1996 (Introduction from Ph.D. Thesis)

Recognition of specific DNA sequences is a crucial step associated with many cellular processes and among eucaryotes, the zinc finger is a ubiquitous DNA-binding motif involved in this step (Klug and Rhodes, 1987). Since this motif was first i dentified in Xenopus laevis transcription factor IIIA (TFIIIA) in 1985 (Miller et al., 1985), literally hundreds of proteins, many of them transcription factors, have been shown to contain zinc fingers. In fact, all eucaryotes tested to date cont ain one or more sequences that hybridize to zinc finger-derived probes (Chowdhury et al., 1987; Schuh et al, 1986) and over 100 human DNA sequences encoding putative zinc finger protein genes have been isolated (Hoovers, et al., 1992). Among the better c haracterized zinc finger proteins are yeast transcription factor ADR1 (Bemis and Denis, 1988), human transcription factor Sp1 (Gidoni et al., 1984), mouse transcription factor Zif268 (Christy et al., 1988) and the gene product from human gene ZFY, the testis-determining factor (Kochoyan at al., 1991). The zinc finger motif has also been implicated in Drosophila development, being encoded by the segmentation genes Kruppel (Rosenberg et al., 1986) and Hunchback (Tautz et al., 1987) and by genes thought to be involved in tumorigenesis, including GLI (Kinzler et al., 1988) and a candidate for the Wilm's tumor gene (Gessler et al., 1990). A survey of zinc finger proteins shows a significant variation in the number of zinc finger motifs per protein, ranging from 2 (ADR1), 3 (Sp1, Zif268), 9 (TFIIIA), 13 (ZFY) and up to 37 (Xfin, a Xenopus RNA-binding protein).

TFIIIA is considered the archetypal zinc finger protein. Not only was it the first protein discovered to contain the zinc finger motif, but it was also the first eucaryotic transcription factor to be purified. Because of the great abundance of TFI IIA in Xenopus oocytes, where it exists in a complex with 5S RNA (the cytoplasmic 7S storage particle), it was purified to homogeneity from ovarian extracts (Engelke et al., 1980; Pelham and Brown, 1980). TFIIIA was purified as a 40-kDa protein that was specifically required for accurate transcription of 5S RNA genes by RNA polymerase III (Engelke et al., 1980). It binds specifically to the internal control region (ICR) of the 5S RNA gene, a region approximately 50 base pairs long that begins about 45 base pairs downstream of the transcription initiation site (Engelke et al., 1980; Sakonju and Brown, 1981). TFIIIA alone binds reversibly to the ICR and once bound, it promotes the subsequent formation of a complex involving other transcription factors, TFIIIC and TFIIIB (Bogenhagen et al., 1982; Lasser et al., 1983). TFIIIC does not specifically bind to the ICR itself, but binds to the TFIIIA-5S RNA gene complex and stabilizes it. TFIIIB then binds to complete the preinitiation complex and RNA polymer ase is then recruited to initiate transcription (Bieker et al., 1985). The transcription complex remains bound to the DNA through multiple passages of the RNA polymerase (Bogenhagen et al, 1982).

Because TFIIIA protected a 45-50 base pair region from DNase I digestion, yet was only 38-40-KDa in size, it was initially argued that more than one protein bound to the ICR (Engelke et al., 1980). However, a lack of dyad structure in the ICR coup led with low protein-to-gene molar ratios necessary for binding suggested that one or two protein molecules were sufficient for binding along the entire length of the ICR (Sakonju and Brown, 1982). Footprint titrations, transcription assays, and mixing e xperiments determined that TFIIIA binds to DNA with a binding stoichiometry of 1:1(Smith et al., 1984), suggesting that it has a very elongated structure in association with its binding site.

When the amino acid composition of TFIIIA was determined (Picard and Wegnez, 1979), the surprisingly frequent occurrence of cysteines led Hanas et al. (1983) to suspect that metal ions were involved in TFIIIA function (sulfhydryl groups are capable of metal coordination). They discovered that zinc was necessary for the transcription of 5S RNA genes by RNA polymerase III. It was also determined that preparations of TFIIIA treated with chelating agents could not bind to the 5S RNA gene, yet binding could be restored by the addition of zinc ions to the TFIIIA preparations (Hanas et al., 1983). The stability of 7S particles was also improved by excluding chelating agents (Miller et al. 1985). Atomic absorption spectroscopy analyses showed that prep arations of 7S particles contained 7-11 atoms of zinc per mole of particle (Klug and Rhodes, 1987). While is was clear that zinc was essential for TFIIIA function, the actual role of zinc in DNA-binding was not known.

It was not until the amino acid sequence of TFIIIA was deduced from a cDNA clone that the basis of TFIIIA's zinc-dependence was determined (Miller et al., 1985). Analysis of the amino acid sequence of TFIIIA revealed nine tandem similar units (Figu re 1), each containing about 30 amino acids, including two cysteines and two histidines located at invariant positions within each unit. These four residues were proposed to be zinc-ligands which interact with zinc to form a tetrahedal complex necessary for the proper tertiary structure of the minidomain, termed a "zinc finger." This interpretation was supported by proteolysis experiments, where trypsin digests of TFIIIA yielded fragments that differed in size by about 3 kDa (Miller et al, 1985). A ser ies of individual zinc coordinating domains, each containing about 30 amino acids, would then account for both the zinc content of the protein and a periodic arrangement of 3-kDa fragments.

Further support for the importance of the zinc ligands was forthcoming. An EXAFS (extended X-ray absorption fine structure) study confirmed that approximately nine zinc atoms are contained within a 7S particle and the coordination sphere of the zi nc ions contains two cysteines and two histidines (Diakun, et al., 1986). A peptide, containing finger two of TFIIIA, was also purified and shown to fold properly in the presence of zinc (Frankel et al, 1987). Site-directed mutagenesis was used to repla ce the first conserved histidine with asparagine in each of the nine fingers (Del Rio et al., 1993). The mutant proteins

demonstrated localized regions of disruption associated with the region of mutation, a decreased affinity for the 5S RNA gene, and when compared wild type TFIIIA, allowed for localized regions of increased susceptibility to DNase I digestion in DNA-TFI IIA complexes. Zinc has also been determined to be essential for the proper folding and binding activity of transcription factor Sp1 (Kuwahura and Coleman, 1990).

Several other features of the amino acid sequence of TFIIIA are functionally relevant. The consensus sequence of the nine repeats is

Each repeat is joined by a linker region of 2-6 amino acids. The linker regions between fingers 1-2 and 2-3 contain a pentapeptide sequence, TGEKP/N, that is also found in Kruppel, Zif268, and many other zinc

Figure 2. Analysis of average distribution of amino acids along the length of each repeated unit. The height of the histogram corresponds to the number of times that class of amino acid occurs within a particular position along the repeated unit. Cl assification of amino acids are as follows: a) DNA-binding - Lys, Arg, His, Asn, Gln, Thr; b) Basic - Arg, Lys, His; c) Acidic - Asp, Glu; d) Hydrophobic - Leu, Ile, Val, Phe, Tyr, Trp. The four zinc-ligands (cysteines and histidines) are represented in each set of histograms. (Figure taken from Miller et al., 1985).

finger proteins (Schuh, et al., 1986). There is no obvious pattern among the linkers except that fingers 1,3,5,7,and 9 have more basic residues in

their C-terminal halves than fingers 2,4,6, and 8 (Churchill et al., 1990). Compared to other fingers, the linkers going into and out of the sixth

finger are unusually short. This suggests that finger six probably interacts with the DNA in a manner that is different from the other fingers.

Following the nine zinc finger domains is a C-terminal region that contains about 70 amino acids. The first half of this region is enriched

with basic residues (Figure 1). Miller et al. (1985) proposed that this region corresponds to the 10-kDa domain that is revealed at one end of TFIIIA by mild proteolysis (Smith et al., 1984). This domain can be removed by proteolysis without affectin g the affinity of DNA-binding or the pattern of protection seen with DNase I footprinting (Smith et al., 1984; Vrana et al., 1988). However, the C-terminal domain is required for efficient RNA synthesis. The 30-kDa fragment has only about 20% of the act ivity of the full-length 40-kDa fragment (Smith et al., 1984). This C-terminal region probably interacts with TFIIIC in the formation of the preinitiation complex (Lasser et al., 1983).

Two-dimensional NMR studies of peptides from ADR1 containing either one or two zinc fingers (Parraga et al., 1988), finger 31 from Xfin (Lee et al., 1989), finger 2 from yeast transcription factor SWI5 (Neuhaus et al., 1992) and the C-terminal fing er of human enhancer binding protein (Omichinski et al., 1990) essentially confirmed this model (Figure 3). Specifically, the N-terminal region of the finger forms a two-stranded beta sheet, containing both metal-binding cysteine residues, that terminat es, bends, and gives rise to an alpha helix containing the metal-binding histidine residues. The two cysteine residues and two histidine residues coordinate the zinc ion and the conserved hydrophobic

Figure 3. Schematic illustration of the tertiary structure of a zinc finger domain. Interactions between zinc-ligands and conserved hydrophobic residues are shown. (Figure taken from Krizek et al., 1991).

residues do indeed form a core, which further stabilizes the tertiary structure of the finger. The linker residues connect the adjacent fingers. While all zinc fingers characterized share this common structure, finger one from SWI5 has an additional beta strand N-terminal to the formal start of the finger motif (Neuhaus et al., 1992). This third strand plays a role in stabilizing the structure of this finger.

TFIIIA binds specifically to an intragenic region, called the internal control region (ICR), of the 5S RNA gene and protects nucleotide positions 45-96 (where 1 denotes the transcription start site) from DNase I digestion (Pelham and Brown, 198 0; Engelke et al., 1980). A series of deletion mutants initially defined residues 50-83 to be required for transcription initiation (Bogenhagen et al., 1980). Methylation interference experiments showed that the main DNA contact points with TFIIIA occur at the extreme 3' end of the ICR, where six critical G residues on the noncoding strand were identified between positions 80-90 (Sakonju and Brown, 1982). These data suggest TFIIIA interacts with almost every nucleotide on the noncoding strand from regi on 80-90. A series of point mutations were introduced into the 5S RNA gene to determine their effect on transcription initiation (Pieler et al., 1985a; Pieler et al., 1985b). As a result, the ICR was defined to include residues 45-97 and discovered to b e tripartite in structure. The three regions are defined as box A (54-60), the intermediate element (67-72) and box C (80-97). Box A is a common, conserved class III promoter domain, while the intermediate element and box C are 5S-gene-specific. The sp acer regions between these elements are quite tolerant of sequence alterations (Pieler et al., 1985b) and are variably sensitive to insertions and deletions (Pieler et al., 1985b; Pieler et al., 1987).

A series of 5S RNA gene point mutants were also analyzed in terms of their effect on TFIIIA binding affinity (You et al., 1991; Veldhoen et al., 1994). In agreement with the data derived from methylation interference experiments (Sakonju and Brown , 1982), extensive protein-DNA contacts were detected in the C box. Substitution of GC base pairs at positions 81, 85, 86, 89, and 91 resulted in three to fivefold reductions in TFIIIA binding affinity (Veldhoen et al., 1994). Other bases within the C bo x (residues 80-91) were also important in binding. Within the intermediate element, a significant reduction in TFIIIA affinity was seen when positions 70 and 71 were mutated, but no reduction in affinity was seen when positions 67-69 were mutated. (Veldh oen et al., 1994). Substitutions of sequences 57-62 in box A resulted in a three to fourfold reduction in binding (You et al., 1991).

The 5S RNA gene also has the ability to bind zinc in a sequence-specific fashion (Martinez-Balbas, et al., 1995). The strongest zinc-binding sites correspond to GGG trinucleotide repeats, although the strength of binding depends on flanking sequen ces. It is interesting to note that residues determined to be important for TFIIIA binding (79-82 and 84-86 in the C box and 70-71 in the intermediate element) have a strong affinity for zinc. These data raise the intriguing possibility that the structu ral zinc ions contained within TFIIIA may also interact with the DNA.

Two models of TFIIIA-ICR interaction were initially proposed (Fairall et al., 1986). In model I, TFIIIA snakes around the DNA with the fingers following the helical path of the major groove (Figure 4a) (Smith et al.,

Figure 4. Various models representing the proposed interactions between TFIIIA and the 5S RNA gene. a) Two models originally proposed by Fairall et al. (1986). Model I shows the the fingers of TFIIIA wrapping around the DNA so that they follow the helical path of the major groove. Model II shows TFIIIA positioned on one side of the helix.

b) Modified version of model II, where every other linker residue (shown as dotted line) crosses the minor groove (Figure taken from Churchill et al., 1990). c) Representative of non-uniform model, where fingers 1-3 and 7-9 wrap around the DNA much like Zif268 (Pavletich and Pabo, 1991). In this particular model, finger 5 interacts with the major groove and fingers 4 and 6 span the minor groove (Figure taken from Clemens et al., 1992). d) Sites of DNase I footprint alterations associated with par ticular broken finger mutants (Figure taken from Del Rio et al., 1993).

1984; Berg, 1988). In this model, each finger would make structurally equivalent contacts with the major groove. In model II, TFIIIA lies on one face of the helix with successive fingers interacting with alternate sides of the DNA (Figure 4a) (Fairal l et al., 1986). In this model, only every other finger would make structurally equivalent contacts every 10 base pairs. Model II suggests the amino and carboxyl terminus of each finger

would exit from the same end of the finger, while model I suggests the amino and carboxyl termini would exit from opposite ends of the finger. The predicted three-dimensional structure of the zinc finger domain (Berg, 1988), that was later confirmed b y 2-D NMR studies, clearly supported model I. On the other hand, methylation and micrococcal nuclease protection studies showed alternating strong and weak sites of protection every five nucleotides (Fairall et al., 1986). DNase I footprinting also show ed that primarily one face of the DNA is protected (Rhodes, 1985). These data clearly supported model II. Hydroxyl radical footprinting further supported model II, as regions of protection every 10 base pairs were seen, along with asymmetric protection of the two strands (Churchill et al., 1990). Churchill et al. (1990) then proposed a model to reconcile the experimental data with the known structure of the zinc finger domain. In this model, a two finger periodicity was proposed where each finger sits on the same side of the helix, but after every second finger, the linker region would span the minor groove (Figure 4b). All of these models assume essentially equivalent interactions between each finger (or pairs of fingers) and the DNA, yet the tripar tite structure of the ICR, determined by mutagenic studies of the 5S RNA gene, did not support this assumption.

Several studies demonstrated a non-uniform interaction between the nine zinc fingers of TFIIIA and the ICR. By considering the length of the linkers connecting finger six to its adjacent fingers, Berg (1990) proposed that finger six is itself a li nker connecting two sets of fingers (1-5 and 7-9) that wrap along the major groove. Truncated versions of TFIIIA were used in hydroxyl-radical footprint experiments and further showed non-uniform interaction between TFIIIA and the ICR (Vrana, et al., 198 8; Hayes and Clemens, 1992). These experiments revealed two continuous regions of protection at both ends of the ICR, corresponding to roughly one turn of the helix. However, the middle 20 base pairs showed a more complex pattern of protection, where a short stretch of DNA is protected, but is flanked by two unprotected regions. These three clustered regions of protection were in good agreement with the tripartite structure of the ICR determined by Pieler et al. (1985a,b). Various further models were t hen proposed. While they differ slightly in their positioning of fingers 4-6, they all agree that fingers 1-3 and 7-9 wrap around the DNA, while fingers 4-6 span only one side, crossing the minor groove twice (Figure 4c). These non-uniform models were further supported by a reconsideration of the DNase I footprinting experiments (Fairall and Rhodes, 1992), missing nucleoside experiments (Hayes and Tullius, 1992) and further refined hydroxyl-radical footprinting experiments using truncated versions of T FIIIA (Clemens et al., 1992; Hayes and Clemens, 1992). To account for the 22 or so base pairs that interact with fingers 4-6, it has been noted that TFIIIA bends the DNA upon binding (Schroth et al., 1989, Schroth et al., 1991). Most of the experimental evidence used to propose this model of non-uniform interactions employed truncated versions of TFIIIA. Therefore, the above interpretations depend upon the assumption that truncated forms of TFIIIA interact with the 5S RNA gene essentially as they woul d in the context of the full-length protein.

Site-directed mutagenesis was used to generate mutant versions of TFIIIA containing structural disruptions of individual zinc fingers (Del Rio et al., 1993). DNase I footprinting analyses of these 'broken finger' mutants suggest that TFIIIA's mod e of interaction with the ICR may be even more complex than that which is proposed in the non-uniform model (Figure 4d). For example, while the pattern of footprint alterations associated with zinc-ligand mutations in finger 2 or 5 correspond to the pos itioning of these fingers in the non-uniform models (Clemens et al., 1992; Hayes and Clemens, 1992), the same does not hold true for fingers 6,7,8, and 9. The sites of loss-of-protection associated with these finger mutants were positioned further toward s the 5' end of the ICR than would be predicted by the non-uniform models. In fact, disruption of finger 8 was associated with two separate patches of loss-of-protection suggesting that finger 8 interacts with two sites on the ICR well separated along th e linear dimension of the DNA. These data do, however, show that TFIIIA does interact with the helix along the length of the ICR as a series of sequence-specific, localized contacts.

Cross-linking experiments have shown that of the 20 T residues that span a 61-bp fragment containing the ICR, >90% of the cross-linking occurs in two positions corresponding to residues 84 and 88 on the noncoding strand and coding strands, respe ctively (Lee et al., 1991). Amino acid analysis of the peptides cross-linked to the photoactive, noncoding strand indicates that finger two interacts with position 84 on the noncoding strand. This is in good agreement with the footprinting results obtai ned with broken finger mutants (Del Rio et al., 1993) and truncation mutants (Vrana et al., 1988).

Although zinc finger proteins are thought to be essentially modular, TFIIIA interacts with its binding site in a manner apparently different than the sum of its parts. An individual finger from TFIIIA does not give a clearly defined DNase I footpr int (Frankel et al., 1987) and is probably not sufficient for sequence-specific binding (Parraga et al., 1988). And while polypeptides containing fingers 1-3 demonstrate sequence-specific binding, this is not seen with polypeptides containing fingers 1-2 or 2-4 (Liao et al., 1991). Deletion mutants and broken finger mutants demonstrate that the binding energy contributed by each finger is not equal (Clemens et al., 1994; Del Rio et al., 1993). In fact, TFIIIA binding is not merely the additive effect of the nine separate fingers interacting with the DNA (Clemens et al., 1994; Del Rio et al., 1993).

A TFIIIA-DNA co-crystal structure has yet to be described; given the conserved aspects of the zinc finger motif, however, descriptions of TFIIIA-like protein-DNA crystal structures are likely to be relevant. To date, the crystallographic analyses of three such proteins have been published: Zif268 (Pavletich and Pabo, 1991), tramtrak (Fairall et. al; 1993), and GLI (Pavletich and Pabo, 1993).

Each finger makes primary contacts with a 3-bp subsite of the DNA through the N-terminal portion of the alpha helix (in every case, the peptide interacts with guanines on the same strand of DNA). While the

Figure 5. Crystal structure of the Zif268-DNA complex (From Pavletich and Pabo, 1991).

helix fits snugly into the major groove, it is tipped at a steeper angle than the angle of the major groove. The beta-sheet is oriented away from the major groove on the back of the alpha helix. The first strand of the beta sheet does not make any co ntacts with the DNA while the second strand contacts the sugar phosphate backbone along one strand of the DNA.

As a whole, the residues involved in making base contacts are found immediately preceding the start of the alpha helix (-1) and at the second(+2), third (+3), and sixth (+6) position of the alpha helix. Fingers 1 and 3 have identical residues at t hese positions (Arg, Asp, Glu, Arg) and interact with the same 3-bp subsite sequence (GCG). In both of these fingers, the arginines found at positions -1 and +6 make primary interactions with the GCG subsite (Figure 6a). The aspartic acid at position +2 (which is conserved in all three fingers) does not make direct contact with the DNA, but instead stabilizes the side chain of arginine at position -1 by forming hydrogen bond-salt bridges with the guanidinium group already hydrogen-bonded to the first G of each subsite. The second G in the subsite (at the 5' end) is hydrogen bonded to the arginine at position +6.

Finger 2 interacts with the DNA differently than fingers 1 and 3 (Figure 6b). It recognizes a TGG subsite and contains Arg, Asp, His, and Thr at positions -1, +2, +3, and +6 respectively. Like fingers 1 and 3, the arginine at position -1 hydrogen bonds with the first G of the subsite with stabilizing interactions from the aspartic acid at position +2. However, instead of glutamic acid, finger 2 contains histidine (not a zinc-ligand) at position +3. This histidine forms a hydrogen bond with the second G in

Figure 6. Schematic representations of amino acid-nucleotide contacts seen in the fingers of ZIF268, tramtrak, and GLI. Numbers represent position of amino acids with respect to the start of the alpha helix.

the subsite. The threonine at position +6 does not interact with the DNA.

In addition to the base contacts made by Zif268, several contacts are made with the sugar-phosphate backbone of the DNA. In each finger, the first histidine that contributes to the coordination of the zinc ion also hydrogen bonds to a phosphate on the primary strand of the DNA

Contacts between fingers were also observed. An arginine at the ninth position of the alpha helix of finger 1 forms a hydrogen bond with the backbone carbonyl of a serine at the -2 position of finger 2. The same interaction is seen in fingers 2 a nd 3, where an arginine at position 9 of the alpha helix of finger 2 interacts with the carbonyl of an alanine at position -2 of finger 3.

Several generalized features of the Zif268 complex are evident: 1. Each zinc finger constitutes an independent, modular unit which recognizes a distinct 3 bp contact site; 2. Most contacts are made with a single strand of the DNA; 3. The contacts i nvolve basic or polar amino acids that are found at specific positions in the alpha helix, namely, positions -1, and either +3 or +6. These features, in turn, have been suggested as part of a generalized recognition code between TFIIIA-like proteins and their binding sites (Berg, 1992; Jacobs, 1992). However, the analyses of two other crystalized zinc finger-DNA complexes indicate it is unlikely there is any simple, general code describing zinc finger-DNA interactions.

Tramtrack (TTK) is a transcriptional regulator of the Drosophila development gene fushi-tarazu. A 66-residue portion of this protein's DNA-binding domain, containing two zinc finger motifs, was crystallized in the presence of an 18-b p oligonucleotide and the complex was solved to 2.8 A (Fairall et. al; 1993). As with Zif268, the zinc fingers function as independent modular units that recognize a specific 3-bp sequence, but unlike Zif268, the binding sites overlap slightly. The fing ers fit into the major groove and make contact through amino acids in the N-terminal region of the alpha helix.

The structure of finger 2, and its interaction with the DNA, is essentially the same as that seen in Zif268 (Figure 6d). An arginine at position -1 (relative to the start of the alpha helix) hydrogen bonds with the second G of the 5'AGG3' subsite. An asparagine at position +3 makes contact with the 5' A. Like Zif268, there is an aspartic acid at position +2, but instead of merely stabilizing the arginine interaction, this residue actually hydrogen bonds with the C on the opposite strand of DNA t hat is base-paired to the most 3' G in the subsite (thus, finger 2 of TTK contacts three bases compared to two contacted by Zif268). Phosphate contacts are also similar to those seen in Zif268 and involve the first conserved histidine and a lysine residu e on the second beta-sheet (two residues downstream of the second conserved cysteine). The other phosphate contact involves a lysine in the alpha helix at position +1.

While finger 2 is very similar to the fingers of Zif268, finger 1 showed several unanticipated features. The residues N-terminal to the conventional finger motif fold to form a third strand to the beta-sheet. This feature has been previously seen in the first finger of protein SWI5 (Neuhaus et al., 1992). While these residues do not interact with the DNA, they are required for DNA-binding. The zinc-histidine-phosphate contact observed in all three fingers of Zif268 and finger 2 of TTK was not s een in finger 1. This was surprising as this interaction was expected to be widely conserved among zinc finger-DNA complexes since it involved the invariant histidine. In finger 1 of TTK, this interaction is substituted by a phosphate contact with tyros ine at position -3 (relative to the start of the alpha helix). Unlike the three fingers of Zif268 and finger 2 of TTK, there are many other phosphate contacts with residues in the second beta strand and turn region between this second beta strand and the alpha helix (arginine at position -5, threonine at position

-2, and histidine at position-1, all relative to the start of the alpha helix). The numerous contacts with the phosphate backbone are probably due to a significant bend in the DNA.

The recognition code employed by Zif268 and finger 2 of TTK does not strictly apply to finger 1. At position -1, instead of a base contact, there is a backbone contact. As with all fingers noted thus far, a 3-bp subsite (5'GAT3') is contacted. A n arginine at position +6 hydrogen bonds to the 5' G, an asparagine at position +3 hydrogen bonds to the middle A, and a serine at position +2 hydrogen bonds to the 3' T (Figure 6c).

The third protein that has been crystallized and complexed to DNA is the product of the human GLI oncogene, a DNA-binding protein that is amplified in human glioblastomas. Five zinc fingers of GLI were bound to a 21-bp DNA fragment obtained by in vitro selection from genomic DNA (Pavletich and Pabo, 1993). Since it is not known whether this DNA fragment contains the biologically relevant binding site, caution is required when interpreting these results. The complex was solved to 2.6 A.

There are several features of GLI structure and its interaction with DNA that are quite different from Zif268 and tramtrak. The first zinc finger does not contact the DNA, but instead makes extensive protein-protein contacts with finger 2. In ad dition, fingers 1,2 and 3 together make only one base contact (a tyrosine at position +2 in finger 2). Given that this complex may not contain the biologically relevant binding site, the anomalous features of the first three fingers may not be all that m eaningful.

Fingers 4 and 5 do make extensive base contacts in a 9-bp region, more so than those seen in either Zif268 or TTK. As with Zif268 and TTK, the contacts predominate in the N-terminal portion of the alpha helices of fingers 4 and 5. Finger 4 has fo ur residues that make contact with 5 bases (Figure 6e). An alanine at the position that corresponds to the first position (+1) of the alpha helix makes a van der Waals contact with the methyl group of T (the most 5' base in the recognition sequence). A serine at the second position (+2) hydrogen bonds with a T that is immediately 3' to the first T that interacts with Ala. The same serine also makes van der Waals contact with a G that is immediately 3' to the second T. This is the first example of a si ngle amino acid residue contacting more than one nucleotide. An aspartic acid at the third position (+3) hydrogen bonds to a C that is on the other strand of DNA. Finally, a lysine that corresponds to the sixth position in the alpha helix hydrogen bonds to a G immediately 3' to the third G in the recognition sequence (it is also base-paired to the C recognized by the aspartic acid).

Finger 5 involves five residues that make contacts with five bases (Figure 6f). The first residue, aspartic acid, is found at the position immediately preceding the alpha helix (position -1). It accepts hydrogen bonds from two adjacent cytosines . The next residue is a serine that corresponds to the second position (+2) of the alpha helix. It hydrogen bonds with a G that is base-paired with the most 5' C bound to aspartic acid. A serine in the third position (+3) hydrogen bonds to an A that is 5' to C. An arginine at the fifth position (+5) in the alpha helix hydrogen bonds to the same G bonded to the serine in the second position. Finally, a lysine at position +6 hydrogen bonds with a G that is 5' of the A (bond to the serine in the third p osition).

While the residues that make contact in fingers 4 and 5 show a correlation with the base contact positions in Zif268, many of these contacts could not have been predicted from a knowledge of Zif268 contacts. Furthermore, while Zif268 recognizes ba ses on only one strand of the DNA, GLI recognizes bases from both strands, where the bases recognized by finger 4 of GLI are found on the opposite strand from the bases recognized by finger 5. Figure 7 summarizes the base contact positions in Zif268, TTK , and GLI demonstrating that previous attempts to assign base contact residues to positions -1, +3, and +6 (Berg, 1992; Jacobs, 1992) are too simplistic.

In addition to base contacts, several residues in fingers 4 and 5 make contact with the phosphodiester oxygens. In finger 4, such contacts are made with residues which correspond to the following positions (relative to the start of the alpha helix ): positions +4, +7, +11, and +16. In finger 5,

residues at positions -3, -2, and +5 make contact with phosphate groups.

TFIIIA has been most extensively analyzed in Xenopus laevis. However, it has also been sequenced and partially characterized in four other species of frogs (Xenopus borealis, Rana pipiens, Rana catesbeiana, and Bufo americanus), human, and Saccharomyces cerevisiae. While all proteins are composed of an N-terminal region that

Among frogs, biochemical assays and sequence analysis indicate divergence between genera is greater than that which has occurred within genera. Immunoblot assays revealed that TFIIIA from both R. catesbeiana and R. pipiens reacted with anti-X.laevis TFIIIA antisera, but to a considerably lesser extent than TFIIIA from X. laevis and X. borealis (Gaskins et al., 1989). DNase I- protection analyses demonstrated that the Rana TFIIIAs imparted less protect ion of the 5' half of the Xenopus ICR than either Xenopus TFIIIA. In vitro transcription assays showed that both Rana TFIIIAs were less efficient at promoting transcription of a Xenopus 5S RNA gene than was either of the Xenopus TFIIIAs (Gaskins et al., 1989). The cDNAs for TFIIIA from Xenopus borealis and Rana catesbeiana were cloned and sequenced (Gaskins and Hanas, 1990). X. borealis TFIIIA is composed of 339 amino acids and shar es 84% sequence homology with X. laevis., while R. catesbeiana TFIIIA is composed of 335 amino acids and exhibits 63% sequence homology with X. laevis. The C-terminal halves of both proteins contain more non-conservative changes (wit h respect to the X. laevis sequence) than the N-terminal halves, with the most conserved region consisting of the first two zinc fingers and the most divergent region contained in the C-terminal domain. The Rana TFIIIA contains an insertio n of 4 amino acids between the 6th and 7th finger not seen in either Xenopus TFIIIA.

The cDNA clones of TFIIIA were later obtained from both Bufo americanus and Rana pipiens (Gaskins et al., 1992). Like X. borealis and R. catesbeiana, each gene contained nine tandem repeats corresponding to the C2H2 zinc finger motif seen in Xenopus laevis. In both cases, the zinc finger domains were found in the N-terminal thre e-fourths of the protein and both contained a non-finger C-terminal domain. TFIIIA from B. americanus contains 339 amino acids with 136 changes relative to X. laevis TFIIIA, resulting in a homology of 61%. TFIIIA from R. pipiens con tains 335 amino acids with 131 changes relative to X. laevis TFIIIA, resulting in a homology of 62%. R. pipiens TFIIIA also has a 13 amino acid deletion in the C-terminal domain, while B. americanus has single amino acid deletions in the N-terminal region and the seventh zinc finger, along with a three amino acid deletion in the C-terminal domain. Among X.laevis, B. americanus, and R. pipiens, finger one shows the highest overall conservation and the C-terminal domain has a highly conserved overall positive charge. In addition, there is also significant sequence conservation between a region comprising the last 5 amino acids of finger nine and the first 5-6 amino acids of the C-terminal domain. Footprinting st udies also showed that B. americanus TFIIIA protects the same region of the X. borealis somatic-type 5S gene ICR from DNase I digestion as does X. laevis TFIIIA (Gaskins et al., 1992). The only difference between the two is the loss of a hypersensitive site (position +43) seen with X. laevis TFIIIA-binding but not seen with B. americanus TFIIIA-binding. R. pipiens TFIIIA gives strong protection from nucleotide +96 to +78, but weaker protection from nucleotide +7 8 to +43 (corresponding to weaker protection imparted by the C-terminal fingers).

A clone isolated from a human brain cDNA library (Drew et al., 1995) possesses a deduced amino acid sequence with 57% homology to X. laevis TFIIIA and 59% homology to R. pipiens TFIIIA. The gene product encodes 363 amino acids that contain nine C2H2-type zinc finger motifs which, like Xenopus TFIIIA, include a linker region preceding the second and third fingers containing the TGEK/RP amino acid element. These data strongly suggest this is the human TFIIIA gene, but the gene product of this clone has not yet been expressed and characterized. The greatest homology between human and frog sequen ce is found in fingers 8 and 9.

TFIIIA is a rapidly evolving zinc finger protein. While human and frog TFIIIA show only 57-59% homology, significantly less divergence is seen in another zinc finger protein, erythroid transcription factor (Eryf 1) (Trainor et al., 1990). Mouse a nd human Eryf 1 demonstrate only 2 conservative substitutions among the 108 amino acids that comprise the two C2C2 zinc f ingers of these protein. Human and chicken Eryf 1 have just three conservative and five non-conservative substitutions. The rapid rate of TFIIIA evolution is surprising given that the ICR sequence remains largely conserved among species. For example, w hile there are 131 amino acid differences between R. pipiens and X. laevis TFIIIA (resulting in 62% homology), there are only 7 nucleotide substitutions in the 5S RNA gene from nucleotides +40 to +100. TFIIIA's rapid rate of evolution sugg ests it might serve as a good molecular clock for determining phylogenetic relationships between species (Nei, 1987).

A sequence analysis of 221 C2H2 zinc finger proteins containing 1340 zinc fingers shows that fingers in the same p rotein tend to be more similar to each other than fingers from other proteins (Jacobs, 1992). This suggests that the primary mode of zinc finger evolution involves internal duplication. This view is supported by the gene structure of Xenopus TFII IA, where introns exist between each of the domains corresponding to the first seven fingers (Tso et al., 1986).

Alanine-scanning site-directed mutagenesis was used to determine the residues of ADR1 (a yeast transcription factor) that are important for DNA-binding (Thukral et al., 1991). ADR1 possesses two zinc fingers of the TFIIIA type (C2H2). The results were largely consistent with the mode of binding seen in Sp1 and Zif268. When Arg at position -1 in both fingers was mu tated to alanine there was a significant reduction in affinity, although the loss of the Arg in finger two had a more severe effect. In addition , His at position +3 and Arg at position +6 in finger one were determined to be important for binding. Finge r two also has Arg at position +6, but it contributes less binding energy than the Arg at the same position in finger one. An Asp in position +2 of finger two is also important for binding and probably functions (like Zif268) in stabilizing the sidechain of Arg at position -1. A more extensive mutagenic analysis of ADR1 showed that positions -1, +3 and +6 in finger one would not tolerate several other amino acids in addition to alanine (Thukral et al., 1992).

The differences in the contribution to the free energy of DNA binding by individual zinc fingers seen in the broken finger analysis was supported by a series of "finger-swap" experiments. In order to identify amino acid residues of TFIIIA that are important for DNA-binding, a series of finger-swap, scanning, and single amino acid substitution mutants were constructed and binding activities were determined using a nitrocellulose filter-binding assay (Zang et al., 1995). Finger-swap mutants were co nstructed by subcloning cDNA sequences from either finger 31 of Xfin or finger 3 of p43 into the Xenopus TFIIIA gene so that a chimeric TFIIIA molecule was produced. Both Xfin and p43 exhibit RNA-binding activity with little or no DNA-binding acti vity ( Andreazzoli et al., 1993; Darby and Joho, 1992; Joho et al., 1990). Both contain C2H2 zinc fingers and, apart fro m the conserved zinc-ligands and hydrophobic residues, the amino acid sequences are vastly different from the TFIIIA zinc fingers. The finger-swap mutants showed that substitution of finger 1 or 7 had little or no effect on the interaction of TFIIIA with its binding site. In contrast, substitutions of fingers 2, 3, and 4-6 significantly reduced the chimeric TFIIIA's affinity for its binding site. Substitution of finger 3 with the p43 finger led to a 100-fold reduction in its DNA-binding activity. Fing er 3 was previously shown to make a large contribution to TFIIIA's binding energy through analysis of 'broken finger' mutants (Del Rio et al., 1993).