Communication Vol. 264, No. 15, Issue of May 25, pp. 8463-8466,1989 THE JOURNAL OF BIOLOGICAL CHEMISTRY

0 1989 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

Asymmetric DNA Bending Induced by the Yeast Multifunctional Factor TUF*

(Received for publication, November 28, 1988)

Marie-Luce Vignais and Andre Sentenac From the Departement de Biologie, Service de Biochimie, Centre d%tudes Nuclkaires de Saclay, 91191 Gif-sur- Yvette Ceden, France

TUF is a yeast regulatory factor that binds to con- served DNA sequence elements involved in gene acti- vation or silencing as well as in telomere function. Using gel electrophoresis analyses, we show here that TUF induces DNA bending at a site located upstream of the recognition sequence (rpg box). Several point mutations in the rpg box reduced TUF binding strength without affecting the extent of bending. Selective pro- teolysis of TUF-DNA complexes further suggested the existence of two separate protein domains involved in DNA bending and specific DNA recognition. DNA bending may be an important feature of multifunc- tional factors that could help them to recruit other proteins for the formation of multiprotein complexes.

TUF (1, 2) is a yeast general factor that binds to DNA at specific sites involved in multiple functions. TUF binding sites act as an upstream activating sequence (UASrpg)’ (3,4) in a large family of housekeeping genes (5), form an essential part (E element) of the silencer of the mating type loci (6, 71, and are also found as repeated motifs (&A) in yeast telo- meres (7). The functional diversity of TUF recognition sites suggests a pivotal role for TUF in recruiting other factors on neighboring sequences, possibly by changing DNA confor- mation in these regions. Protein-induced DNA bending at specific sites is considered to be an important mechanism for promoting the multiprotein interactions involved in regula- tion of gene activity (8-lo), initiation of replication (11-13), or site-specific recombination (14, 15). The best studied ex- ample of a DNA-bendmg regulatory factor is the bacterial CAP protein (8, 16, 17). Electrophoretic analysis of lac pro- moter. CAP complexes revealed the induced bend (16,18) and allowed its mapping near the center of the binding site (8). This powerful technique is based on the dependence of gel mobility on the position of the bending locus in identically sized fragments. Following the strategy of Wu and Crothers (16) we found that TUF is inducing DNA bending. Evidence is presented that two separate protein domains are involved in specific DNA recognition and DNA bending.

* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

‘The abbreviations used are: UASrpg, upstream activating site with a rpg box; rpg box, ribosomal protein genes box; ARS, autono- mous replicating sequence; rpg-1, ACACCCATACATTT; R super- script refers to the orientation (reverse) of the rpg cassette cloned at the BamHI site; bp, base pair(s).

RESULTS

To examine the effect of TUF binding, we constructed a series of linear DNA fragments of identical size harboring a unique TUF binding site at various locations. The high effi- ciency binding site (5) ACACCCATACATTT (rpg-1) was cloned into a plasmid, and a HinpI-HinpI 432-bp DNA frag- ment harboring this sequence was excised, labeled with 32P, circularized by ligation, and cleaved at unique sites using different restriction enzymes. The rpg-1 cassette being cloned in either orientation, two sets of linear circularly permuted fragments were obtained (rpg-1 and rpg-lR in Fig. 1). These fragments were incubated with purified TUF, and the migra- tion of TUF. DNA complexes was analyzed by electrophoresis on a polyacrylamide gel. No difference was detected in the migration of the naked DNA fragments, which displayed no intrinsic sequence-dependent curvature (16) (Fig. lA 1. On the other hand, TUF-DNA complexes were retarded to different extents, depending on the position of the rpg box within the fragments. A plot of the relative migration rates (RF) of the complexes against the map position of the cleavage sites is shown in Fig. 1B. The faster mobilities were observed for DNA fragments with TUF bound near one extremity. This was the case with the rpg-1 fragments produced by EcoRI or BamHI cuts. The largest retardation effect was obtained when the TUF binding site was centrally located, as when fragments were produced by HinpI, PuuI, or PuuII cuts. This migration behavior was typical of a protein-induced DNA bend (16).

By extrapolating the nearly linear portions of the curve in Fig. 1B the bending site appeared to be asymmetrically located on the 5’ side of the rpg sequence. This is illustrated by the fact that although the BamHI site was closer to the center of the rpg cassette, the fragments cleaved by HincII, on the other side of the cassette, migrated faster (compare lanes 7 and 8 of the rpg-lR series). This conclusion was confirmed by compar- ing the migration rates of 285-bp EcoRI-PuuII DNA frag- ments harboring the rpg cassette cloned in different orienta- tions at the BamHI site, close (21 bp) to the EcoRI end. If the bending site were asymmetrically located relative to the rpg box a significant difference in migration should ensue. In the reverse orientation (rpg-lR) the bend would be shifted by about 25 bp (ie. the size of the TUF-protected region (1, 2)) toward the center of the fragment and the complex should be more retarded. Indeed, the two classes of complexes had markedly different mobilities, with RF values of 0.42 for the forward orientation and 0.37 for the reverse orientation (see lanes 1 and lR on the right part of Fig. 2). A mixture of the two complexes was well resolved by the gel retardation tech- nique (lane 1 + lR). These results confirmed the excentric location of the bending site on the 5‘ side of the rpg box. Note that the sequence upstream of the rpg box differed in the forward and reverse orientations.

In the same experiment, we investigated the sequence de- pendence of TUF-induced DNA bending by analyzing the effect of point mutations on the mobility of mutant complexes. The mutant sequences were generated by random incorpora- tion of incorrect bases during chemical synthesis of rpg-1.’ We compared six cassettes, with mutations spanning the rpg sequence, that showed relative affinities for TUF ranging

M-L. Vignais, unpublished results.

8463

8464 As>*mmttric DNA Bending by TUF Protein

1 6 15 2215 1 1 3R MR ISR MR lR lR

FIG. 2. Effect of point mutations and binding site orienta- tion on the mobility of TUF-DNA complexes. Variants of the rpg-1 sequence were ohtained hv saturation mutagenesis using syn- thetic oligonucleotides essentially as descrihed (19). The douhle- stranded rpg cassettes were cloned, as descrihed in the legend to Fig. 1, in either orientation into the HamHI site of the RS plasmid. DNA prohes (285 hp) harhoring the mutant cassettes were ohtained by cutting at the I.:coRI site (see Fig. 1H) and at a PuulI site located heyond the Hind111 site (not shown in Fig. 1). The probes were "P- labeled at the F h R I end hv filling in with a DNA polymerase Klenow fragment. DNA.protein complexes were formed with 20 or 2.50 ng of TLJF, as indicated, and subjected to electrophoresis as described in Fig. 1. The rpg variants were: rpg-1, the wild type sequenre

CCATACATTT; rpg-14, ACACTCCTACATTT; rpg-15, ACACCCT- TACATTT; rpg-16, ACACCCAAA-CATTT rpg-22, ACACCCATT

rpg cassette cloned at the RamHl site (see Fig. 1R). The mutants rpg-3, -6, -14. -15, -16, and -22 had relative affinities of 0.9, 0.1, 0.07, 0.12. 1.0, and 0.04, respectively, as compared to rpg-1.' T o achieve sufficient complex formation with the mutant probes, the eight com- plexes starting from the left were formed with 250 ng of T U F factor, while the control rpg-1 prohes (the last three lanes) were incubated with 20 ng of TUF. When indicated, two DNA probes were mixed hefore incubation with TUF.

ACACCCATACATTT; we.?, TCACCCATACATTT; rpg-6, ACCC

- TATTT. The R superscript refers to the orientation (reverse) of the

from 100 to 5% of that of the wild type rpg-1 sequence (see legend to Fig. 2). In view of the low affinity of T U F for some mutant sequences, a 12-fold higher concentration of factor was used to form mutant complexes. This difference in protein concentration did not affect the migration of the control rpg- 1 probe (lanes I ). T h e 285-bp DNA probes harboring the rpg- mutated cassettes cloned in different orientations were incu- bated with TUF, and the migration of the complexes was analyzed by electrophoresis through a nondenaturing gel (Fig. 2). Remarkably, none of the mutations changed significantly the rate of migration of the complexes as compared to the control rpg-1 sequence. The complexes migrated with RF values of 0.42 or 0.37, depending on the orientation of the rpg cassette, as discussed above. These observations contrasted with the results of Gartenberg and Crothers (17), who found that changes in DNA sequence within the distal binding site of the CAP protein altered the mobility (ie. bending) of mutant complexes, with variations of RF values of up to 30%. In the case of TUF, apparently, the extent of bending was not correlated with binding strength at least for the mutations examined. These results suggested that two domains of the protein were involved in the bending process, one being re- quired for specific binding t.o the core recognition sequence and another distal bending domain contacting DNA on the 5' side, probably nonspecifically a s suggested by the high sequence variabi1it.y of TUF binding sites in this region.

T o explore this hypothesis, we subjected TUF.DNA com- plexes to limited proteol-ytic treatment and analyzed the elec- trophoretic mobilities of the truncated protein complexes. Proteinase K generates a protease-resistant fast migrating complex retaining a prot,ein binding core of 50 kDa, which is about one-third the size of T U F (2). This truncated polypep-

I,* DNA

rpg.1'

,#"-="- -."-a

I I i

/ 0.5.

I \I

wg.1 ACACCCATACATTT

w mob

FIG. 1. Electrophoretic mobility of TUF-DNA complexes formed with circularly permuted DNA fragments. A , to gen- erate circularly permuted Irngments. with one strong TlIF hinding site. the rpg-I srquence (overlined).

GATCACACCCATACATTTCG TGTGGGTATGTAAAGCCTAG

was cloned into t he HnmHl site ofthe HS plasmid vector (Rluescrihe, Stratagene). A / f i n p l - f f i n p l fragment (Ai :?? hp) was isolated, '"P- lahrled l)y kinasing. and srlf-ligated. Only one HarnHI site was restored hy the cloning prncess. The circles were cleaved at unique sites with tlifferrnt rrstriction enzymes and the DNA fragments incuhated with factor 'T'IIF purified a s previously desrrihed (2) . Ap- proximately 6 fmol o f each DNA fragment (5000 cpml were incubated in 20 p I o f a mixture containing 21) mM Tris-HCI, 50 mM KCI, 5 mM MgCl,. 03 mM CaCI2, 0 . 5 mhl dithiothreitol, 0.1 mbf EDTA, 12% glycerol ( v / v l , 20 11% ofcarrier pI31<:122 D S A , and 20 ng of TIJF. After I O min at 25 "(*, romplexed and free DSAs were separated by elec- trophorrsis on a 5"; polyacrylamide gel at 4 "C for 6 h at 1.5 V/cm, as descrihed 11 I . The gel WIS autoradiographed at -70 "C with an intensifying screen. Restriction enzymes were: I , HinpI; 2, I'cuI; 3, / 1 ~ ~ ~ ~ 1 1 : 4, b ' r h l : 5 . Mocll: 6. EcoRI: 7. HornHI: X , HincII: 9, Hindlll; 10 , none. Complexed and free DSA hands are indicated. The nrrotu and nrrorrlwod show the nligrat ion of free (lanc I I ) and complexed (Innc, 1 0 ) D S A circles. respectively. Partial cleavage with Mac11 rpsulted in the formation of free and complexed circles. R, mapping 'IY~F-hentling locus. The curc'c, shows the relative mohilities of T I J F . DNA complexes plotted as a function of the restrictinn site used to open DNA circles. The same results were ohtained in two independent rxperiments. The tu.0 sets of circularly permuted fragments rpg-1 and rpg-1" differed hy thr orientation of the rpg cassette cloned at the IiomHI sitc. The map positinn o f the restriction sites relative to t hr rpgcassettr ihlnrl; orrow) is shown for the rpg-1 and rpg-I" linear fragments gener:lted h y Hinpl cleavage. A. relative mohility of rom- plexes formed with t h e diffrrent restricted rpg-1 fragments: 0, with rpg-I" fragments. The relntivc mohility ( H F ) is the ratio of the migration distances of protein. TINA complexes and the correspond- ing free I)NA hanrl.

Asymmetric DNA Bending by TUF Protein 8465

A rw-1 rpg -1

C I F:

C' 5'

r: s

r: p3 a free wp. w

0 1 2 3 4 5 6 7 1 2 3 1 5 6 7

C F: ;P'

'4 $ '5 P5 I

a B s

FIG. 3. DNA bending property is lost upon proteolytic treat- ment of TUF-DNA complexes. Factor TUF was incubated for 10 min in 20 pl with the 285-hp EcoRI-PuuII rpg-l or rpg-1" probes as descrihed in Fig. 2 in the presence of varying amounts of cu-chymo- trypsin (panrl A ) or proteinase K ( p o n d R ) . After incuhation, the samples were directly loaded on the gel and subjected to electropho- resis as in Fig. 1. The amount of protease added was: I , 0; 2, 5 pg; 3 , 10 pg; 4 , 25 pg: 5, 50 pg; 6, 125 pg. L a n a 7, DNA probe alone. C and C' indicate the unproteolvzed complexes. PI, Pz, . . ., Pe refer to different proteolyzed complexes (no correspondence is meant between the complexes generated hy the two proteases).

tide gives the same extended footprint pattern on the TEF2 promot.er as the native factor (2). It was therefore of interest to investigate the DNA bending properties of this DNA bind- ing domain. We used the differential migration of the rpg-1 and rpg-1'' complexes with TUF bound in opposite orienta- tions to monitor DNA bending. The rpg-l and rpg-lR probes were incubated with TUF, and the complexes were subjected to proteolysis using varying concentrations of n-chymotrypsin (Fig. 3A) or proteinase K (Fig. 3R) and analyzed by the gel retardation assay. Increasing protease concentrations gener- ated DNA. protein complexes of increasing velocity, with a band pattern characteristic of the protease used (Fig. 3, lanes 1-6). The limited protease treatments did not affect DNA binding, and no free DNA accumulated. Upon proteolysis by tu-chymotrypsin, three distinct complexes were formed. Inter- estingly, when the rpg-1 and rpg-lR probes were compared, only the native and the first proteolyzed intermediat.e PI displayed the differential migration characteristic.of bending. The migrations of Ps and Pa complexes were identical for the two probes (&values of 0.66 and 0.86, respectively). Although proteinase K generated a more complex band pattern, the same phenomenon was clearly apparent; the fast migrating extensively proteolyzed complexes P, to P6 ( R P 0.64, 0.67, 0.79, and 0.85) lost the differential mobilities observed in the case of the native or slightly proteolyzed complexes (Fig. 3B). Note that the lowermost complex P, corresponded to the fast migrating protease-resistant core analyzed previously (2). The formation of sharp DNA.protein complexes indicated that the specific DNA-binding and bending properties were dis- jointed by a rather selective proteolytic cleavage. The results strongly suggested the existence of a protein domain required for DNA bending, separate from the specific DNA recognition domain.

DISCUSSION

TUF, also called GRFl (general regulatory factor) ( 7 ) or RAP1 (repressor activator protein) (6), is an abundant mul- tifunctional DNA-binding protein that recognizes sequences implicated in gene regulation (1, 7, 20) as well as in plasmid mitotic stability (20, 21) and telomere function (7, 22). Such

functional versatility could reflect the aptitude of factor TUF either to interact with a variety of protein components or to recruit other factors by altering DNA conformation (or both). The same intriguing question arises with ot,her DNA-binding factors, CTF/NF-I (23) and OTF-I/NF-111 (241, that act,ivate both transcription and DNA replication. The present results suggest that DNA bending could be an important parameter as a nucleation site for mult.iprotein complex formation. It has been speculated that initiation of DNA replication a t t h e silencer was important for establishing transcriptional repres- sion (25). As bent DNA has been found to act as a replication enhancer in yeast (26, 2'3, it may be significant that a TUF- binding site lies next to an ARS element within the silencer region (20). DNA bending could also facilitate the observed attachment of the silencer region to the nuclear scaffold (28), which is probably responsible for the mitotic stabilization of plasmids in which the silencer is used as replication origin (20, 21). TUF-binding sites are also found as repeated motifs (close to an ARS) at the ends of yeast chromosomes. A cloned terminal repeat (81 bp) can bind up to four molecules of TUF, a s evidenced by the formation of four complexes in the mo- bility shift assay.:' A role in telomere function could involve the formation of a condensed structure to preserve telomere integrity and possibly favor its replication and interaction with nuclear structures.

The different TUF-binding sites may not all require bend- ing to carry out their diverse functions. This may be the case for the UASrpg sequences. However, together with the heat shock transcription factor HSTF (9) and the viral protein E? (29) , TUF is the third case of a eukaryotic transcriptional activator reported t.o bend DNA. No similar data are available yet concerning the prototypical GAL4 or GCN4 factors. The current model, derived from delet ion and fusion experiments, suggests that the DNA-binding site onlv serves to position an acidic activating domain near the gene (30, 31). However, these experiments do not rule out a role for bending in gene activation, since, as shown here by proteolysis of TUF, the DNA-bending property could be lost without affecting DNA binding. It was recently found that synt.hetic curved DNA sequences replacing the CAP-binding sit,e in t.he galactose promoter can act as transcriptional activators in E. A requirement for a DNA conformation change at promoters could explain the surprising observat.ion by Graves et al. (32) tha t a mutation at the first residue of CCAAT boxes severely impaired promoter activity while increasing binding of C B P protein. A similar paradoxical mutation effect was reported by Cohen et al. (33) for another transactivator.

Acknorc.lrdpmrnts-We thank Dr. .J. Huet for providing purified factor TUF and for suggesting the protease experiment, Dr. . J - M . Ruhler for help and advice in the isolation of mutated oligonucleo- tides, and Drs. P. Fromageot. M. Paule, P. Oudet, and H. Ruc for fruitful discussions.

REFERENCES

1. Huet, J . , Cottrelle, P.. Cool, M., Vignais, M. I,., Thiele, D., Marck, C., Ruhler, J . M., Sentenac, A,, and Fromageot, P. (1985) E M H I ) J . 4, 3.539-3547

2. Huet, .J., and Sentenac. A. (1987) ['roc. Nat l . Acad. Sci. 1'. S. A. 84,3648-3652

3. Woudt, L. P., Smit, A. R., Mager, \V. H., and Planta. R. .J. (1986) E M H I ) J . 5 , 10:%7-1040

4. Rotenberg, M. O., and Woolford. J . Id., ,Jr, (1986) Mol. C d . Riol. 6 , 6 7 4 -687

5. Vignais, M. I+ LVoudt, I,. P., Wassrnaar. C. M., Mager, W. H., Sentenac, A,, and Planta. R. J . (1987) E M H O J . 6, 1451-14.57

"~ ________ ~ ~ _ _ _ _ _ ~ ~ ~ ~~ ~ ~ _ _ _ _ ~

:' "1,. Vignais, unpuhlished ohservation. H. Ruc, personal communication.

8466 Asymmetric DNA Bending by TUF Protein 6. Shore, D., and Nasmyth, K. (1987) Cell 51, 721-732 7. Buchman, A. R., Kimmerly, W. J., Rine, J., and Kornberg, R. D.

8. Liu-Johnson, H. N., Gartenberg, M. R., and Crothers, D. M. 310,154-157

9. Shuey, D. J., and Parker, C. S. (1986) Nature 323, 459-461 Nature 334,218-224

G., and Kelly, T. S. (1988) Science 241, 1210-1213

21. Kimmerly, W., Buchman, A,, Kornberg, R., and Rine, J. (1988)

22. Shampay, J., Szostak, J. W., and Blackburn, E. H. (1984) Nature

23. Santoro, C., Mermod, N., Andrews, P. C., and Tjian, R. (1988)

10. Echols, H. (1986) Science 233, 1050-1056 11. Mukhergee, S., Patel, I., and Bastia, D. (1985) Cell 43, 189-197 24. O'Neill, E. A., Fletcher, C., Burrow, C. R., Heintz, N., Roeder, R.

12. Zahn, K., and Blattner, F. (1985) EMBO J. 4, 3605-3616 13. Koepsel' R' R'7 and A' Science 2339 1316-1318 26. Williams, J. S., Eckdahl, T. T., and Anderson, J. N. (1988) Mol.

25. Miller, A. M., and Nasmyth, K. A. (1984) Nature 312,247-251 14. Hatfull, G. F., Noble, S. M., and Grindley, N. D. F. (1987) Cell

(1988) Mol. Cell. Biol. 8, 210-225

(1986) Cell 47,995-1005

EMBO J. 7,2241-2253

49,103-110 Cell. Bid. 8, 2763-2769 15. Robertson, C. A., and Nash, H. A. (1988) J , Bioi. ckm, 263, 27. Snyder, MV Buchman, A. R.9 and Davis, R. w. (1986) Nature

16. Wu, H. M., and Crothers, D. M. (1984) Nature 308, 509-513 28. Amati, B. B.l and Gasser* s. M. (1988) Cell 5 4 9 967-978 17. Gartenberg, M. R., and Crothers, D. M. (1988) Nature 333,824- 29. Moskaluk, c., and Bastia9 D. (lgM) proc. Acud. Sei.

829 U. S. A. 85,1826-1830 18. Kolb, A., Spassky, A., Chapon, C., Blazy, B., and Buc, H. (1983) 30. PtaShne, M. (1988) Nature 335, 6&3-689

Nucleic Acids Res. 11, 7833-7852 31. Struhl, K. (1987) Cell 49,295-297 19. Hill, D. E., Hope, I. A., Macke, J. P., and Struhl, K. (1986) 32. Graves, B. J., Johnson, p. F., and McKnight, s. L. (1986) Cell

Science 234, 451-457 44,565-576

719 Biol. Chem. 263.10377-10385

3554-3557 324987-89

20. Brand, A. H., Micklem, G., and Nasmyth, K. (1987) Cell 51,709- 33. Cohen, R. B., Yang, L., Thompson, J. A,, and Safer, B. (1988) J.