). These zinc fingers are compact domains containing a single amphipathic α-helix stabilized by two β-strands and zinc ligation [13]. Like the Myb domains, zinc-finger proteins contain multiple tandem repeats and display varying degrees of inter-domain cooperativity. Fortunately, a subset of zinc-finger proteins, including the murine transcription factor Zif268 and the human protein Sp1, display only minimal — though still troublesome — cooperativity. In these proteins, each finger recognizes a three nucleotide site with relative independence, which has allowed several groups to produce zinc fingers with novel specificities using rational or combinatorial methods (reviewed in [14••]). Early attempts to produce zinc fingers with novel sequence recognition gave hope to the idea that there might be a simple 1:1 amino acid to base recognition code that could be used to build fingers that could specifically recognize any three nucleotide sequence [15,16]. Unfortunately, this simplistic code is poorly predictive of the actual specificity of these domains [17,18•]. In one report of a code-based three-finger protein, only five of the nine nucleotides appear to be correctly specified [17]. There are at least three reasons why a simple code is insufficient for comprehensive recognition. The first is technical. Phage display has proven to be a powerful tool for revealing the repertoire of zinc-finger–DNA interactions. This procedure involves the display of a large (typically 107–109 member) library of randomized proteins on the surface of a filamentous bacteriophage. Because ch4106.qxd 02/15/2000 36 09:01 Page 36 Interaction, assembly and processing Table 1 Affinities of sequentially-selected and stitched three-finger proteins. Protein Target sequence Published KD (nM)‡ Normalized KD (nM)§ Stitched study Zif268* B3* C5* E2C-HS2(SP1)* GCG GGA GGA GCC TGG GGG GGC GCA GCG GAC GGG GTG 10 4 30 25 10 4 30 25 Sequential study Zif268† TATAZF† p53ZF† NREZF† GCG GCT GGG AAG GGG ATA ACA GGT GCG AAA TGT TCA 0.010 0.12 0.11 0.038 10 120 110# 38 *Data from [23••]. †Data from [22]. ‡Standard deviation < 60%. §Published values normalized to Zif268 = 10 nM. #p53 ZF was later reported to bind with an affinity higher than NREZF [18•]. each phage contains the gene for the protein it displays, the sequence of proteins having the desired properties can be identified through repeated cycles of in vitro selection and amplification of the phage library (known as ‘panning’). There is, however, a common misconception that any sequence that is strongly selected from panning must be the optimal sequence. The flaw in that logic is that the researcher is not always aware of what the actual selection pressure is during an experiment. Our approach [14••] has been to optimize the output of phage display using sitedirected mutagenesis. This study revealed that some residues were selected during panning because they increased the affinity of the interaction at the cost of specificity. The end product of this study was a collection of 16 well-characterized domains recognizing each of the possible GNN binding sites. A number of these domains demonstrated exquisite specificity and discriminated between sequences that differ by one in nine bases with > 100-fold loss in affinity. Sequential selection versus stitchery The second reason for the problems with simple coded interactions relates to cooperativity. A central theme in recent zinc-finger research has been to refocus on the question: just how independent are the zinc fingers of Zif268? An early clue was the difficulty that every laboratory encountered in selecting fingers that could recognize sequences of the type ANN or CNN, instead of GNN — the sequence type recognized by the natural protein. With Zif268 as their scaffold protein, a consensus in the field emerged that aspartate in position 2 of one α-helix contacted the binding site of the finger next to it, forcing recognition at that neighboring site to be GNN or TNN [19,20•]. The significance of this contact was underappreciated in the original Zif268 crystal structure [13] but is now widely acknowledged and constitutes what is known as the target site overlap problem in zinc fingers [14••]. Limitations imposed by this contact can be addressed by randomization of residues at the interface of two α-helices including position 2 [20•,21], or by replacement of the offending domain with a finger that does not contain aspartate in position 2 and is not expected to have any cooperative effects. Concerns regarding cooperativity can also be addressed by the selection methodology if sequential selections are applied [22]. In this approach (Figure 2a) finger 1 of a three-finger protein is combinatorially selected in the context of two ‘anchor’ fingers. Subsequently, the terminal anchor finger is removed, finger 1 becomes finger 2, and a new finger 1 is selected to bind the next three nucleotides in the DNA sequence. After one additional round of exchanges, a new three-finger protein is created in which all of the fingers have been selected in the ‘context’ of the finger next to it. This is in contrast to direct selection of defined fingers from a single library that are then ‘stitched’ together (or at least the contact residues) to make a new three-finger protein (Figure 2b). The advantage of the stitchery method is that once fingers of high affinity and specificity are created, they can be rapidly reassembled to recognize any desired sequence. However, this aim requires that each finger be completely modular and independent. On the other hand, sequential selection requires that three sequential libraries be constructed and selected for each three-finger protein, making zinc-finger technology inaccessible to most researchers. Is sequential selection necessary? Several stitched threefinger proteins have been constructed from an optimized set of GNN-binding fingers [14••,23••]. In these fingers, cooperativity effects caused by aspartate in position 2 are minimized as all of the target sites are of the GNN type. The published specificity of sequentially selected proteins appears to be no better than that of the stitched proteins [18•,23••]. The affinity of the sequentially selected proteins was reported to be in the sub-nanomolar range [22]. Regrettably, however, the issue of affinity has become somewhat distorted in the zinc-finger field. As one example, the affinity of Zif268 for its operator ‘improved’ 600-fold by modification of the binding measurements [13,22]. Although the use of non-standard binding conditions when measuring affinities is legitimate, the values obtained can not be directly compared to affinities measured in a more conventional way. A more accurate comparison can be obtained by normalizing the published affinities to a common value for Zif268 as this measurement is also reported. As shown in Table 1, the sequentially selected proteins bind their target with a lower affinity than the stitched proteins. On the basis of this comparison the advantages of sequential selection are not clear. The comparison is not entirely fair, however, as the sequentially selected proteins contain fingers that recognize ANN and TNN sites, which may be inherently more difficult for zinc fingers to bind with high affinity and specificity. Studies in live cells suggest that specificity is a problem as the TATA protein was not effective in reporter gene studies [24]. ch4106.qxd 02/15/2000 09:01 Page 37 Design of novel sequence-specific DNA-binding proteins Segal and Barbas 37 Figure 2 Sequential and parallel selection methods. Boxes represent zinc finger domains. Letters beneath the boxes represent DNA sequences. See text for details. (a) Sequential selection (b) Anchor Anchor Library GCG TGG ABC DEF Anchor ABC Library TGG ABC DEF GCG ABC GCG TGG ABC ABC DEF DEF DEF GHI Parallel selection and stitchery Anchor Library Anchor GCG ABC GCG Anchor Library Anchor GHI GCG DEF Library Anchor Library GHI GCG GHI GHI ABC DEF GCG Anchor GCG GHI Current Opinion in Chemical Biology Framework effects A third reason for the problems with simple coded interactions is that a code fails to consider the influence of the linker region, the β-strands, and the carboxy-terminal end of the α-helix (collectively referred to as the framework). Framework effects on specificity and affinity are only beginning to be appreciated and will undoubtedly represent a new area of research in this field. Structures of Zif268 mutants binding their designed targets clearly show a repositioning of the α-helix relative to the DNA, suggesting that the α-helix might require reorientation for the optimal presentation of its contacting residues to variant DNA sites [25••]. Variation in helical presentation is also clearly evident in recent structural studies of three-finger and six-finger proteins derived from TFIIIA [26,27]. In our own studies [23••], a 50-fold increase in affinity was observed when the same six α-helices were displayed on different frameworks. Ryan and Darby [28] reported an eightfold loss in affinity and a 50-fold loss in specificity after modifying only the linker regions of TFIIIA. Appropriate display of the α-helix becomes particularly important as the number of tandem domains is increased to six, the number required to recognize a unique sequence in the human genome [29]. No natural zinc finger proteins have been found that bind specifically to 18 contiguous nucleotides. Additionally, the DNA duplex is unwound slightly upon binding, causing concern that the contact residues in a long protein may come out of register with the nucleotides, resulting in a loss in specificity and affinity. Our studies [23••,29] suggest that the affinity of a six-finger protein is highly dependant on the framework. In another example, the use of an 11-amino-acid linker peptide produced a six-finger protein with an affinity > 6000-fold greater than that of its three-finger components [30•]. Framework mutations have tended to be context-dependent in the zinc finger area and application of this 11-amino-acid linker to other proteins has not resulted in other examples with exceptional affinity. Additional studies that correlate structure with quantitative affinity and specificity data will be required to further reveal how framework effects influence the ability of zincfinger proteins to both recognize specific sequences and exclude highly related ones. Despite these complexities, a polydactyl protein specifically recognizing an 18-nucleotide sequence in the 5′-untranslated region of the human erbB-2 gene has been constructed. This protein bound the target sequence with 0.5 nM affinity. When expressed as a fusion protein with repressor or activator domains, the transcription factor was able to specifically regulate the erbB-2 promoter [23••]. Extension of this work to the study of the regulation of the endogenous erbB-2 gene has resulted in the first demonstration of the modulation of endogenous genes with designed transcription factors (CF Barbas III, unpublished data; see Note added in proof). These proteins were shown to be effective in cells derived from three mammalian species that conserve the erbB-2 binding site. Amazingly, the erbB-3 gene is not regulated despite a 15 of 18 nucleotide match to the erbB-2 target site at the same relative position. Furthermore, construction of an erbB-3-targetting transcription factor enables specific regulation of the gene as well. Thus it appears that genome-specific regulation is now in reach. Conclusions With the demonstration that polydactyl zinc-finger proteins can be readily assembled to recognize 18 base-pair sequences of the 5′-(GNN)6-3′ type, and that these proteins can specifically regulate an endogenous gene, much additional study will be needed to extend the type of sequences that can be addressed by this approach. Some sequence motifs may require the use of other DNA-binding proteins or combinations of domains of different types. This might be accomplished through the selection of novel dimerizing regions that serve to link varied proteins noncovalently [31••]. Now that we have begun to be able place ch4106.qxd 02/15/2000 38 09:01 Page 38 Interaction, assembly and processing proteins at specific addresses within complex genomes, what activities might these proteins direct? Given the modular nature of proteins that act on DNA, a vast array of tools that facilitate the manipulation of genomes is possible. Appending activation or repression domains onto a DBD allows for transcriptional modulation [23••,29], while endonuclease [32,33•], methylase [34], topoisomerase [35], and HIV-1 integrase [36] domains have allowed for some degree of site-directed action on DNA. Although many of these novel functions are still too inefficient for practical application, they are certain to have an impact on basic and applied biology in this new era of the genome. Note added in proof The paper referred to in the text as (CF Barbas III, unpublished data) has now been accepted for publication [37]. References and recommended reading Papers of particular interest, published within the annual period of review, have been highlighted as: • of special interest •• of outstanding interest 1. Winkler FK, Banner DW, Oefner C, Tsernoglou D, Brown RS, Heathman SP, Bryan RK, Martin PD, Petratos K, Wilson KS: The crystal structure of EcoRV endonuclease and of its complexes with cognate and non-cognate DNA fragments. EMBO J 1993, 12:1781-1795. 2. •• Lanio T, Jeltsch A, Pingoud A: Towards the design of rare cutting restriction endonucleases: using directed evolution to generate variants of EcoRV differing in their substrate specificity by two orders of magnitude. J Mol Biol 1998, 283:59-69. This paper describes one of the very few modifications of DNA recognition in an enzyme using a non-α-helical recognition motif. DNA shuffling was used to evolve the specificity of the enzyme to a longer sequence than its native target. Extension of the sequence specificity of restriction enzymes to longer sequences promises many practical applications in cloning and recombination. 3. 4. 5. O’Neil KT, Hoess RH, DeGrado WF: Design of DNA-binding peptides based on the leucine zipper motif. Science 1990, 249:774-778. Ellenberger TE, Brandl CJ, Struhl K, Harrison SC: The GCN4 basic region leucine zipper binds DNA as a dimer of uninterrupted α-helices: crystal structure of the protein–DNA complex. Cell 1992, 71:1223-1237. Sera T, Schultz PG: In vivo selection of basic region-leucine zipper proteins with altered DNA-binding specificities. Proc Natl Acad Sci USA 1996, 93:2920-2925. 6. •• Zondlo NJ, Schepartz A: Highly specific DNA recognition by a designed miniature protein. J Am Chem Soc 1999, 121:69386939. This study demonstrates minimalist design at its best. Simple grafting of the recognition residues of GCN4 onto the well-studied avian pancreatic polypeptide and a little fine-tuning provided a small 42-residue protein with exquisite affinity and specificity for the target DNA, much greater than the native protein. 7. 8. 9. •• Schwabe JW, Chapman L, Finch JT, Rhodes D: The crystal structure of the estrogen receptor DNA-binding domain bound to DNA: how receptors discriminate between their response elements. Cell 1993, 75:567-578. Luisi BF, Xu WX, Otwinski Z, Freedman LP, Yamamoto KR, Sigler PB: Crystallographic analysis of the interaction of the glucocorticoid receptor with DNA. Nature 1991, 352:497-505. Chusacultanachai S, Glenn KA, Rodriguez AO, Read EK, Gardner JF, Katzenellenbogen BS, Shapiro DJ: Analysis of estrogen response element binding by genetically selected steroid receptor DNA binding domain mutants exhibiting altered specificity and enhanced affinity. J Biol Chem 1999, 274:23591-23598. This study is the most extensive analysis of the elements of recognition for Cys4-type zinc-finger binding motifs. Novel randomization and in vivo selec- tion methods are described. The authors are poised to discover if these Cys4 zinc fingers are capable of the versatility of their Cys2–His2 cousins. 10. Ogata K, Morikawa S, Nakamura H, Sekikawa A, Inoue T, Kanai H, Sarai A, Ishii S, Nishimura Y: Solution structure of a specific DNA complex of the Myb DNA-binding domain with cooperative recognition helices. Cell 1994, 79:639-648. 11. Williams CE, Grotewold E: Differences between plant and animal Myb domains are fundamental for DNA binding activity, and chimeric Myb domains have novel DNA binding specificities. J Biol Chem 1997, 272:563-571. 12. Oda M, Furukawa K, Sarai A, Nakamura H: Construction of an artificial tandem protein of the c-Myb DNA-binding domain and analysis of its DNA binding specificity. Biochem Biophys Res Commun 1999, 262:94-97. 13. Pavletich NP, Pabo CO: Zinc finger-DNA recognition: crystal structure of a Zif268-DNA complex at 2.1 Å. Science 1991, 252:809-817. 14. Segal DJ, Dreier B, Beerli RR, Barbas CF III: Toward controlling •• gene expression at will: selection and design of zinc finger domains recognizing each of the 5′′-GNN-3′′ DNA target sequences. Proc Natl Acad Sci USA 1999, 96:2758-2763. The domains reported in this paper were optimized using a combination of phage display and rational mutagenesis. Domains are disclosed that recognize any GNN target sequence (guanine followed by any two bases. Many of these domains exhibit exquisite specificities not previously observed for zinc-finger proteins. Together with [23••], this paper describes a system by which any laboratory can build a zinc finger protein capable, perhaps, of uniquely targeting any gene in a complex genome. 15. Choo Y, Klug A: Selection of DNA binding sites for zinc fingers using rationally randomized DNA reveals coded interactions. Proc Natl Acad Sci USA 1994, 91:11168-11172. 16. Desjarlais JR, Berg JM: Use of a zinc-finger consensus sequence framework and specificity rules to design specific DNA binding proteins. Proc Natl Acad Sci USA 1993, 90:2256-2260. 17. Corbi N, Libri V, Fanciulli M, Passananti C: Binding properties of the artificial zinc fingers coding gene Sint1. Biochem Biophys Res Commun 1998, 253:686-692. 18. Wolfe SA, Greisman HA, Ramm EI, Pabo CO: Analysis of zinc • fingers optimized via phage display: evaluating the utility of a recognition code. J Mol Biol 1999, 285:1917-1934. This paper reports the specificity of the proteins constructed by sequential selection (see [22]). The authors also argue that many of the interactions they observed were not predicted by a recognition code. 19. Isalan M, Choo Y, Klug A: Synergy between adjacent zinc fingers in sequence-specific DNA recognition. Proc Natl Acad Sci USA 1997, 94:5617-5621. 20. Isalan M, Klug A, Choo Y: Comprehensive DNA recognition through • concerted interactions from adjacent zinc fingers. Biochemistry 1998, 37:12026-12033. By randomizing residues at the interface of two domains, the authors showed that they could select residues that would allow the recognition of sequences starting with any base. It is still not clear if these results can be generalized, however, and the authors have yet to show that they can apply their methods to make multi-finger proteins of high affinity and specificity or that purified rather than phage-bound proteins actually bind the target sequence. 21. Jamieson AC, Wang H, Kim S-H: A zinc finger directory for highaffinity DNA recognition. Proc Natl Acad Sci USA 1996, 93:1283412839. 22. Greisman HA, Pabo CO: A general strategy for selecting highaffinity zinc finger proteins for diverse DNA target sites. Science 1997, 275:657-661. 23. Beerli RR, Segal DJ, Dreier B, Barbas CF III: Toward controlling •• gene expression at will: specific regulation of the erbB-2/HER-2 promoter by using polydactyl zinc finger proteins constructed from modular building blocks. Proc Natl Acad Sci USA 1998, 95:14628-14633. This paper presents the assembly, affinity, specificity, and biological activity of six-domain polydactyl zinc-finger proteins using the domains described in [14••]. Transcription factors recognizing an 18 base pair sequence are assembled and demonstrated to specifically control the erbB-2 promoter in human cells, the first time that transcription factors have been fashioned to regulate a specific promoter. Together, these papers provide the necessary tools by which any laboratory can build a zinc finger protein capable, perhaps, of uniquely targeting any gene in the human genome. ch4106.qxd 02/15/2000 09:01 Page 39 Design of novel sequence-specific DNA-binding proteins Segal and Barbas 24. Kim JS, Pabo CO: Transcriptional repression by zinc finger peptides. Exploring the potential for applications in gene therapy. J Biol Chem 1997, 272:29795-29800. 25. Elrod-Erickson M, Benson TE, Pabo CO: High-resolution structures •• of variant Zif268–DNA complexes: implications for understanding zinc finger–DNA recognition. Structure 1998, 6:451-464. This paper presents the structures of phage display selected mutants. The structures show interactions in cognate and non-cognate sites. The paper also describes the repositioning of the α-helix relative to the DNA. This is an excellent and necessary descriptive work, although the paucity of quantitative affinity and specificity data for these proteins allows for few conclusions regarding specific interactions that provide for binding one sequence while excluding others. 26. Nolte RT, Conlin RM, Harrison SC, Brown RS: Differing roles for zinc fingers in DNA recognition: structure of a six-finger transcription factor IIIA complex. Proc Natl Acad Sci USA 1998, 95:2938-2943. 27. Wuttke DS, Foster MP, Case DA, Gottesfeld JM, Wright PE: Solution structure of the first three zinc fingers of TFIIIA bound to the cognate DNA sequence: determinants of affinity and sequence specificity. J Mol Biol 1997, 273:183-206. 28. Ryan RF, Darby MK: The role of zinc finger linkers in p43 and TFIIIA binding to 5S rRNA and DNA. Nucleic Acids Res 1998, 26:703-709. 29. Liu Q, Segal DJ, Ghiara JB, Barbas CF III: Design of polydactyl zincfinger proteins for unique addressing within complex genomes. Proc Natl Acad Sci USA 1997, 94:5525-5530. 30. Kim JS, Pabo CO: Getting a handhold on DNA: design of poly-zinc • finger proteins with femtomolar dissociation constants. Proc Natl Acad Sci USA 1998, 95:2812-2817. This work suggests that modifications to the framework of a zinc-finger protein can lead to dramatic improvements in binding. The authors state that by using a noncanonical linker between zinc fingers 3 and 4, a > 6000-fold increase in affinity can be obtained by a six-finger protein over that of its constituent three-fingers proteins. It remains to be seen if specificity is sacrificed, or if this effect can be reproduced with other proteins. To date, further application of this approach has failed. More importantly, it is not clear that 39 the effect has anything to do with the noncanonical linker, as it appears that the canonical linker was never tested. Kinetic measurements are used to derive dissociation constants. As performed, these measurements can lead to very large errors. Therefore, the use of the term ‘femtomolar’ in the title and abstract of this paper is provocatively misleading. Normalizing the reported affinities to a more traditional value for Zif268 produces dissociation constants in the picomolar range. Such affinities are still impressive, but are achievable with canonical linkers. 31. Zhang Z, Murphy A, Hu JC, Kodadek T: Genetic selection of short •• peptides that support protein oligomerization in vivo. Curr Biol 1999, 9:417-420. In this study a genetic selection is devised to select novel peptides that result in dimerization or multimerization of the λ repressor. Effective peptide motifs allow cell survival following assault with the lytic phage. An amazingly diverse selection of novel functional peptides results from this. 32. Nahon E, Raveh D: Targeting a truncated HO-endonuclease of yeast to novel DNA sites with foreign zinc fingers. Nucleic Acids Res 1998, 26:1233-1239. 33. Chandrasegaran S, Smith J: Chimeric restriction enzymes: what is • next? Biol Chem 1999, 380:841-848. This review describes fusions of zinc finger proteins and the gal4 DBD with the endonuclease domain of FokI. The zinc finger proteins were constructed entirely by rational design. The fusions exhibit activity, though their ability to catalyze multiple DNA cleavages is not assessed. Efficient site-specific enzymes may be only an evolutionary selection away. 34. Xu G-L, Bestor TH: Cytosine methylation targetted to predetermined sequences. Nat Genet 1997, 17:376-378. 35. Beretta GL, Binaschi M, Zagni E, Capuani L, Capranico G: Tethering a type IB topoisomerase to a DNA site by enzyme fusion to a heterologous site-selective DNA-binding protein domain. Cancer Res 1999, 59:3689-3697. 36. Bushman FD, Miller MD: Tethering human immunodeficiency virus type 1 preintegration complexes to target DNA promotes integration at nearby sites. J Virol 1997, 71:458-464. 37. Beerli RB, Dreier B, Barbas CF III: Selective positive and negative regulation of endogenous genes by designed transcription factors. Proc Natl Acad Sci USA 2000, in press.