Sharp DNA Bends as Landmarks of Protein-Binding Sites on

extending from the apex. The presence of DNA-binding proteins at the apex was verified by atomic force micros- copy. The position of protein binding r...
0 downloads 0 Views 119KB Size
Download PDF
Anal. Chem. 1999, 71, 1663-1667

Accelerated Articles

Sharp DNA Bends as Landmarks of Protein-Binding Sites on Straightened DNA Hiroki Yokota,*,†,‡ Kevin Fung,† Barbara J. Trask,† Ger van den Engh,† Mehmet Sarikaya,§ and Ruedi Aebersold†

Department of Molecular Biotechnology and Material Sciences & Engineering, University of Washington, Box 357730, Seattle, Washington 98195

We have developed a fluorescence-based method for mapping single or multiple protein-binding sites on straightened, large-size DNA molecules (>5 kbp). In the described method, protein-DNA complexes were straightened and immobilized on a flat surface using surface tension. A fraction of the immobilized complexes displayed a sharp DNA bend with two DNA segments extending from the apex. The presence of DNA-binding proteins at the apex was verified by atomic force microscopy. The position of protein binding relative to the ends of the DNA molecule was determined by measuring the length of two DNA segments using fluorescence microscopy. We demonstrate the potential of the fluorescencebased method to localize protein-binding sites on the DNA template and to evaluate relative binding affinity. The proposed protein-binding-site mapping technique is simple and easy to perform. Practical applications include screening for DNA-binding proteins and the localization of protein-binding sites on large segments of DNA.

The specific interactions of DNA-binding proteins to DNA are of fundamental importance in biological systems, and the precise localization of protein-binding sites are critical for understanding complex cellular functions and mechanisms, in particular the control of gene expression.1 For identifying protein-binding sites on DNA, biochemical methods such as the electrophoretic mobility shift assay (EMSA) and DNase footprinting have been

widely used.2-4 However, detection of protein-DNA complexes with EMSA is generally limited to DNA fragments of 100-700 bp or less and is most commonly performed with synthetic oligonucleotides. In DNase I footprinting, the DNA templates are limited to the size of DNA covered by a DNA sequencing gel, typically ∼500 bp. Since eukaryotic promoters can extend over 5 kbp,5 these techniques are not suited for mapping specific DNAbinding sites on large regulatory DNA sequences. It would be useful to develop techniques that allowed physical mapping of elements for DNA-binding proteins on large segments of DNA. By anchoring one end of the DNA molecule, individual DNA segments can be extended and manipulated by various small forces such as electric or dielectric force,6-9 viscous drag,10-12 surface tension,13,14 magnetic force,15 or optical force.16-18 Stretching fluorescently labeled DNA molecules has been used for unraveling physical DNA properties such as entropic elasticity,15,19 (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)

†

Department of Molecular Biotechnology. ‡ Current address: Departments of Mechanical Engineering and Anatomy Indiana UniversitysPurdue University at Indianapolis 635 Barnhill Drive, MS504, Indianapolis, IN 46202; (phone) 317-274-2448; (fax) 317-274-9744; (e-mail) [email protected]. § Material Sciences & Engineering. (1) Ptashne, M.; Gann, A. Nature 1997, 386, 569-577. 10.1021/ac981370x CCC: $18.00 Published on Web 04/01/1999

© 1999 American Chemical Society

(15) (16) (17) (18) (19)

Lane, D.; Prentki, P.; Chandler, M. Microbiol. Rev. 1992, 56, 509-528. Tullius, T. D. Annu. Rev. Biophys. Biophys. Chem. 1989, 18, 213-237. Sleigh, M. J. Anal. Biochem. 1986, 156, 251-256. Martin, D. I. K, Fiering, S.; Groudine, M. Curr. Opin. Genet. Dev. 1996, 6, 488-495. Smith, S. B.; Aldridge, P. K.; Callis, J. B. Science 1989, 243, 203-206. Schwartz, D. C.; Koval, M. Nature 1989, 338, 520-522. Washizu, M.; Kurosawa, O. IEEE Trans. Ind. Appl. 1990, 26, 1165-1172. Asbury, C. L.; van den Engh, G. Biophys. J. 1998, 74, 1024-1030. Zimmermann, R. M.; Cox, E. C. Nucleic Acids Res. 1994, 22, 492-497. Perkins. T. T.; Quake, S. R.; Smith, D. E.; Chu, S. Science 1994, 264, 822826. Perkins, T. T.; Smith, D. E.; Chu, S. Science 1997, 276, 2016-2021. Bensimon, A.; Simon, A.; Chiffaudel, A.; Croquette, V.; Heslot, F.; Bensimon, D. Science 1994, 265, 2096-2098. Bensimon, D.; Simon, A. J.; Croquette, V.; Bensimon, A. Phys. Rev. Lett. 1995, 74, 4754-4757. Smith, S. B.; Finzi, L.; Bustamante, C. Science 1992, 258, 1122-1126. Perkins, T. T.; Smith D. E.; Chu, S. Science 1994, 264, 819-822. Smith, S. B.; Cui, Y.; Bustamante, C. Science 1996, 271, 795-799. Quake, S. R.; Babcock, H.; Chu, S. Nature 1997, 388, 151-154. Bustamante, C.; Marko, J. F.; Siggia, E. D.; Smith, S. Science 1994, 265, 1599-1600.

Analytical Chemistry, Vol. 71, No. 9, May 1, 1999 1663

nonlinear conformational transition,17,20 and dynamic behavior in an aqueous flow.11,12,16,18 The DNA-stretching technique is emerging as a routine biological tool in constructing genetic and physical DNA maps. In fiber-FISH gene mapping, a target genomic DNA is straightened and the location of the genetic marker is detected by fluorescence in situ hybridization.21-28 In optical restriction mapping, the stretched DNA under tensile stress is digested by a restriction endonuclease, and the ordered restriction fragments, with a spatial gap at recognition sites, generated from individual DNA molecules are measured.29-34 Previously we developed methods for straightening and immobilizing DNA and DNA-protein complexes on an atomically flat surface.34,35 During the course of these experiments, we observed that, under conditions leading to the immobilization of straightened DNA molecules, the protein-DNA complexes frequently appeared as a V-shaped structures. In this report, we demonstrate that the sharp bend in the DNA is coincident with the protein-binding site and that the protein-induced DNA bend can be used to localize the site of protein binding to DNA. MATERIALS AND METHODS Preparation of DNA Templates and Proteins. For imaging protein-DNA complexes by atomic force microscopy (AFM), we used the 5.6-kbp plasmid 1631 DNA template. The plasmid contains one GAL4-binding site, 5′-CGGAGGACAGTACTCCG-3′, located 2.1 kbp from one end (Figure 1A). The DNA template was incubated with GAL4 protein (a gift from I. Sadowski, University of British Columbia, Canada) as previously described,35 prior to immobilization on the sample plate. For determining protein-binding sites in protein-DNA complexes by fluorescence microscopy, we used a 19.0-kbp Stu I-digested λDNA fragment and 48.5-kbp λDNA (GibcoBRL) (Figure 1B). λDNA has three AP-1 consensus sequences (Jun (20) Cluzel, P.; Lebrun, A.; Heller, C.; Lavery, R.; Viovy, J.-L.; Chatenay, D.; Caron, F. Science 1996, 271, 792-794. (21) Parra, I.; Windle, B. Nature Genet. 1993, 5, 17-21. (22) Haaf, T.; Ward, D. C. Hum. Mol. Genet. 1994, 3, 629-633. (23) Bengtsson, U.; Altherr, M. R.; Wasmuth, J. J.; Winokur, S. T. Hum. Mol. Genet. 1994, 3, 1801-1805. (24) Florijn, R. J.; Bonden, L. A. J.; Vrolijk, H.; Wiegant, J.; Vaandrager, J.-W.; Baas, F.; den Dunnen, J. T.; Tanke, H. J.; van Ommen, G.-J. B.; Raap, A. K. Hum. Mol. Genet. 1995, 4, 831-836. (25) Rosenberg, C.; Florijn, R. J.; van de Rijke, F. M.; Blonden, L. A. J.; Raap, T. K.; van Ommen, G.-J. B.; den Dunnen, J. T. Nature Genet. 1995, 10, 477479. (26) Heiskanen, M.; Hellsten, E.; Kallioniemi, O.-P.; Makela, T. P.; Alitalo, K.; Peltonen, L.; Palotie, A. Genomics 1995, 30, 31-36. (27) Weier, H.-U. G.; Wang, M.; Mullikin, J. C.; Zhu, Y.; Cheng, J.-F.; Greulich, K. M.; Bensimon, A.; Gray, J. W. Hum. Mol. Genet. 1995, 4, 1903-1910. (28) Michalet, X.; Ekong, R.; Fougerousse, F.; Rousseaux, S.; Schurra, C.; Hornigold, N.; van Slegtenhorst, M.; Wolfe, J.; Povey, S.; Beckmann, J. S.; Bensimon, A. Science 1997, 277, 1518-1523. (29) Samad, A.; Huff, E. J.; Cai, W.; Schwartz, D. C. Genome Res. 1995, 5, 1-4. (30) Schwartz, D. C.; Li, X.; Hernandez, L. I.; Ramnarain, S. P.; Huff, E. J.; Wang, Y.-K. Science 1993, 262, 110-114. (31) Wang, Y.-K.; Huff, E. J.; Schwartz, D. C. Proc. Natl. Acad. Sci. U.S.A. 1995, 92, 165-169. (32) Meng, X.; Benson, K.; Chada, K.; Huff, E. J.; Schwartz, D. C. Nature Genet. 1995, 9, 432-438. (33) Cai, W.; Aburatani, H.; Stanton, V. P.; Housman, D. E.; Wang, Y.-K.; Schwartz, D. C. Proc. Natl. Acad. Sci. U.S.A. 1995, 92, 5164-5168. (34) Yokota, H.; Johnson, F.; Lu, H.; Robinson, R. M.; Belu, A. M.; Garrison, M. D.; Ratner, B. D.; Trask, B. J.; Miller, D. L. Nucleic Acids Res. 1997, 25, 1064-1070. (35) Yokota, H.; Nickerson, D. A.; Trask, B. J.; van den Engh, G.; Hirst, M.; Sadowski, I.; Aebersold, R. Anal. Biochem. 1998, 264, 158-164.

1664 Analytical Chemistry, Vol. 71, No. 9, May 1, 1999

Figure 1. DNA templates used in the study. (A) The 5.6-kbp plasmid 1631 DNA has one GAL4 binding site 2.1 kbp apart from the left end. (B) The 19.0-kbp λDNA fragment digested by StuI has one Junbinding site A, 6.7 kbp from the left end, and 48.5-kbp λDNA contains three Jun-binding sites A, B and C, located 19.1 kbp, 36.6, and 41.9 kbp from the end, respectively. The distances normalized to the total DNA length for the predicted Jun binding sites are shown.

binding sites), represented by the sequence 5′-TGAGTCA-3′, at sites A, B, and C, which are 19.1, 36.6, and 41.9 kbp from the end, respectively. The 19.0-kbp StuI-digested λDNA fragment has one AP-1 consensus sequence at site A (6.7 kbp from the left end). A 50-ng sample of DNA template in a 50-µL total volume was incubated with 0.1 µL of Jun protein stock (human c-jun produced in Escherichia coli with a total protein concentration of 550 ng/ µL, E3061, Promega) in a buffer containing 20 mM Hepes (pH 7.5), 5 mM MgCl2, 10% glycerol, 10 mM ZnCl2, 50 mM NaCl, 0.1 mM DTT, and 0.1 mM EDTA at 22 °C for 30 min. After incubation for forming Jun-λDNA complexes, 5 µL of 10 mM DNA dye, YOYO-1 (Molecular Probes Inc.), was added. Straightening and Immobilization of Protein-DNA Complexes. We straightened protein-DNA complexes using a linear receding meniscus as previously described.34,35 In brief, a freshly cleaved mica sheet (25 mm × 25 mm, Ted Pella Inc.) was soaked in 100 mM MgCl2 solution for 1 min, rinsed in clean running water (Millipore Ultrapure water system), and dried in a flow of air. Sample solution (10 µL) containing protein-DNA complexes was dispensed at the intersection of a mica plate and an 18-mm2 cover glass, and a wedge-shaped meniscus was moved on the mica surface by dragging the cover glass connected to a linear motor drive. The mica plate with straightened protein-DNA complexes was rinsed in clean running water and dried in a flow of air. Imaging GAL4-DNA Complexes by AFM. The mica sheet containing the GAL4-DNA complexes was cut into a 10-mm2 sheet and mounted on a 12-mm metal disk sample holder. We used a Nanoscope III AFM (Digital Instruments, Inc.) to capture the height images of individual GAL4-DNA complexes.35 The AFM with a Si3N4 tip was operated in the tapping mode at ∼200 kHz of a scan rate of 1-2 Hz. An artifactual dark streak appeared along a scanning line when the oscillating AFM tip encountered a relatively high object such as a GAL4 protein. AFM images shown are not modified, except for flattening to remove the

background curvature of the mica surface, and are displayed using NIH 1.60 image analysis software. Imaging Jun-λDNA Complexes by Fluorescence Microscopy. Mica sheets containing straightened Jun-λDNA complexes prepared as described above were further treated with ∼5 µL of antifade solution containing 90% glycerol and 0.1% phenyldiamine (pH 7.5). Fluorescently labeled molecules were imaged by a CCD camera (model TEA/CCD-1517-K/1, Princeton Instruments, Inc.) mounted on a Zeiss Axiophot fluorescence microscope with a 100×, 1.3 NA plan-Neofluar objective. The width of one pixel corresponded to 0.065 µm in the specimen. The field of view was 67 µm × 86 µm, and the exposure time was 5-10 s. The positions of sharp DNA bends within the DNA templates were determined from CCD images using IP Lab software (SignalAnalytics Corp.). For a DNA bend with two linearly extending DNA segments, the bend was defined as “sharp” when the angle formed by two DNA segments was less than 90°. The normalized sharp bend position was defined as d1/(d1 + d2), where d1 and d2 are an end-to-end length of DNA segments with d1 < d2. When more than one sharp DNA bend is observed, d1 and d2 are determined independently for each DNA bend. The normalized bend position ranges from 0 to 0.5, and it does not distinguish the site on the left half DNA from the right half. RESULTS Using AFM, we first examined GAL4-DNA complexes to verify the presence of GAL4 protein at the sharp bend on the straightened protein-DNA complex. Using fluorescence microscopy, we next examined Jun-λDNA complexes and analyzed the correlation between the site of the observed sharp bend on stretched protein-DNA complexes and three Jun-binding sites. Characterization of the Sharp DNA Bend at the GAL4Binding Site. To investigate the morphology of protein-DNA complexes after stretching and immobilization on a flat surface using surface tension and to verify the presence of protein at the apex of the frequently observed bend, we examined topographical images of the GAL4-DNA complexes using AFM. In the absence of the GAL4 protein, the straightened plasmid DNA template containing one GAL4-binding site appeared as a straight line (data not shown). When the same DNA template was incubated with the GAL4 protein and the complexes were subsequently straightened and immobilized on a MgCl2-soaked mica plate, the AFM image showed a V-shaped complex with DNA segments of different lengths protruding from the apex. The AFM image showed that GAL4 protein was bound at the apex of the bent DNA with two DNA segments extending from the bound protein (Figure 2). Of five such V-shaped protein-DNA complexes imaged by AFM, each one showed the presence of a protein at the apex (data not shown). Thus, the AFM image of the GAL4DNA complex shows a clear landmark demonstrating a proteinbinding site as a sharp DNA bend. These results suggested that protein-binding sites on DNA templates could be determined by detecting the position of sharp DNA bends in straightened, immobilized DNA. Since straightened long DNA is more easily detected by fluorescence microscopy than with AFM, we next used fluorescence microscopy for the analysis of DNA-protein complexes. Observation of a Sharp DNA Bend on λDNA Fragments at the Jun-Binding Consensus Sequence. To validate the use

Figure 2. AFM image of GAL4-DNA complex extended by surface tension. The image is illustrated in a three-dimensional coordinates to highlight the position of the GAL4-binding site. The dark streak next to GAL4 is an artifact created by a high object on scanning lines. The normalized position of the GAL4 binding site is determined as 0.38 by measuring the length of two DNA segments, in agreement with the predicted value of 0.38 (2.1 kbp/5.6 kbp). Arrowheads indicate the ends of two DNA segments. The DNA image taken by AFM in a dry condition usually gives a smaller height than the expected value of 2.3 nm.

of sharp bends in straightened, immobilized DNA as an indicator for protein binding, we investigated the correlation of an observed DNA bend position to a predicted protein-binding site by fluorescence microscopy. We examined the protein-DNA complexes consisting of a 19.0-kbp StuI-digested λDNA fragment and human c-Jun proteins (Figure 1B). A total of 302 fluorescently labeled DNA images that contained a sharp DNA bend were captured. Measurements of the length of two DNA segments, d1 and d2 (d1 < d2), were performed and used to determine the normalized bend position between 0 and 0.5 such as d1/(d1 + d2). Since neither end of the λDNA fragment was labeled, the normalized bend position did not distinguish a bend site near one end from a symmetrically located bend near the other end. The histogram (Figure 3) with an increment of 0.04 (∼0.8 kbp) shows that a number of DNA molecules have a normalized bend position between 0 and 0.5. The graph displays a peak frequency at the normalized position of 0.36. The value of 0.36 implies that a sharp DNA bend is most frequently found 6.8 kbp from one end of the 19.0-kbp λDNA fragment. Although there is a directional ambiguity and the frequency distribution is relatively broad, the measured distance of 6.8 kbp is in good agreement to 6.7 kbp, which is the distance of the AP1 consensus sequence from the left end of the DNA segment (Figure 1B). Localization of Protein-Binding Sites on DNA Templates with Multiple Binding Sites. To investigate whether the binding of multiple proteins to the same DNA template can be detected and localized by the position of sharp DNA bends on straightened protein-DNA complexes, we next examined Jun-λDNA complexes. A 48.5-kbp λDNA segment containing three AP-1 consensus binding sequences was used (Figure 1B). The entire λDNA was incubated with Jun proteins, processed as described above, and observed by fluorescence microscopy. A total of 398 fluorescence images of straightened DNA molecules containing one or two sharp DNA bends was acquired. Typical images are shown in Figure 4, and an analysis of the data is shown in Figure 5. The probability of finding a DNA molecule with one or two sharp bends Analytical Chemistry, Vol. 71, No. 9, May 1, 1999

1665

0.40 (19.4 or 29.1 kbp), where the numbers in parentheses indicate the observed physical distance of Jun-binding site from the left end (bp-1) of λDNA. The increment of the histogram, i.e., 0.02, corresponds to the resolution of this experiment. This resolution can be enhanced by increasing a sample number. Again there is a directional ambiguity, but the measured positions of 19.4, 35.9, and 41.7 kbp are in good agreement to three actual AP-1 consensus sites at 19.1, 36.6, and 41.9 kbp from bp-1. Thus, the results show that multiple protein-binding sites can be determined simultaneously using large intact DNA molecules.

Figure 3. Histogram showing the normalized position (0-0.5) of sharp DNA bends on Jun-StuI-digested λDNA complexes. An arrow at 0.37 indicates the known position of the Jun-binding site, a binding site located 7.0 kbp from the left end or at a fractional position of 0.37.

Figure 4. A gallery of fluorescent CCD images of Jun-λDNA complexes with a sharp DNA bend. The bar is 10 µm.

Figure 5. Histogram showing the normalized position (0-0.5) of sharp DNA bends on Jun-λDNA complexes. Three arrows at 0.14, 0.25, and 0.39 mark an actual position of three Jun-binding sites at 14%, 25%, and 39% of the length from one or the other end of λDNA.

was ∼10% and ∼1%, respectively. We did not detect a DNA molecule with more than two sharp DNA bends. From the measured length of the DNA segments protruding from the apex, we determined the normalized position of sharp DNA bends. A histogram representing the distribution of the normalized DNA bend positions is shown in Figure 5. The histogram had an increment of 0.02 (∼1 kbp) and exhibited three major peaks at 0.14 (6.8 or 41.7 kbp), 0.26 (12.6 or 35.9 kbp), and 1666 Analytical Chemistry, Vol. 71, No. 9, May 1, 1999

DISCUSSION This paper presents a novel fluorescence-based method for the determination of protein-binding sites on long DNA segments from the position of sharp DNA bends created in a straightening and immobilizing process of the protein-DNA complexes. In the described method, DNA molecules were incubated with unlabeled DNA-binding proteins, and the protein-DNA complexes were straightened on a flat surface by surface tension. DNA molecules without any bound protein appeared as straight segments. In contrast, DNA molecules incubated with DNA-binding proteins frequently exhibited a sharp DNA bend that had a peak distribution at the protein-binding site expected from the nucleotide sequence. The molecules were observed by AFM or fluorescence microscopy, and the binding sites were determined from the length of the DNA segments protruding from the apex. Here, we present the AFM images of a GAL4-DNA complex with GAL4 protein positioned at the apex of the bent DNA and the fluorescent images of Jun-λDNA complexes containing sharp bends at Junbinding sites. The described fluorescence-based method is expected to provide a simple tool for building protein-binding site maps. For identifying protein-binding sites, the described method complements conventional tools such as EMSA and DNase footprinting or an emerging AFM-based imaging method.36-40 Unlike EMSA or DNase footprinting, the topographical method proposed here enables us to use large DNA templates. The DNA from a bacterial artificial chromosome of >100 kbp in length, for instance, can easily be straightened by the device we have developed. Compared to the direct detection of bound proteins by the AFM-based method,41 the fluorescence-based method offers an indirect way, but the detection is rapid and easy. The resolution in position and binding specificity can be enhanced by a larger sample number than the AFM-based method. Since neither end of the DNA molecule was labeled in the experiments reported here, a protein-binding site had a directional ambiguity. It is possible to remove this ambiguity by cutting the bent DNA unevenly using restriction endonucleases in a parallel experiment.29-34 (36) Bustamante, C.; Rivetti, C. Annu. Rev. Biophys. Biomol. Struct. 1996, 25, 395-429. (37) Hansma, H. G.; Kim, K. J.; Laney, D. E.; Garcia, R. A.; Argaman, M.; Allen, M. J.; Parsons, S. M. J. Struct. Biol. 1997, 119, 99-108. (38) Hu, J.; Wang, M.; Weier, H. G.; Frantz, P.; Kolbe, W.; Ogletree, D. F.; Salmeron, M. Langmuir 1996, 12, 1697-1700. (39) Allison, D. P.; Kerper, P. S.; Doktycz, M. J.; Spain, J. A.; Modrich, P.; Larimer, F. W.; Thundat, T.; Warmack, R. J. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 8826-8829. (40) Allison, D. P.; Kerper, P. S.; Doktycz, M. J.; Thundat, T.; Modrich, P.; Larimer, F. W.; Johnson, D. K.; Hoyt, P. R.; Mucenski, M. L.; Warmack, R. J. Genomics 1997, 41, 379-384. (41) Binnig, G.; Quate, C. F.; Gerber C. Phys. Rev. Lett. 1986, 56, 930-933.

The physical basis for the formation of the sharp DNA bend needs to be further investigated. In the DNA-straightening force field, the bend may be a preferred conformation, or protein molecules such as GAL4 and Jun may have higher affinity to the MgCl2-soaked mica surface than DNA. Alternatively, the sharp bend may result from a local DNA bend induced by protein binding. However, the amount of DNA bending documented by protein binding is relatively small.42-44 Furthermore, the binding angles observed in the force field were not uniform and were similar even when the different DNA-binding proteins were tested. We therefore conclude that the bend is formed by interaction of the protein-DNA complex with the shear force. Consequently, the method should be general and not dependent on the ability of a protein to induce a bend in DNA. Although it remains to be determined whether a broad panel of proteins exhibits a similar DNA bend on a variety of surface substrates, a clear landmark of protein-binding sites was observed using two DNA-binding transcription factors. Since the presence of other proteins in the Jun protein stock does not interfere with the specific DNA-bend formation, we think that the method described here can be applied to determine specific binding sites even in the presence of nonspecific proteins such as are present in nuclear extracts. Obstacles on the surface probably create nonspecific DNA bends as in gel electrophoresis where DNA molecules alternately contract and lengthen and they often hook around obstacles.6 Besides three major peaks corresponding to the predicted Jun binding sites, there were four minor peaks in Figure 5. These minor peaks may represent nonspecific DNA bends or weak DNA-binding sites. The peak height may be used as an indicator of relative binding affinity, when a larger number of molecules than the number treated in this report is analyzed. The relatively broad distribution in Figure 3 using the shorter

19.0-kbp DNA fragment may result from the inaccuracy of length determination by fluorescence microscopy. The fact that determination of the length of the longer DNA molecule is easier allows the described method to cover long regulatory DNA segments. In conclusion, we describe a simple topographical method of building a protein-binding-site map from individual protein-DNA complexes. Using GAL4-DNA complexes, we demonstrate a clear AFM image showing a sharp DNA bend at the position of bound protein. Histograms of DNA bend indicating the frequency positions of bend location as determined by fluorescence microscopy are in good agreement with the predicted protein-binding sites. In elucidating complex molecular and cellular mechanisms, localizing protein-binding sites is critical. An efficient experimental tool for building a protein-binding-site map becomes increasingly important for determining functional significance in accumulating DNA sequences.45,46 The topographical technique described here is expected to provide an easy and reliable tool for investigating DNA and protein interactions.

(42) Heitman, J. BioEssays 1992, 14, 445-454. (43) Rees, W. A.; Keller, R. W.; Vesenka, J. P.; Yang, G.; Bustamante, C. Science 1993, 260, 1646-1649. (44) Werner, M. H.; Gronenborn, A. M.; Clore G. M. Science 1996, 271, 778784.

AC981370X

ACKNOWLEDGMENT We thank Ivan Sadowski and Martin Hirst (University of British Columbia, Canada) for a kind gift of GAL4 proteins and DNA templates containing a GAL4-binding site, and James Sunwoo (University of Washington) for critical reading of the manuscript. This work was supported in part by the National Science Foundation, Science and Technology Center for Molecular Biotechnology (NSFBIR9214821AM) and Sumitomo Electric Industries.

Received for review December 11, 1998. Accepted March 12, 1999.

(45) Rowen, L.; Mahairas, G.; Hood, L. Science 1997, 278, 605-607. (46) Hieter, P.; Boguski, M. Science 1997, 278, 601-602.

Analytical Chemistry, Vol. 71, No. 9, May 1, 1999

1667