High mobility group proteins HMG-I(Y) and HMG-1, as well as histone H1, all share the common property of binding to four-way junction DNA (4H), a synthetic substrate commonly used to study proteins involved in recognizing and resolving Holliday-type junctions formed during in vivo genetic recombination events. The structure of 4H has also been hypothesized to mimic the DNA crossovers occurring at, or near, the entrance and exit sites on the nucleosome. Furthermore, upon binding to either duplex DNA or chromatin, all three of these nuclear proteins share the ability to significantly alter the structure of bound substrates. In order to further elucidate their substrate binding abilities, electrophoretic mobility shift assays were employed to investigate the relative binding capabilities of HMG-I(Y), HMG-1 and H1 to 4H in vitro. Data indicate a definite hierarchy of binding preference by these proteins for 4H, with HMG-I(Y) having the highest affinity (Kd ~6.5 nM) when compared with either H1 (Kd ~16 nM) or HMG-1 (Kd ~80 nM). Competition/titration assays demonstrated that all three proteins bind most tightly to the same site on 4H. Hydroxyl radical footprinting identified the strongest site for binding of HMG-I(Y), and presumably for the other proteins as well, to be at the center of 4H. Together these in vitro results demonstrate that HMG-I(Y) and H1 are co- dominant over HMG-1 for binding to the central crossover region of 4H and suggest that in vivo both of these proteins may exert a dominant effect over HMG-1 in recognizing and binding to altered DNA structures, such as Holliday junctions, that have conformations similar to 4H.
Recognition and alteration of DNA structure plays a significant biological role in regulating transcription, replication, genetic recombination and repair in both prokaryotes and eukaryotes. In eukaryotes the overall folded structure of DNA is determined primarily by its association with `core' and H1-type linker histones to form nucleosomal chromatin (1 ,2 ), but with additional and often localized contributions being made by both transcription factors and ancillary chromatin proteins such as members of the high mobility group (HMG) proteins.
Histone H1 (Mr ~23 kDa) and its variants (e.g H5, Hlo, etc.) are commonly referred to as linker histones because of their ability to preferentially bind to the linker DNA of nucleosomes (1 -4 ), although their exact placement with respect to the nucleosome core particle is still unknown (5 -13 ). The crystallographic structure of the globular domain of histone H5 (6 ) indicates that its three [alpha]-helices assume a `winged' configuration with two potential regions of contact with DNA, a prediction confirmed by recent mutagenesis studies (11 ). The linker histones play a pivotal role in chromatin compaction (recently reviewed in 4 ) and can exert either a positive or negative role on gene transcription both in vitro (14 ,15 ) and in vivo (16 ,17 ). H1 binds to DNA in a sequence- independent fashion yet has been shown to preferentially bind to supercoiled plasmids over either linear or relaxed DNAs (18 -20 ). In addition, H1 is known to preferentially bind to certain altered DNA structures such as four-way junctions (4H) (21 ) [The nomenclature for four-way junction used here is that defined in Lilley et al. (22 ) and in Materials and Methods.] H1 not only recognizes certain altered forms of DNA but is able to induce changes in DNA structure, as demonstrated by its ability to both unwind naked DNA (23 ,24 ) and to compact chromatin (1 -3 ).
HMG-1/-2 proteins (Mr ~25 kDa) are the largest and most abundant of the HMG proteins and have been implicated in both positive and negative regulation of gene transcriptional activity in vitro and in vivo, but their precise biological functions are still uncertain (reviewed in 25 ). They interact with DNA through their two DNA binding domains, known as HMG-1 boxes, a conserved set of amino acids folding into three [alpha]-helices forming an L-shaped structure (26 ,27 ). HMG-1/-2 proteins bind to the minor groove (28 ) of double-stranded DNA in a sequence-independent manner (25 ). HMG-1/-2 binding affects DNA structure by inducing bends in linear substrates (29 -31 ) and by introducing supercoils into topologically constrained molecules (32 -34 ). Recently HMG-1 has been shown to preferentially bind to altered DNA structures which contain sharp bends, such as 4H (35 ) and those found in cisplatin-DNA adducts (36 ).
HMG-I (Mr ~11.7 kDa) and HMG-Y (Mr ~10.5 kDa) are isoform proteins produced by alternative splicing of mRNA transcripts from a single gene (37 ,38 ). For convenience and to distinguish them from HMG-1/-2 proteins, members of this family will be referred to as HMG-I(Y). HMG-I(Y) proteins have been demonstrated to be involved in both positive and negative gene regulation, possibly by functioning as accessory `architectural transcription factors' (reviewed in 25 ). HMG-I(Y) proteins preferentially bind to the minor groove of A[middot]T-rich B-form DNA by recognition of substrate structure rather than nucleotide sequence (25 ,39 -44 ). In vivo HMG-I(Y) recognizes the structural features of DNA through its DNA binding domains, known as A[middot]T hooks, which also bind to altered DNA structures such as 4H (45 ), those found on the front face of native (random sequence) nucleosome core particles (46 ) and on the surface of nucleosomes reconstituted from defined sequence DNAs (40 ). In addition to recognizing structure, HMG-I(Y) has the ability to induce structural changes in DNA substrates. For example, HMG-I(Y) can unwind DNA, induce both positive and negative supercoils into topologically constrained plasmid DNAs (47 ) and introduce bends and/or other distortions into linear substrate molecules (48 ; G.Schroth and R.Reeves, unpublished data).
Even though the three-dimensional structures of the DNA binding domains of the H1/H5 linker histones (6 ), of the HMG-1/-2 proteins (26 ,27 ,31 ) and of HMG-I(Y) (49 -51 ) are all quite different, they share many common DNA binding characteristics, including the ability to bind 4H substrates. The structure of 4H has been rigorously analyzed and many structural features elucidated and characterized (reviewed in 52 ).
Although synthetic 4H substrates are not associated with core histones, they do contain two converging DNA strands which have been proposed to imitate the structure of DNA found at, or near, the entrance and exit points of the nucleosome and have therefore been suggested as a simple model system for studying chromosomal proteins that bind to such regions (4 ,21 ,30 ,53 ). Perhaps more importantly, 4H simulates in vivo Holliday junctions and has therefore been used extensively as an intermediate substrate for genetic recombination events (52 ,54 -58 ).
With these potential biological contexts in mind, we performed quantitative in vitro binding and competition experiments to investigate the relative strengths and specificities of the physical interactions of HMG-I(Y), HMG-1 and H1 proteins with 4H. Our results demonstrate that HMG-I(Y) binds most tightly to 4H with a Kd of ~6.5 nM, followed by H1 and HMG-1 with Kd values of ~16 and ~80 nM respectively. We also demonstrate that HMG-I(Y) is far more effective in competing with H1 than is HMG-1 for binding to 4H substrates. Hydroxyl radical footprinting of HMG-I(Y) on 4H indicates that the protein protects 4H at the crossover, thereby demonstrating for the first time that HMG-I(Y) preferentially binds to the altered structure at the center region of the 4H structure. Our findings suggest that HMG-I(Y) and H1 exert a dominant effect over HMG-1 in binding to altered DNA structures in vivo and therefore have important implications for their possible biological roles in regulation of chromatin structure and function, as well as for participation in recombination and integration events.
Oligonucleotides 1-4 previously described by Bianchi (35 ) were used to create 4H:
Leg 1, CCCTATACCCCTGCATTGAATTCCAGTCTGATAA;
Leg 2, GTAGTCGTGATAGGTGCAGGGGTTATAGG;
Leg 3, AACAGTAGCTCTTATTCGAGCTCGCGCCCTATCACGACTA;
Leg 4, TTTATCAGACTGGAATTCAAGCGCGAGCTCGAATAGAGCTACTGT.
The oligonucleotides were synthesized using an Applied Biosystems DNA synthesizer and subsequently purified by electrophoresis on a denaturing 1.5 mm thick, 1* TBE, 16% polyacrylamide (19:1 bisacrylamide) gel (59 ). Complete 4H, as well as incomplete junctions, were formed by annealing equal molar amounts of the appropriate gel-purified oligonucleotides in annealing buffer (10 mM Tris, pH 7.2, 50 mM NaCl, 10 mM MgCl2) and treating the products as previously described (35 ). Prior to annealing, if needed, one of the oligonucleotides was 5'-end-labeled using T4 kinase. Stable formation of complete 4H (as well as incomplete junctions) was demonstrated by gel electrophoresis (data not shown). Nomenclature for four-way DNA junctions are defined in Lilley et al. (22 ) to be HHHH or 4H, representing four helixes converging to a central point. However, in addition to using complete 4H constructs in this study we also used incomplete junction constructs, i.e. those lacking one or more legs. The standardized nomenclature for our incomplete junction molecules is as follows: incomplete junction formed by annealing only oligonucleotides 3 and 4 (legs 3:4), HS14/S20; the incomplete junction formed by annealing only oligonucleotides 1, 3 and 4 (legs 1:3:4), 2HS14/S16. This nomenclature indicates that there are either one or two converging DNA helixes with two single-strand extensions, the length of which are numerically indicated as subscripts (see Fig. 3 for an illustration).
HMG-1 and H1 proteins were purified from calf thymus and fractionated using a polybuffer BPE-94 column as described in detail by Adachi et al. (60 ). Recombinant human HMG-I(Y) was prepared as previously described by Nissen et al. (61 ).
Electrophoretic mobility shift assays (EMSA) were performed by described procedures (59 ) by incubating 8 or 17 fmol labeled 4H with a target protein at room temperature in a total volume of 20 [mu]l protein binding buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA, 25 mM NaCl, 0.28 [mu]g sheared salmon sperm DNA, 2.8 [mu]g BSA). Competitions were performed by adding unlabeled competitor DNA (either 4H or incomplete junctions) or competitor proteins to the reaction and incubating the combined mixture for an additional 10 min prior to electrophoresis. Glycerol was then added to 2% final concentration and the samples loaded onto a polyacrylamide gel (1.5 mm thick, 15 * 15 cm and comprised of 0.5* TBE, 4 or 6.5% polyacrylamide at 29:1 bisacrylamide). Gels were pre-run for at least 1 h at 10 V/cm and the electrode buffer replaced before loading samples. Samples were electrophoresed at 10 V/cm for 3-6 h at 4oC, following which the gels were dried on Whatman filter paper and exposed to either Amersham hyperfilm or to a Molecular Dynamics (Sunnyvale, CA) phosphorimager screen. Visual reproductions of the gels were created using a Textronix Phaser 440 printer. Quantification of band densities was performed using ImageQuant Peakfinder (Molecular Dynamics Corp., Sunnyvale, CA) to locate the bands and then the individual peaks were deconvoluted and analyzed using Peakfit (Jandel Scientific Corp., Corte Madera, CA). Further analysis, manipulations or plots were performed using Excel (Microsoft), Sigmaplot (Jandel Scientific) or Enzfitter (Niles & Associates, Berkeley, CA) software.
Disassociation constants for HMG-I(Y), HMG-1 and H1 were determined by evaluating the relative band densities from EMSAs produced by titrating 4H with HMG-I(Y), HMG-1 and H1 (Fig. 1 A-C). Prior to electrophoresis the reaction mixtures were incubated sufficiently long to reach equilibrium, so standard equilibrium equations describing ligand-substrate complex formation were used (62 ,63 ). The equilibrium equation used to describe the first, or tightest binding, DNA-protein complex (bands a, c and e in Fig. 1 ) is P + A <-> PA, where P is the substrate (4H), A is the ligand (protein) and PA is the substrate-ligand complex. The average number of bound ligand molecules per substrate molecule is v, where v = [PA]/[P] + [PA] and is related to the dissociation constant Kd by the equation v = ([A]/Kd)/(1+[A]/Kd). Kd determinations were confirmed by plotting v versus [A], which produces a rectangular hyperbola that approaches vmax (total ligands bound per substrate molecule). Extrapolation through the curve at [1/2]vmax gives the Kd for the test protein (62 ,63 ).
Hydroxyl radical footprinting of HMG-I(Y) on 4H was carried out following published protocols (64 ). Each 100 [mu]l reaction mixture containing 0.068 nM labeled 4H and enough HMG-I(Y) to form either one or two DNA-protein complexes was incubated at room temperature for 10 min prior to sequential addition of 4 [mu]l each of: (i) a freshly prepared mixture of 0.75 mM Fe(II) and 1.5 mM EDTA; (ii) 37 mM ascorbate; (iii) 1.25% hydrogen peroxide. The complete reaction mixture was incubated at room temperature for exactly 2 min and stopped by addition of glycerol to 5%. To demonstrate the formation of a 4H-HMG-I(Y) complex, 5 [mu]l of the mixture was loaded onto an EMSA gel (previously described) and electrophoresed. The DNA in the remaining reaction mixture was extracted with phenol:chloroform, dried, dissolved in formamide loading buffer, boiled for 5 min and then equal counts were loaded onto a 15% polyacrylamide sequencing type gel. The gels were electrophoresed at 18 000 V for 1.5 h, fixed in 8% acetic acid, 8% methanol, dried on Whatman paper and exposed to Amersham hyperfilm or a Molecular Dynamics phosphorimager screen.
Competition experiments were performed to determine the binding specificity of HMG-I(Y) to 4H. In these experiments HMG-I(Y) was incubated with labeled 4H and then titrated with increasing amounts of various unlabeled competitor DNAs. The competitor DNA molecules were either complete 4H or incomplete constructs, i.e. those lacking one or more legs (Fig. 3 ). In the first competition series (Fig. 3 A) 20 nM HMG-I(Y) was incubated with 4H to form only the first high affinity 4H-HMG-I(Y) complex (a), which was subsequently competed with either a 100- or 500-fold molar excess of unlabeled competitor DNA. These results clearly show that the 4H-HMG-I(Y) complex (a) is not effectively competed with either of the incomplete junctions HS14/S20 (legs 3:4) or 2HS14/S16 (legs 1:3:4) (Fig. 3 A, lanes 2-6), however, it is effectively competed with 4H (Fig. 3 A, lanes 7 and 8). Similar results were obtained when incomplete junctions containing other combinations of legs were used in the competition assays (data not shown). These competition EMSAs demonstrate that high affinity binding of HMG-I(Y) to 4H requires an intact crossover-containing structure and also show that the HMG-I(Y) protein does not bind tightly to either the single-stranded regions or the elbows of the incomplete junction molecules.
A second competition series (Fig. 3 B) was performed exactly like the first except that in this case a much higher concentration of HMG-I(Y) (70 nM) was initially incubated with labeled 4H to form both the high affinity (a) and the lower affinity (b) 4H-HMG-I(Y) complexes. This competition series demonstrates that the second 4H-HMG-I(Y) complex (b) is competed slightly with the incomplete junction HS14/S20 (legs 3:4) and to a greater extent with the incomplete junction 2HS14/S16 (legs 1:3:4). This result indicates that incomplete junctions possess some structural component that is weakly recognized by HMG-I(Y) but that this binding is not strong enough to dissociate the tightly bound protein from the first complex. The oligonucleotides used to form the incomplete 4Hs used in these competition assays contain a 5 bp stretch of A[middot]T residues, the minimal length of duplex DNA required for specific binding of the HMG-I(Y) protein (39 -44 ). However, the binding of HMG-I(Y) to this stretch of A[middot]T residues in the duplex of legs 3:4 is relatively weak since neither of the incomplete junction constructs effectively competed with the first 4H-HMG-I(Y) complex, as demonstrated in both Figure 3 A (lanes 2-6) and Figure 3 B (lanes 2-6). Only the complete 4H construct effectively competed with the high affinity 4H-HMG-I(Y) complex (a) (Fig. 3 B, lanes 7 and 8).
To investigate a possible hierarchy of binding of HMG-I(Y), HMG-1 and H1 we performed a number of competition EMSAs by first incubating labeled 4H with one protein and then titrating the preformed 4H-protein complex with a second protein. By design, in all of the competition experiments protein concentrations were such that only competition for binding to the highest affinity site for each protein was monitored. Two possible results were anticipated under these binding conditions: (i) direct competition between the titrated protein and the prebound protein for 4H binding, as evidenced by depletion of the original preformed DNA-protein complex band and concomitant appearance of a new complex band comprised of only the added competitor protein and 4H; (ii) simultaneous binding of the two proteins to the same 4H as evidenced by depletion of the preformed complex band and concomitant appearance of a new band representing a ternary complex comprised of 4H and both proteins.
The results of competition EMSAs between HMG-1 and H1 demonstrated that: (i) H1 and HMG-1 cannot bind simultaneously on the same 4H substrate; (ii) H1 can effectively displace HMG-1 from 4H (Fig. 4 ). The competition was conducted by incubating labeled 4H with only enough HMG-1 to form the first complex (c) and then titrating with increasing amounts of H1 (Fig. 4 A). This gel clearly shows that during titration the intensity of the 4H-HMG-1 complex band (c) decreases concomitantly with an increase in intensity of the band corresponding to the 4H-H1 complex (e). Experiments using higher initial concentrations of bound HMG-1 (e.g. 100 nM) were also conducted and produced similar results (data not shown), however, slightly higher amounts of H1 were necessary to completely compete the HMG-1 from the preformed complex, which is expected under the given equilibrium conditions (62 ,63 ,65 ). These results indicate that H1 and HMG-1 compete with each other for the same binding site on 4H and that H1 easily out-competes HMG-1 for this site. Plotting the relative band intensities of the 4H-HMG-1 complex (c) and the 4H-H1 complex (e) from Figure 4 A as a function of HMG-1:H1 molar ratio demonstrates that it takes an ~5 molar excess of HMG-1 over H1 to compete equally for the 4H substrate (Fig. 4 B). We also conclude from these experiments and the fact that the Kd of H1 for 4H is ~16 nM, whereas that of HMG-1 is ~80 nM (Fig. 1 ), that binding of the H1 and HMG-1 proteins to 4H is mutually exclusive. Furthermore, it is evident from these results that H1 is clearly dominant over HMG-1 in in vitro competition assays using 4H as substrate and they suggest that this may likewise be the case for binding of these proteins to DNA crossover structures existing in vivo.
Competition EMSAs between HMG-I(Y) and H1 demonstrated that HMG-I(Y) is very effective at competing with H1 for binding to 4H. Under equilibrium conditions 10 nM HMG-I(Y) was incubated with labeled 4H (i.e. enough protein to form only the tightest binding complex (a) and increasing amounts of H1 were then added (Fig. 5 A). The result of this competition clearly shows disappearance of 4H-HMG-I(Y) band a with a concomitant appearance of 4H-H1 band e, indicating that H1 directly competes with HMG-I(Y) for the same site on 4H and that binding to this site by the two proteins is mutually exclusive. Other titrations using higher HMG-I(Y) concentrations (20 nM) were also performed and similar results were obtained (data not shown), however, somewhat greater amounts of H1 were needed to completely compete with the preformed complex of HMG-I(Y), a result that is expected under equilibrium conditions (62 ,63 ,65 ). Again, as shown in Figure 5 B, we plotted the relative band densities versus the molar ratio of protein concentrations. This graph clearly indicates that H1 must be in at least a 2.5 molar excess over HMG-I(Y) to compete equally for substrate binding, a result consistent with the relative Kd values of these two proteins for 4H substrates (Fig. 1 ). It is therefore evident from these results that HMG-I(Y) is able to out-compete H1 for binding to 4H in vitro and, most likely, also in vivo.
Competition EMSAs were conducted using HMG-I(Y) and HMG-1 proteins. However, because the 4H-HMG-I(Y) and 4H-HMG-1 complexes have nearly identical mobilities under the several different gel electrophoretic conditions examined, we were unable to determine the precise HMG-I(Y):HMG-1 molar ratios contained in the shifted band, rendering the gel systems employed unsuitable for quantitative competition comparisons. Nevertheless, we did not observe any ternary complex formation in any of the competition titrations, indicating that HMG-I(Y) also competes with HMG-1 for binding to the high affinity site on 4H (data not shown).
Hydroxyl radical DNA cleavage experiments were employed to determine the placement of binding of HMG-I(Y) protein on 4H at base pair resolution. Our footprinting of HMG-I(Y) on 4H is the first of its kind, since, to our knowledge, no other mammalian HMG proteins have been footprinted on 4H. The results shown in Figures 6 -9 indicate that HMG-I(Y) binds most strongly to 4H near the junction of the four oligonucleotide strands. Figures 6 and 7 show the results of hydroxyl radical footprinting of a 1:1 4H-HMG-I(Y) complex (i.e. complex a in Fig. 1 ), whereas Figures 8 and 9 depict the results of footprinting a 1:2 4H-HMG-I(Y) complex (i.e. complex b in Fig. 1 ).
The results of a typical hydroxyl radical footprint of HMG-I(Y) on 4H can be seen in the autoradiogram shown in Figure 8 B, where 4H containing a single labeled oligonucleotide (i.e. 3) was incubated with enough HMG-I(Y) protein to form both high and low affinity DNA-protein complexes (i.e. complex b in Fig. 1 ). The band intensities from the hydroxyl radical cleavage reactions of naked 4H (lane 2) and protein-complexed 4H-HMG-I(Y) (lane 3) were quantitatively analyzed by densitometry, normalized and the resulting scans overlaid to produce the composite results shown in Figure 8 A. The bars indicate the regions of greatest protection of 4H by HMG-I(Y) from cleavage and the dots represent less protected regions. All four legs of the oligonucleotide strands comprising the 4H structure were independently radiolabeled, formed into DNA-protein complexes, subjected to hydroxyl radical footprinting, analyzed and the composite results of numerous such experiments are depicted in Figure 9 .
Four-way junction DNA was used as a substrate in our in vitro protein binding studies because it is thought to mimic the Holliday-type recombination crossover structures found in vivo (52 ,66 -69 ) and also because it has been hypothesized to simulate the structure of the linker DNA strands near the entrance and exit points of nucleosomes (21 ,30 ,70 ). With these biological connections in mind, it is significant that the HMG-1/-2 proteins have recently been shown to participate in genetic recombination events during immunoglobulin gene rearrangements in vitro (71 ). Because HMG-I(Y) out-competes HMG-1 for binding to 4H it is also reasonable to expect that HMG-I(Y) may likewise facilitate genetic recombination in vitro. In addition, HMG proteins have recently been demonstrated to be required for integration of retroviral cDNAs into host cell DNA substrates in vitro, a process that in many respects is homologous to recombination events involving 4H structures. For example, HMG-1 has been shown to be required for efficient integration of avian sarcoma virus into host cell DNA in vitro (72 ) and, likewise, HMG-I(Y) has been demonstrated to be required for integration of HIV-1 cDNA into DNA in vitro (73 ). Although little information is currently available, in the light of our present findings it seems reasonable to suspect that H1 might likewise participate in genetic recombination and/or retroviral integration events.
As discussed above, the HMG-I(Y), HMG-1 and H1 proteins have all been demonstrated to specifically associate with nucleosomal chromatin in vitro and 4H has been suggested to structurally resemble linker DNA near the entrance and exit points of nucleosomes. Nevertheless, results from our quantitative binding studies for each of these individual proteins to 4H do not closely correlate with the previously reported binding affinities of these proteins to nucleosomal substrates. For example, we find that H1 binds to 4H with a Kd of ~16 nM, which is similar to the Kd value of ~18 nM reported by others for H1 binding to naked B-form DNA in vitro (14 ). In contrast, H1 has been reported to have a higher in vitro binding affinity for both reconstituted Xenopus 5S rDNA mononucleosomes (Kd ~2 nM) (10 ) and dinucleosomes (Kd ~7.4 nM) (14 ). In the case of HMG-1 we find that it binds to 4H with a Kd of ~80 nM, whereas HMG-1 has been reported to have a much lower affinity for reconstituted 5S rDNA dinucleosomes (Kd ~300 nM) (14 ). In marked contrast to histone H1, HMG-I(Y) protein binds more tightly to 4H (Kd ~6.5 nM) than to random sequence nucleosome core particles (Kd ~50 nM) (46 ). Thus our data suggest that in vitro the binding of HMG-I(Y), HMG-1 and H1 to 4H probably more closely relates to in vivo biological events such as genetic recombination or retroviral integration than to binding of these proteins to linker DNA near the entrance and exit points of nucleosomes.
Our protein binding competition studies are also consistent with a previous report indicating that HMG-I(Y) has the ability to displace H1 from duplex linear B-form DNA containing A[middot]T-rich MAR/SAR sequences (74 ,75 ). In combination, these findings indicate that HMG-I(Y) is able to out-compete H1 for binding to different substrates, whether they are the A[middot]T-rich, B-form DNA sequences found in MAR/SAR regions or the kinked, distorted DNA structures found at the crossover of 4H. In both cases HMG-I(Y) has the capacity to effectively displace H1 from such substrates. These observations are significant because DNA in the living cell is believed capable of assuming many different configurations and it is likely that, regardless of its biological form, HMG-I(Y) will have the ability in vivo to out-compete, or displace, H1 from such DNA structures.
Finally, it should also be noted that our hydroxyl radical footprints of the high affinity binding site of HMG-I(Y) on 4H are consistent with a previously reported footprint of a prokaryotic HMG-1-like protein, HU, on 4H. In this instance HU, which associates as a dimer, binds on opposite sides of the crossover region of 4H (76 ). Both our results and those with HU protein (76 ) are consistent with the suggestion by Lilley (30 ,52 ) that 4H in vitro has a 2-fold symmetrical structure that provides independent binding sites, located on opposite sides of the junction, for structure-recognizing proteins.
This work was supported by National Science Foundation grant MCB-9506878.
*To whom correspondence should be addressed at: Department of Biochemistry and Biophysics, Washington State University, Pullman, WA 99164-4660, USA. Tel: +1 509 335 1948; Fax: +1 509 335 9688; Email: reevesr@mail.wsu.edu
REFERENCES