Structure of the TFIIIA-5 S DNA complex

I.

Afol.

Biol.

(1992)

227,

407.-417

Structure of the TFIIIA-5 Jeffrey Laboratory IVational

Institutes

S DNA Complex

J. Hayes

of Molecular

Embryology

of Child Health and Human Development

Bethesda, MD 20892. 1T.B.A.

and Thomas Department

D. Tullius of Gh,emi.stry

The Johns Hopkins I!niversity Baltimore, MD 21218, U.S.A. (Received 25 January

1992, accepted 18 May 1992)

The missing-nucleoside experiment, a recently developed approach for determining the positions along a DNA molecule that make energetically important contacts with protein, has been used to investigate the structure of the complex of transcription factor TIIA with a somatic 5 S RNA gene from Xenopus borealis. We detect three distinct regions of the 5 S promoter that are contacted by TFIIIA, corresponding t’o the A-box, intermediate element and C-box regions previously identified by mutagenesis experiments. The advantage of the missing-nucleoside experiment over mutagenesis is that additional information, directly related to the structure of the complex, is obtained. Of most importance is that contacts to each strand of DNA are determined independently. and can be assigned unambiguously as interactions with TFTTIA. Throughout the binding site the strongest contacts are made with the non-coding strand of the 5 S gene. The two groups of contacts at either end of the binding site (boxes A and C) are comprised of sets of approximately ten contiguous nucleosides for which the contacts are reflected, without stagger, from one strand to the other. In contrast. contacts in the center of the promoter (the intermediate element) are staggered about five base-pairs in the 5’ direction with respect to each strand. These results. when analyzed in conjunction with the hydroxyl-radical footprint of the complex, support’ a, model in which TFIIIA wraps around the DR;A in the major groove of the helix for one turn at the two ends of the complex in boxes A and C, and lies on one side of the DNA helix in the rent,er of the complex at the intermediate element.

Keywords: DNA-protein

complexes; transcription fact,or IIIA; experiment: zinc finger

1. Introduction The isolation of transcription factor IIIA (TFIIIAT), by virtue of its ability to complement 5 S gene transcription in an otherwise incompetent activated egg extract from Xenopus (Engelke et aE., 1980), made possible the first physical studies of a purified eukaryotic transcription factor. DNase I footprinting showed that the protein binds to a site 50 base-pairs (bp) in length that comprises an intrai Abbreviations used: TFIIIA, transcription factor IDA; bp, base-pair(s); ICR, intragenic control region; n.m.r., nuclear magnetic resonance; IE, intermediate element: borealis

Xlo, Xevwpus somatic.

laevis

oocyte;

Xbs,

Xenopus

missing-nucleoside

genie control region (ICR) in the 5 S gene (Engelke et aE., 1980). Since a single molecule of TFIITA appeared to be too small to cover such a large stretch of DNA, it was thought likely that more than one molecule of the protein binds to the ICR. That TFITIA is a protein with an unusual structure began to emerge from the results of careful titration experiments, which showed that TFITIA in fact binds to the ICR in a 1 : 1 molar ratio (Smith et aZ., 1984). Further, it was found that a 30,000 M, amino-terminal proteolytic fragment of TFIIIA is able to protect the entire 50 bp binding site in the footprinting assay. These observations led to the proposal that TFTIIA is highly elongated when bound to DNA, with a diameter calculated to be

407 0022%2836j92/

180407

I 1 $08.0010

0

1992 Academic

Press Limihd

408

J. ,I. NaIyes and 7’. Il. Tullius -

about, half that of DNA (Smith et aZ., 1984). Biophysical studies of the protein provided additional support for the view that TFIIIA exists in an extended conformation (Bieker & Roeder, 1984). The finding that Zn2+ was absolutely required for TFIIIA to bind to DNA (Hanas et al.. 1983), and the subsequent determination of the complete sequence of the protein (Ginsberg et al., 1984), led to the recognition within the TFIITA protein of what has come to be known as the zinc finger motif (Brown et al., 1985; Miller et aE., 1985). Nine of these motifs, repeating in tandem, constitute almost all of the DNA-binding portion of TFIIIA. Zinc fingers have since been identified in the sequences of a large number of proteins, of which TFIIIA is now the prototypical example (Berg, 1988). A detailed model for the structure of the zinc finger was proposed, based on sequence homologies between segments of the putative Zn2+-binding domains, and metal-binding sites in other proteins for which crystal structures were known (Berg, 1988). The proposed structure is a compact domain with a hydrophobic core, made up of an antiparallel beta sheet connected to an alpha helix, with pairs of highly conserved histidine and cysteine residues bound to zinc. The main features of this model were soon supported by nuclear magnetic resonance (n.m.r.) spectroscopic analysis of single, synthetically derived, zinc fingers (Parraga et al.: 1988; Lee et al., 1989). A recent paper reports the X-ray crystal structure of the complex of the threefinger protein Zif268 with DNA (Pavletich & Pabo, 1991). The architecture of each of the Zif fingers closely resembles the original model and t’he n.m.rderived structures. Achieving the structure of a single zinc finger at atomic resolution still does not solve the problem of how a protein that contains several fingers (9 in the case of TFIIIA) might bind to DNA. Two general models for the structure of the TFITIA-DNA complex have been considered (Fairall et al., 1986). One model proposes that the protein binds mainly along one side of the DNA, parallel to the helix axis (Miller et al., 1985). The nine zinc fingers would contact successive parts of the major groove of the DNA on alternate sides of the helix (Miller et al., 1985; Fairall et al., 1986). Methylation-protection and micrococcal-nuclease footprinting experiments have been interpreted as supporting this model (Fairall et al., 1986). The second model suggests that the zinc fingers of TFIIIA are connected in an end-to-end fashion, like sausages in a link (Fairall et al., 1986). The protein would wrap around the DNA helix, with the nine fingers following the major groove (Smith et al., 1984; Fairall et al., 1986). This model closely resembles the mode of binding to DNA of the three fingers of the Zif268 protein that was found in the X-ray cocrystal structure (Pavletich & Pabo, 1991). A third model for the TFIIIA-DNA complex (Churchill et aE., 1990) incorporates aspects of the two previously mentioned. This model proposes that two adjacent fingers constitute the repeating

unit of TFIIIA. Pairs of tingers, arranged head-t,otail, would bind in the major groove of I)NA ‘l’hta segment of polypeptide that links two pairs of’ fingers is suggested to cross over the adjacent minor groove of the DNA. TFIIIA would lie primarily along one side of the helix, in a way similar t.o thr proposal of the first model. These three models have in common t,he notion that’ all nine fingers (or at least pairs of fingers) make equivalent contacts with the DNA. Other experiments have provided evidence that the strucbture might not be so uniform. The hydroxyl-radical footprint, of TFIIIA reveals that t’he DNA backbone at the two ends of the binding site is prot,ected from cleavage in a roughly even and continuous manner for more than one turn of the helix (Churchill. 1987; Vrana et d., 1988: Hayes et al., 1989). ln c+ont,rast. the footprint at t,he center of the binding site rnore closely resembles the pattern observed in earlier footprinting experiments with bacteriophage j, repressor (Tullius & Dombroski, 1986). A short stretch of DNA (at positions + 67 on the nor)-c*odinp and +64 on the coding strand) that is strongly protected is flanked by two regions, each a half-turn away. that are exposed to cleavage. Such a footprint is characteristic of protein bound to one side of DNA, as is 1 repressor. Analysis of the hydroxvlradical footprint led t’o the proposal of a tripartite st’ructure for the TFIIIA-DNA complex, in which the protein surrounds the DNA at the ends of t,ht, ICR, while at the center the protein lies along ant’ side of the helix (Churchill, 1987). Additional support, for a non-uniform st,ruct,ure came from a modeling study (Berg, 1990) in which a key observation was that* there are fewer amino acid residues csonnecting t’he sixth iinger to its adjacent fingers than is found for the other fingers. These short “linkers” were proposed to interfere with the ability of t,he fingers t,o follow one another in binding continuously along the major groovy. Instead. the sixth finger would itself serve as a link between set,s of fingers wrapping around t’hr I)SA at the two ends of the complex. Fingers 1 to 5 NIP 7 to 9 were proposed to constitut#e the two scl s ot’ tingflrs that wrap along the rnajor groove’ (much as was found for the three Zif268 fingers in the c*ocarystal structure (Pavletich &, Pabo. 1991)). Hecaausr it was thought that) the distanc*c between the two srts of fingers wrapped around the DKA c~ulti not bta spanned by a single finger if the 1)NA wt‘rt‘ st)raight. it was proposed that a bend occurs at the carnt’er of the ICR in the TFIIIA-DNA complex. lC:arlirl experiments had suggested that binding of TFTIIA bends DNA (Schroth pf al.. 1989; Bazett-*Jones & Brown, 1989). Another aspect, of this model is that finger 6 crosses over the minor groove. accounting for t,he repressor-like hydroxyl-radical footprint found in this region. A knowledge of the nucleotides in the I(IR that make energetically important contacts with TFIITA would help to distinguish among the various models for the structure of the complex. The DNA e1ement.s that comprise the 5 S promoter have been investjim

TFIIIA-5

S RNA Gene Complex

gated in extensive DNA mutagenesis experiments (Pieler et al., 1987). Three crucial sequence elements were identified in this fashion: the A-box, at the 5’ end of the ICR; the C-box, at the 3’ end; and an intermediate element (IE) between the A and C-boxes. Most of these mutations were evaluated by their effect on transcription; only a small fraction have been tested directly for their effect on TFIIIA binding. These experiments also were complicated by the fact that a change of sequence at a position contacted by TFIIIA does not always reduce the binding affinity of the protein for all possible substitutions (Pieler et al., 1987). More recently, a seriesof linker-scanning mutants of the Xenopus laevis oocyte (X10) 5 S gene has been tested directly for TFIIIA binding in vitro (You et al., 1991). These results. while of somewhat lower resolution since sets of changed base-pairs and not point mutations were tested, are consistent with the results of t’he experiments we describe in this paper. However, a general limitation of mutagenesis is that the effect of a particular mutation on binding cannot! be independently ascribed to one base of a pair, as both must be changed simultaneously. We report here the use of a new approach to such questions. the missing-nucleoside experiment (Hayes & Tullius, 1989), in order to describe in finer detail the structure of the complex formed by TFIIIA with t’he 5 S RNA gene. In this experiment, a radiolabeled DNA molecule is treated with the hydroxyl radical to make random single-nucleoside gaps. Protein is added, free DNA is separated by native gel electrophoresis from DNA bound to protein, and DNA is excised from the gel bands. The DNA from each band is subjected to denaturing gel electrophoresis to determine which “missing nucleosides” interfered with the ability of protein to bind, and which were irrelevant. A key advantage of t’his experiment is that each member of a base-pair may be interrogated individually as to its effect on protein binding, since the base opposite the missingnucleoside remains in the DNA. Our results impose very specific structural constraints on the TFTJTA-5 S gene complex. We use these data to propose a model for the complex that differs in important

details

from

models

advanced

previously.

2. Materials and Methods (a)

DNA

mokcules

and

TFIIIA

Plasmid DNA was prepared as described (Tullius & Dombroski, 1986) and stored as a precipitate in ethanol, or as a frozen solution in TE buffer (10 m&i-Tris. HCl (pH %O), 1 mM-EDTA) at -20°C. A DNA molecule containing the Xenopus borealis somatic (Xbs) 5 S gene was obtained by digestion of plasmid pXbs201 (Bogenhagen et al.. 1980) with the restriction endonuclease HindHI. This restriction fragment was labeled with radioactive phosphorus. A 246 bp-long DNA fragment. radiolabeled at the Hind111 site located 50 bp upstream from the beginning of the 5 8 gene, was obtained by subsequent restriction at the BarnHI site located at position + 195. Chromatographically pure TFIIlA was the gift of Dr Alan P. Wolffe (National

409

Institutes of Health). TFIIIA was prepared as described (Smith et al., 1984).Protein wasaliquoted at a concentration of @5 mg/ml and stored mobility shift gel electrophoresis that the preparations of TFIIIA length protein.

(b)

Footprinting

at, -70°C. Analysis by and SDS-PAGE showed contained >90%, full

the TFIIIA-DNA

complex

Hydroxyl-radical footprinting was performed as previously described (Churchill, 1987), except for the addition of 2.5 mM-dithiothreitol to the solution just before addition of the protein. In brief, 20 x lo-l5 mol of radiolabeled DNA and 50 x 10-l’ to 100 x 1OV l5 mol of

TFIIIA

were mixed in buffer for 10 min at room t,empera-

ture in a volume of 35 ~1. Five ~1 of each of the three hydroxyl-radical footprinting reagents (iron(I1) EDTA, hydrogen peroxide and sodium ascorbate) were added as described (Tullius et al., 1987). The reaction mixture was allowed to incubate for 2 min before addition of a quenching agent. The DNA cleavage products were analyzed by denaturing gel electrophoresis using standard methods. (~3) Missing-nucleoside

asmy

This experiment was accomplished as described (Hayes & Tullius, 1989). Briefly, a 5-fold molar excess of TFIIIA was incubated at room temperature for 10 min with a radiolabeled 5 S DNA molecule that’ previously had been randomly gapped by reaction with the hydroxvl radical. Various amounts of specific unlabeled competitor DNA (linear pXbs 201) was then added and t,he mixture allowed to incubate for 2 min. Exposure of the TFTIIA-DNA complex to excess competitor DNA at room temperature was found to enhance t,he signal in the missing-nucleoside assay, probably due to the relatively slow dissociation rate of TFIIIA bound in the native complex (Hanas et al., 1983). The sample then was either loaded immediately onto a native polyacrylamide gel, or cooled on ice before loading, and then clectrophoresed. The native (mobility shift) gel was run as described (Wolffe, 1988). This gel consisted of 47” (w/v) acrylamide, @080/, (w/v) bisacrylamide, and 5% (v/v) glycerol! with 20 mM-Hepes (pH 8.3), @l mM-EDT,4 as bot,h the gel and running buffer. The gel was run at room temperature or at 4°C’ for hetween 1 and 2.5 h at 80 to 120 1’. After electrophoresis. the wet gel was exposed to X-ray film for 1 11. Segments of the gel containing free DNA and DNA hound to protein were identified b>- aut)oradiography. Bands containing DNA were excised and DNA was axt,racted from the gel slices. The gaps that occurred in bound and free DNA were determined by subjecting the extracted DNA to denaturing gel elect’rophoresis. The aut,oradiograph of the denaturing gel was scanned wit’h a Joyce-Loebl Chromoscan 3 densitometer. Software included with the instrument was used to obtain integrated intensities for each of the bands in the unbound lane. These peak integrals were corrected for background by subtracting from each integral the average value found for nucleotides out’side the binding site. The resulting corrected integrals were then plotted. Data from several experiments were used in the analysis. Missing nucleosides that’ are all derived from one of the DNA strands mav be compared directly for their effect on protein binding, since the relative values of the integrals are accurately represented in the plot. Strict quantitative comparison of the effects of nucleosides between strands is not, possible. however.

II0

.I. .I. Jla!ys

trtld 7’. 1). T~tlli~cs

12345678

c

,45

Figure 1. Autoradiograph of a DNA styurncing gel revealing the nuclrosidw in the 5 S gene promoter that art’ cwntact,ed by TFIITA. (a) Results from the missing-nuclrosidc. assay with the nowcwding strand of’ the 5 S RN.4 gene radiolabrlrd. T‘ane I cwntains the prnducats of a Maxarw(~ilbrrt guanim-specifics sequencing reaction; law 2. thtx lab&d I)XA fragment, after rract)ion with the hydroxgl radical (cwntrol): lane 4. same as 1 rxc,capt that the I)h-X fragmrnt was subjected to native (mobility shift) gel cktrophoresis beforr the sequewing gel: lanes 1 and 5. thr bound J)zJ;\ fraction from the niissing-nuclrositie assay: lanes 6 and 7. the ~~rrbound DNA frac,tion from the missing-nlrc,leoside assay; 1ant.s X and 9. 5 S DIVA c~leavrd by the hydroxyl radicsal in thr presence and absence. respwtively. of hound TFI1I.A: lanes IO and 11. DEase I footprint of TFTTIA. and c,ontrol caleavagr pabtern of 5 S I)NA. respr~tively. The samples showt~ in lanes 5 and 7 were cahallenged with cwmprtitor DNA just prior to loading on the mobility shift gel (SW Materials and Methods). (b) Results from the missing-nuc~leosidr assay- with the coding strand of the 5 S RSA gent’ radiolabeled. I,ane I wntains the products of a Maxam 4:ilbrrt guaninr-specific sequencing rrac:t,ion; lanes 2 and 8. t,he latwk~tl I)N\IA fragment was allowed to react with the hpdroxyl radical in the absence or presencr of TFTTTX. rt.spe’.tivt~ly: Iant. :3. tht satne as lane 2 except t)hat the DNA fragment was subjected to native (mobility shift) grl t~lec~trophorrsis hrfore the sequencing gel; lanes 4 and 5. the bound I)KA fraction from t,hr missing-rluc~leosidr assay: lanes 6 and i. the unbountl I)iKiA fraction from the missing-nurleoside assay. The missing-nucleosidr samples shown in lanes 5 atld 7 w~rr c*hallenged wit,h cwmpetitor DNA ,just prior to loading on t)he mobilit,y shift gel (we Materials and Methods). sinc.e it is difficult to put integrals for bands from difkent~ strands on t,hr same intensity scale. However. it is apparent from the primary data t’hat the contacts are in general stronger on the non-coding compared to t,he coding strand. as reflected in the plot shown in Fig. 3.

3. Results (a) i~issing-nucleoRide

analysis

of

the TFIIIA-5

S

gene complex The missing-nucleoside experiment was used to examine the energetically important contacts made by TFIIIA to both strands of a somatic 5 S RXA gene from Xenopus borealis. The pattern of DNA fragments that resulted from this experiment is shown in the autoradiographs in Figure 1. Densitometric scans of the lanes on these autoradio-

graphs art depicted in Figure 2. A graphical summary of the data. in which the approsimat~e strength of a contact is represented by the length of a vert~ical bar, is presented in Figure 3. The part of the TCR wit’h which TFTIIA makes energetically important contacts is slightly smaller than the region protect’ed in a footprinting experiment. Contacted nucleosidrs are detected on both strands of thr DNA between positions +50 and $92 (Fig. 3), while the hydroxyl-radical footprint ext’ends from about +40 to +9X (C’hurchill. 1987: Vrana et al.. 1988: Churchill et al., 1990). (h) TPI

1 IA

interacts most cstrongl:y with coding strand of the JCR

the trw-

The sensitivity of the missing-nucleoside assay (aan be adjusted by varying the amount of unlabeled

TFIIIA-5

S RNA

Gene

Complex

U

f4G

(b)

(0)

Figure 2. Densitometer scans of the missing-nucleoside analysis of TFIIIA binding to the 5 S ICR. (a) Scans of the shown in Fig. I(b) (coding strand). autoradiograph in Fig. l(a) (non-coding strand). (b) S cans of the autoradiograph C, scans of lanes containing DNA cleaved by the hydroxyl radical (control) (lanes 3 in Fig. l(a) and (b)); F, hydroxylradical footprints of TFIIIA (lanes 8 in Fig. l(a) and (b)); U, scans of lanes containing the unbound DPU’A fractions (from lanes 6 in Fig. l(a) and (b)); B, scans of lanes containing the bound DNA fractions (from lanes 5 in Fig. I(a) and (b)). DNA added to the protein-binding solution. When the strongest contacts are examined, by adding the least amount of competitor DNA, only the nucleosides from positions + 79 to +87 on the non-coding strand are observed to give contacted signals (data not shown, but seeFig. 1). These data agree with the results of previous ethylation and interference experiments, which methylation suggested that the strongest contacts of TFIIIA with DNA oc(~ur in the region +79 t,o +89, and mainly on the non-coding strand (Sakonju & Brown. 1982). When the scope of the missing-nucleoside assay is broadened to examine both strong and weak contacts, by adding more competitor DNA to the

competitor

binding mixture, interactions are detected throughout the ICR (Figs 1 and 2). While mutagenesis experiments also found an extended array of

contacts (Pieler ct al., 1987), such experiments are unable to discern which member of a base-pair is more important for binding. Densitometric analysis of the

autoradiographs

in

Figure

1 shows

that,

although contacts occur on both DXJA strands

throughout the binding site, contacts to t’he noncoding strand appear to be more important energet#ically than those to the coding strand (Fig. 3). Thus the earlier conclusion that the strongest contacts are made with the C-box on the non-coding strand, perhaps to facilitate the passage of polymerase on the coding strand, can now be extended to include the entire TFIIIA binding site. (c) TFIIIA

makes three

groups

of contac’ts

with

the

ICR Inspection

of Figure

3 reveals

that the conta&s

of

TFIIIA with 5 S DNA are grouped together at three locations in the ICR. These groups of contacts are coincident with the A-box, the C-box, and the intermediate element that had been identified previously because of the deleterious effect of mutations in these regions on transcription efficiency or templateexclusion ability (Pieler et al., 1987). Each of the three clusters of contacts identified in the missingnucleoside experiment is comprised of a set of signals from the two DNA strands, with t,he weaker

Figure 3. Summary of contacts made by TFIIIA with the int)ragenic control region of the 5 S gene. Integrals were determined for peaks in the scans of gel lanes containing the unbound DNA fraction from the missingnucleoside analysis of the TFIIIA-5 S DNA complex (such as in Fig. l(a), lane 6). Data from several experiments were pooled and used to construct the bar graph. See Materials and Methods for details of the analysis. The length of a bar indicates the approximate strength of the contact, at that position. The bars above the line correspond to data from the non-coding strand. and the bars below the line represent data from t,he coding strand. Thp locations of the A-box. the C-box. and the IE are marked.

contacts on t)he coding strand approximately mirroring the stronger contacts on the non-coding strand (Fig. 3). These groups of contacts will be referred to as the A-box. (*-box, and IE contacts. (d) The C-box Previous work had demonstrated that the 3’ end of the TCR is the primary locus of interaction of TFTTIA with DNA. Deletion of this region eliminates TFITIA binding, while some weak binding is retained if all but the 3’ end of the ICR is delet’ed (Sakonju & Brown, 1981). As mentioned above, most of the contacts to DNA made by TFIIIA that were detected by alkylation-interference experiments were found in the region +79 to +89 (Sakonju & Brown, 1982). Ethylation interference signals were detected at six of the eight phosphate groups between positions +79 and -1-87, and all were from the non-coding strand. Methylation interference was seen for the guanine residues which occur at +81, +82, +85, t86, +87, and +89 on the non-coding strand, and at +91 on the coding strand. The set of contacts in the C-box detected by the missing-nucleoside assay generally agrees with these results. Figure 3 shows that contacts occur from approximately + 78 to +92, with the strongest signals found between positions + 79 to + 86 on the non-coding strand. Weaker signals are detected at positions +80 to + 86 and + 89 to t91 on the coding strand, and at positions +89 to +92 on the non-coding strand. The contacts in this region span more than one turn of the helix. However, a closer inspection of the

missinp-nuc.lc.osi(~e data in t,he (‘-box suggest h t ha,t 5ome st.ruct ural details are present wit hitI this pattern. The nucleoside at posibion -t X8 on thc~ non coding strand appears not t)o br> contact~rtl 1)). protein. This result is consistent, wit,h the obst~rvation that, no phosphate (sontact is detected to either sitIt> of this nuclroside (Sakonju $ Brown. 1982). This non-contacted position separat’es two regions of’ contact. + 79 t,o + 86 and + 89 to +9%. ()I] the coding strand. three groups of’ signals (‘au btx discerned, centered over positions + Xl. + 81.5. and + 90. This pattern is suggestive of t’he presenc~t~ of at least threcb separate structural units of t hr. t)rot.cirl making cont,ac:t with t’he I)NA in t,he (‘-hex.

(e) 7'h~ =l-box The A-box. which was proposed to extend from position +50 to +61 in the ICR. was first identified because of its similarity to elements in the promoters of other class III genes. Mutational analysis showed that t’his region is important. t)o promoter function (Sakonju &, Brown. 1981; Pieler et nl., 1987). However. methylationand et)hylationinterference experiments did not detect, contac%s in t’his region (Sakonju & Brown. 1982). The missing-nucleoside analysis idrnt’ifies a stretch of contacts which corresponds welt with the region identified as the L-box (Fig. 3). These contacts arc’ observed to extend uninterrupted for more than one turn of the helix, from position + 50 to + 62. A similar pattern of contact,s is det,rcted on both strands. The patterns from t’he two st,rands are reflected almost perfectly across the base-pairs. with no stagger in the pattern from one strand to the other. It has been suggested (Pieler et al.. 1987) that the 3’ border of the A-box in the 5 S gene should be extended to include position +64. The rnissingnucleoside analysis shows that indeed position + 62 is contacted by protein. but only on the non-coding strand. However, positions -i-63 and + 64 do not’ yield a contacted signal but’, along with position $65, comprise a sparer region between t’he A-box and intermediate element that is devoid of (*ontact by protein to either strand of DNA.

(e) The intermrdiatr elrment Interference experiments provided the first evidence for i mport,ant int,eractions between TFIIIA and the phosphodiester backbone in t.he center of t’he ICR. Ethylation of the phosphate groups between positions + 69 to + 71 interfered with binding, as did methylation at positions +70 and +71 (Sakonju & Brown, 1982). Mutagenesis experiments led to the suggestion that’ an intermediate element (IE), important to transcription, encompasses positions +67 to + 72 (Pieler et al., 1987). For example, transcription from a template mutated at position + 70 was reduced to 20 to 30% of the value for wild-type. This mutation nearly

TFIIIA-5

S RNA

abolishes TFIIIA binding, as assayed by DNase I footprinting (Pieler et al., 1987). Nine base-pairs at the center of the ICR, +67 to +75 (Fig. 3), give signals in the missing-nucleoside experiment. The 5’ border of this set of contacts agrees wit,h the 5’ edge of the IE as determined by mutagenesis (Pieler et al., 1987). In fact, a mutation at +66 was found to cause a small (30%) reduction in binding, which corresponds well with the weak contact signal we observe at this position (Fig. 3). Because of the missing-nucleoside signals at +74 and + 75 on the coding strand, we propose that the 3’ border of the IE extends beyond that defined by mutational analysis. Pieler et al. (1987) used the observation that’ a 5 F gene mutated at position + 73 was found to be transcribed at 80% of the efficiency of wild-type t,o demarcate the 3’ limit of the IE. The missing-nucleoside data agrees with this result, since only a small signal is detected at this position. However, a template mutated at + 74 was found to support, efficient transcription, in apparent conflict with the missing-nucleoside results, whirh show a moderate signal on the coding strand. The two assays measure different things (transcription and binding), perhaps accounting foe this discrepancy. As defined by the missing-nucleosidr experiment. the IE is similar in length to the A and C-boxes (Fig. 3). The IE is unlike these ot,her regions. however, in the way contacts on t’he two complement)ary strands are related. Instead of being mirrored on each strand as in the A and C-boxes, t,he strongest csontacts on the two strands in the IE are staggered from each other, centered at’ +69 on the non-coding strand a’nd at +74.5 on the coding st)rand (Fig. 3).

4. Discussion (a) Proposed structure for the TFIIIA-DNA complex The missing-nucleoside analysis reveals that the TFIIIA-DNA complex is not comprised of nine approximately identical. repeating protein subunits bound equivalently to DNA (Fairall d al., 1986), but instead supports a tripartite structure for the complex (Churchill, 1987: Vrana et al.. 1988; Berg, 1990). Three discrete groups of contacts are found, corresponding to the A and C-boxes and the intermediate element previously noted in mutagenesis experiments (Pieler et al., 1987). In all three regions set,s of contacts extend for about one turn of the helix, and the strongest contacts are made with the non-coding strand. However, there is a key difference between the contacts made with the middle of the ICR and those with the ends: the contacts in the A and C-boxes are not staggered from one strand to the other, while peaks in the signals from opposite strands in the IE are staggered by five to six nucleotides in the 5’ direction. Thus, in conjunction with the hydroxyl-radical footprint of TFIIIA (Vrana et al., 1988; Hayes it a,Z., 1989; Churchill et al., 1990).

Gene Complex

413

the results from the missing-nucleoside experiment suggest that at the two ends of the complex the fingers of the protein bind to DNA in a similar way, while in the central region the fingers interact differently with DNA. The missing-nucleoside signals from the A and C-boxes suggest that in those regions TFIIIA makes a continuous set of contacts along the DNA helix. In particular, the lack of stagger of the missingnucleoside signals from one strand to the other is consistent with the protein following a similar helical trajectory as the DNA. The hydroxyl-radical footprint shows a moderate degree of protection that is approximately equal throughout, these regions, and which extends for slightly more than one turn of the helix in each (Vrana et nl., 1988). Protection of the non-coding strand is greater than protection of the coding strand. Others have shown that, in the C-box, all of the phosphat’e contacts are detected on the non-coding strand (Sakonju & Brown. 1982). The Zif268 cocrystal structure (Pavletich B Pabo, 1991) establishes t’hat, the three linked fingers of this protein bind in the major groove of DNA, making contact nearly exclusively with one of the strands. Thus it appears that in t,he A and (‘-boxes of the ICR, TFIITA runs along the major groove for a turn of the DiL’A helix. The protein probably associates more closely with the phosphodiester backbone of t,he non-csodmg strand in these regions. In the intermediat’e element. a different picture of the structure emerges from our experiments. We have used two key observations. the relationship of the footprint) t’o the missing-nucleoside pattern. and the stagger of the missing-nucteoside pattern itself. to work out the structure. The positions of the missing-nucleoside contacts in the IE are related in an int’eresting way to features in the hydroxyt-radical footprint of TFTIIA. Strong footprints are found at +67 on the non-coding strand and at + 76 on t,he caoding strand (Churchill it al., 1990), while the cont.acts found in the missing-nucteoside experiment are cent,ered at $69 and +74.5 on these strands. This relationship of footprints to base contacts is very similar to t,hat observed for the 1” repressor and cro proteins (Hayes dz Tullius. 1989). These two proteins make contact mainly with one side of the D?;A helix. An alpha helix on the surface of the protein is inserted into the major groove of DNA and makes a series of sequence-specific contacts. In t’he missing-nucleoside analysis of the 2 repressor-DNA and cro-DNA complexes (Hayes & Tullius, 1989): the center of the set of contacts on a strand occurs one or t,wo nucleotides 3’ to the center of the footprint caused by the protein crossing over the minor groove adjacent to the major groove in which the contacts are made. The presence of this same pattern in the TE is evidence that a similar protein-I)NA int~eraction occurs t,here in the TFIII&DlVA complex. The stagger of the missing-nucleoside contacts provides additional evidence for the locat,ion of the protein in the TE. The sets of contacts centered at

position +69 (non-coding strand) and + 74.5 (coding strand) occur directly across t’he major groove of DNA from each other. These cont,act,s are separated by a region where the minor groove is very accessible to the hydroxyl radical. centered around position +72. Footprinting data (Vrana rt al.. 1988; Churchill rt al., 1990) suggest that TFIIIA crosses over the minor groove at two locations, positions +65.5 and +78, that flank the contacts detected in the missing-nuclebside experiment. Because the missing-nucleoside contacts are staggered and not mirrored across the DNA strands, t)he protein likely lies nearly parallel to the helix axis in this region, and does not follow along the major groove, as was suggested above for the A and (I-boxes. The aggregate picture of the TFITIA-DNA complex then is: (1) the amino-terminal part of the protein wraps al,ong the major groove of DNA from about + 96 to + 81. but makes energetically important contacts only from + 92 to + 78: (2) the protein leaves the major groove at + 78 and crosses over t,he minor groove; (3) the protein lies nearly parallel to the DNA helix axis in the center of the complex. making contacts on one face of the DNA in thr major groove (at position +74.5 on t,he coding strand, and at position +69 on the non-coding strand); (4) the minor groove at position +72 is on the opposite side of the DNA from the protein, and is open to solvent; (5) the protein crosses over the minor groove at position +65 and wraps along in the major groove from positions +61 to +42. making energetically important contacts with the DNA from +61 to +51, in a manner similar to that seen at the amino terminus. What specific parts of TFIIIA are responsible for each of these features of the proposed structure? In the Zif268-DNA complex (Pavletich & Pabo, 1991), three zinc fingers follow the major groove and make contacts with 9 bp of DNA. Nearly all of the cont,acts are made with the DNA strand that corresponds to the non-coding strand of the 5 S gene. This structure accounts well for our data on TFTIIA in the A and C-box regions of the ICR, in each of which about, one turn of the DNA helix is contacted by the protein. Thus, contacts from finger 1 would start at about + 91, while contacts for fingers 2 and 3 would be centered around +85 and +82, respec tively. These locations also generally agree with the weak but evident, structure in the missing-nucleoside signal from the coding strand in the C-box region (see Fig. 3). The extra region of footprint and t,he weak contacts beyond +92 at’ the 3’ end of the TCR could be caused by the 14 or HO residues t,hat precede the first’ finger in the amino terminus of t,he protein. The linker between fingers 3 and 4 is one or t,wo residues longer than the linkers between fingers 1 and 2 or 2 and 3. Given the break in contacts after +78 and the minor groove footprint at position +78, it appears the 34 linker and finger 4 do not continue in the major groove but instead cross over the minor groove at + 78. We suggest that fingers 4,

5 and 6 form a unit’ t’hat runs nearly parallel to the helix axis of the DNA in the center of the c~ompl~x. This unit would be in a position to make t hcs bast* contacts that are centered at +69 on t h(> noncoding strand and at + 74.5 on the coding strand in the major groove (Fig. 3). Kerg (1990) has suggested that the linkers coming into and out) of fingrr 6 are t’oo short t,o allow the protein to continue> it1 a helical trajectory. The three-finger unit would cross over the minor groove at posit’ion +65. and then fingers 7. X and 9 would make a continuous set of contacts in the major groove, in a manner similar in structure (hut not in energetics) to the caontacts made by fingers 1, 2 and 3. As mentioned &JW, contact’s in wch of these regions extend over about 10 bp and thus each finger would contact about 3 bp, in agreement with t,he Zif268 cocrystal structure (Pavl&ich $ l’abo. 1991). Klevit (1991), using the Zif268 crystal structure as a model, found those amino acid residues that occur at t,he “DT\U’A contact positions“ of’ t’hr TFTIIA fingers and proposed a set of srqu~‘ncespecific contacts between t,he TFTTTA fingers and the ICR DNA. While t,he contacts suggested for the IE are difficult t*o reconcile with our structure, the Dh’A sequences proposed to make cont’act with thcl protein in the A and (I-boxes caorrespond rather well with t,he contacts identifird by the missing-nucleoside experiment. A flIrther interesting point, from the analysis by Klevit is that only one base contact per finger is proposed for the &box. whik two plausible contacts per finger were found in the C-box. This fits well with the observat)ion made here and by ot,hers that TFITIA makes stronger contacts wit,h the 3’ compared to thr 5’ end of the 1(X. We note that the contacats t)o l)SA end ahruptl~ at position + 50 or + 51, while the hvtlrox!;l-riLdic:al footprint extends to about + 42. This extra prot,cction is c~ausetl by the carboxyl terminus of t,he protein. sincbr the deletion of 62 non-finger rwidues from t’he carboxyl end of TFTTTA shortens the foot-. print to t,he edge of t’hr A-box (Vrana rt (11.. 198X). A representation of our model for the TFITIA-DNA complex is shown in Figunh 4. (b) Comparison

with

lkker

scanniny

mutants

There are interesting parallels between thr linkerscanning mutagenesis study of You et al. (1991) and t’he missing-nucleoside results. Of the ten linkerscanning mutants spanning the ICR that’ were tested for TFITTA binding, all but three agree ver? well (in fact. almost, quantitatively) with the contacts we derive from Figure 3. For example, two linker scanning mutants, 78-81 and 82--X6, caused TFIIIA binding to decrease to 6% and 1 ‘&, respectively, relative to the wild-type sequence. These mutants coincide with the strongest missing-nucleoside signals t’hat we observe. The three exceptional mutants are informative. however. The 53-56 mutant has no effect on binding. although it, occurs in a region that shows

TFIIIA-5

8 RNA

C 87 I

Gene C’omples

IE 77 I

415

A 67

57

/

I

POOH

NH2 Figure 4. Model of the TFIIIA-5 S DE’A complex. The approximate positions of the individual fingers of TFIIIA are shown with respect to the intragenic control region of the 5 S gene, modeled as a DNA helix with 10 bp/turn. The configurations of the 2 sets of 3 fingers at the ends of the complex (1-3 and 7-9) were drawn with reference to a view of the Zif268 cocrystal structure (Pavletich & Pabo, 1991) in a similar orientat’ion. Fingers 1. 3, 7 and 9 are slightly foreshortened to give perspective to the drawing, since these fingers are meant to proceed into and out of the plane of the paper. The carboxyl and amino termini of the protein are marked. It should be noted that only t,he !> zinc fingers of TFIIIA are portrayed in the model; there is an additional protein domain at the carboxyl terminus that does not bind to DNA. but that is important in regulation of transcription (Vrana et al.. 1988). Numbers above the strucsture denote nucleotide positions within the 5 S ICR. The letters show the approximate locations of the A-box. (‘-box. and TE. (1. coding strand: NC. non-coding strand. large missing-nucleoside signals. However, You et nZ. (1991) used the Xlo gene in their experiments, while we studied the Xbs gene. Although the 53-56 linker scanning mutation changes two base-pairs of the Xlo gene (53 and 56), in fact these base-pairs are changed to the sequence that is found in the Xbs gene. It, is, therefore, not at all surprising that this mutant would support DNA binding by TFIIIA. The other exceptional mutants, 71-72 and 73-76, when considered along with the missing-nucleoside results, perhaps offer additional insight into the interaction of TFIIIA with DNA in the IE. These two mutants also have no deleterious effect on binding, yet missing-nucleoside signals are seen at 71 and 72 on the non-coding and at 74 and 75 on the coding strand. According to our proposed structure, the 71-72 mutant occurs where the protein spans the major groove and therefore may not be making base-specific contacts. The 73-76 mutant encompasses the two strongest missing-nucleosides on the coding strand, 74 and 75. The linker scanning mutant changes an AG sequence to a GA sequence here, perhaps indicating that either purine is tolerated for the strong coding-strand contacts we observe in the missing-nucleoside experiment. (c) Comparison

with other models

The structure we propose for the complex of TFIIIA with DNA has some elements in common with previous models, but differs in important ways from them. Most models have been based on the assumption that the nine fingers of TFIIIA make more or less equivalent interactions with DNA (Fairall et al., 1986; Churchill et al., 1990). While this idea is borne out in the Zif268 structure (Pavletich & Pabo, 1991), in which the three fingers and their

DNA-binding sites have nearly exact screw symmetry, data from our laboratory and from other laboratories suggest that the TFTIIA-DNA complex might not have such high symmetry. Our structure incorporates the two types of interactions that have been suggested for zinc fingers binding to DNA (Fairall et al., 1986), that is, fingers wrapping around the DNA, and fingers lying along one side of the DNA helix. One of the distinguishing characteristics between models for the TFIIIA-5 S DNA complex is the number of base-pairs predicted to be contacted by each subunit of the protein. The model of Berg (1990) predicts that each finger contacts about 3 bp, while other models (Fairall et aZ., 1986; Rhodes & Klug, 1986; Churchill et al., 1990) suggest that each repeating unit contacts about 5 bp. The results presented here show that the range of contacts with DNA made by TFIIIA is not quite as large as the footprint of the protein. Contacts extend from about +51 to + 92 (about 40 bp) while the footprint is approximately IO bp longer. This conclusion is supported by the hydroxyl-radical footprints of a series of deletion mutants of TFIIIA, which suggest that non-finger regions adjacent to finger 9 contribute to the 5’ end of the protection pattern (Vrana et al., 1988). Thus, estimates of the number of basepairs contacted per finger, if calculated by dividing the number of base-pairs in the footprint by the number of fingers, may well be too large. (d) Implications

of the proposed

structure

Why is the TFIIIA-DNA complex not simply three Zif268-DNA complexes strung together? We think that there would be a serious kinetic problem for the formation of such a structure. To bind to

416

.J. J.

Elayrs

rend

DNA in this fashion, a protein with nine fingers would make nearly three turns around the DNA helix. Dissociation from DNA would require the protein to rotate 1000” or so relative to the DKA. In contrast, in our proposed structure TFITTA never makes a complete turn around DNA. Dissociation of the protein would involve independent unwrapping of two three-finger segments of the prot’ein from around the DNA helix. The entire protein would not be required to rotate completely around the DNA even once. We anticipate that other zinc finger proteins that contain more than three fingers are likely to adopt structures similar to the proposed TFIITA-DNC’A structure, hecausr of these considerations. Berg’s mode1 of the TFIIIA-DNA complex (Berg, 199(l), while superficially similar to our structure, incorporates a set of five fingers wrapping continuously around the DNA helix. This mode1 would therefore suffer from the same difficulties as the (Zif)3 structure, because these five fingers would make more than one turn around the DNA.

5. Summary We have used the results of missing-nucleoside experiments to construct a mode1 of the TFIIIA-5 S DNA complex, in which the protein is envisioned to wrap around the DNA at the two ends of the complex and run parallel to the helix axis in the center of the complex. This mode1 is consistent with previous mutagenesis and high-resolution footprinting experiments. We thank Alan Wolffe for providing TFIIIA protein, and for helpful discussions. This work was supported by PHS grant GM 41930. J. H. acknowledges a postdoctoral fellowship from the National Research Council. T. D. T. is a fellow of the Alfred P. Sloan Foundation, the recipient of a Camille and Henry Dreyfus Teacher-Scholar Award, and a Research Career Development Awardee of the PHS (CA 01208).

References Bazett-Jones, D. P. & Brown, M. L. (1989). Electron microscopy reveals that transcription factor TFIIIA bends 5 S DNA. Mol. Cell. Biol. 9, 336-341. Berg, J. M. (1988). Proposed structure for the zincbinding domains from transcription factor IIIA and related proteins. Proc. Nat. Acad. Sci., U.S.A. 85, 99-102. Berg, J. M. (1990). Zinc finger domains: hypotheses and current knowledge. Annu. Rev. Biophys. Biophys. Chem. 19, 405-421. Bieker, J. J. & Roeder, R. G. (1984). Physical properties and DNA-binding stoichiometry of a 5 S gene-specific transcription factor. J. BioE. Chem. 259, 6158-6164. Bogenhagen, D. F., Sakonju, S. & Brown, D. D. (1980). A control region in the center of the 5 S RNA gene. II. The 3’ border of the region. Cell, 19, 27-35. Brown, R. S., Sander, C. & Argos, P. (1985). The primary structure of transcription factor TFIIIA has 12 consecutive repeats. FEBS Letters, 186, 271-274.

7’.

11. l’ulli~s

(Churchill. Ji. E. A. (1987). I)o~toral dissertstiotr. ‘I’ht, *Johns Hopkins ITniversity, I{altimorr. MI). (‘.S..\. (‘hurchill. M. E. A., Tullius. T. 1). B Klug. A. (l!)!)(j). Mode of interaction of the zinc finger prot,ein TFII IA with a 5 S RNA gene of Xen,opus. I’m.. Sort. .-lmd. A%., I;.S.A. 87. 5528-5532. Engelke, D. R., Ng, S-Y, Shastrg. K. S. & Roedrr. Ii. (:. (1980). Specific interaction of a purified transcript,ion factor with an internal control region of 5 S RXA genes. Cell, 19, 717--728. Fairall? L.. Rhodes, D. & Klug, A. (1986). Mapping of’ the sites of protection on a 5 S RNA gene by the Xenorrus transcription factor IIIA. J. Mol. Biol. 192, 577 591. Ginsberg, A. M., King, B. 0. & Roeder. R,. G. (1984). Xsnopus 5 S gene transcription factor, TFITIA: Characterization of a cDNA clone and measuremmt of R?r’A levels throughout development. (‘p/l. 39. 47Sb489. Hanas, J. S.. Bogenhagen, D. F. & Wu, C.-W. (1983). Cooperative model for the binding of Xeno2)u.s transcription factor A to the .5 S RNA gene. Proc. Sat. Acad. ~S’ci.. .?:.S.A. 80. 214%2145. Hayes, J. *I. & Tullius, T. D. (1989). The missing nucleoside experiment: a new technique to study recognition of DNA by protein. Biochemistry, 28, 9521-9527. Hayes, *J. ,J.. Tullius. T. 1). & Wolffe, A. I’. (1989). A protein-~protein interaction is essential for stable complex formation on a 5 S RNA genr. .I Riol. Chem. 264. 6009-6012. Klevit, R. E. (1991). Recognition of DNA by Cys2.His2 zinc fingers. Science, 253, 1367, 1393. Lee, M. S.. Gippert. G. P., Soman, K. V., Case, 11. A. b Wright, P. E. (1989). Three-dimensional solution structure of a single zinc finger DNA-binding domain. Science, 245, 635-637. Miller, J., McLachlan, A. D. & Klug, A. (1985). Repetitive zinc-binding domains in the protein transcription factor TIIA from Xenopus oocytes. EMBO .I. 4, 1609-1614. Parraga, G.. Horvath, S. J., Eisen, A., Taylor, W. E., Hood, L., Young, E. T. & Klevit, R. E. (1988). Zinc-dependent structure of a single-finger domain of yeast ADRl. Science, 241. 1489-1492. Pavletich. N. P. & Pabo, C. 0. (1991). Zinc finger-DNA recognition: crystal structure of a Zif268-DNA complex at 2.1 A. Science, 252, 809-817. Pieler, T., Hamm, J. & Roeder, R. G. (1987). The 5 S gene internal control region is composed of three distinct sequence elements, organized as two functional domains with variable spacing. Cell, 48, 91-100. Rhodes, D. & Klug, A. (1986). An underlying repeat in some transcriptional control sequences corresponding to half a double helical turn of DNA. Cell, 46, 123-132. Sakonju, S. & Brown, D. D. (1981). The binding of a transcription factor to deletion mutants of a 5 S ribosomal RNA gene. Cell. 23, 665-669. Sakonju, S. & Brown, I>. D. (1982). Contact points between a positive transcription factor and the Xenopus 5 S RNA gene. Cell, 31, 395-405. Schroth, G. P., Cook, G. R., Bradbury, E. M. and Gottesfeld, J. M. (1989). Transcription factor IIIA induced bending of the Xenopus 5 S gene promoter. Nature

(London),

340,

487488.

Smith, D. R., Jackson, I. J. & Brown, D. D. (1984). Domains of the positive transcription factor specific for the Xenopus 5 S RNA gene. Cell, 37, 645-652. Tullius, T. D. & Dombroski, B. A. (1986). Hydroxyl radical “footprinting” : high-resolution information

TFIIIA-5

S RNA

about DNA-protein contacts and application to lambda repressor and cro protein. Proc. Nat. Acud. Sci., U.S. A. 83, 546%5473. Tullius. T. D., Dombroski, B. A., Churchill, M. E. A. & Kam, L. (1987). Hydroxyl radical footprinting: a high-resolution method for mapping protein-DNA contacts. Methods Enzymol. 155, 537-558. Vrana. K. E., Churchill, M. E. A., Tullius, T. D. & Brown, D. D. (1988). Mapping functional regions of tran-

Edited

Gene Complex

417

scription factor TFIIIA. Mol. Cell. Biol. 8, 1684-1696. Wolffe, A. P. (1988). Transcription fraction TFIIIC can regulate differential Xenopus 5 S RNA gene transcription in vitro. EMBO J. 7, 1071-1079. You, Q., Veldhoen, N., Baudin, F. & Romaniuk, P. J. (1991). Mutations in 5 S DNA and 5 S RNA have different effects on the binding of Xenopus transcription factor ITIA. Biochemistry, 30, 249,525OO.

by P. von

Hippel