DNA intersegment transfer, how steroid receptors search for a target site.

The mammalian nucleus contains 6 billion base pairs of DNA, encoding about 100,000 genes, yet in a given cell steroid hormones induce only a handful of genes. The logistical difficulties faced by steroid receptors or other transcription factors of sorting through this much genetic information is further increased by the density of nuclear DNA (approximately 10-50 mg/ml). Standard models propose that steroid receptors find target elements by repeated cycles of dissociation and reassociation until a high affinity site is found (cycling model) and/or by conducting a one-dimensional search along the DNA (sliding model). A third model proposes that steroid receptors search for target sites in the genome by DNA intersegment transfer. In this model, receptor dimers bind nonspecific DNA sequences and search for a target site by binding a second strand of DNA before dissociating from the first, in effect moving through the genome like Tarzan swinging from vine to vine. This model has the advantage that a high concentration of DNA favors, rather than hinders, the search. The intersegment transfer model predicts, in contrast to the cycling and sliding models, that the dissociation rate of receptor from DNA is highly dependent on DNA concentration. We have employed the purified DNA binding domain fragment from the rat glucocorticoid receptor to perform equilibrium and kinetic studies of the DNA dependence of receptor-DNA dissociation. We find receptor dissociation from DNA to be highly dependent on the concentration of DNA in solution, in agreement with the intersegment transfer model. We also find that this interaction is primarily electrostatic, because DNA-like polyanion chains (e.g. heparin and polyglutamate) can mediate the transfer. These studies provide evidence that direct DNA transfer aids the target site search conducted by steroid receptors in their role as inducible transcription factors.

Many eukaryotic transcription factors function by binding to specific DNA sequences upstream of transcription start sites. In order to bind these DNA elements, the protein must first locate these sites in the genome. In humans this search entails locating a few functional binding sites from over 6 billion base pairs of DNA, requiring the protein to sample a vast number of possible binding sites in a very short period of time. Neverthe-less, transcription factors are able to find their specific binding sites very rapidly. What are the mechanisms used to locate specific binding sites in chromatin?
One model envisions sampling of potential target sites by repeated cycles of dissociation/reassociation of proteins with DNA. This mechanism is commonly assumed to be the search strategy used by site-specific DNA-binding proteins. The theoretical underpinnings of this model were developed in the 1970s (see Ref. 1 for review), describing the situation where the protein binds nonspecifically to the DNA, dissociates at some intrinsic rate, and then diffuses through the aqueous medium to sample a second site. This cycle is repeated until a high affinity target site is found. This model predicts that the rate of search will be controlled solely by the intrinsic rate of protein dissociation from nonspecific DNA sites and the speed of diffusion through the nuclear volume.
The need for alternative models was made apparent by the early observations of Riggs et al. (2), indicating that the association rate for Escherichia coli lac repressor binding to its regulatory element was several orders of magnitude faster than that allowed by a simple diffusion-controlled mechanism (3,4). Further experiments with lac repressor-operator (5,6) and E. coli CAP protein (7) confirmed and extended this observation.
To account for this discrepancy, a second model for protein-DNA search was devised. This strategy describes an intersegment transfer mechanism requiring two binding sites on the protein (or a protein dimer with one site on each monomer). One of the binding sites releases while the other remains bound, allowing the free binding site to interact with a second strand of DNA. This model, termed intersegment transfer (4,8), predicts that increasing concentrations of DNA will increase the apparent dissociation rate of the protein; this is in contrast to the dissociation/reassociation model, in which the rate of protein dissociation is intrinsic and not affected by exogenous DNA.
The third mechanism involves the protein sliding along the DNA chain until it encounters a specific binding site. This model reduces the search space to a one-dimensional random walk and has the advantage that it is highly efficient for locating a binding site over a limited distance along the DNA molecule. This model has been successfully applied to restriction endonucleases (9,10) and the lac operator (11). This mechanism and the first two are not mutually exclusive. Indeed, it is likely that most proteins conduct a short range one-dimensional search in combination with one of the other search mechanisms. Theoretical work indicates that this mechanism alone is not an efficient strategy for a long distance search (12).
Our studies do not directly address the sliding model; instead, they seek to distinguish whether the intersegment transfer or the cycling search mechanism may be employed by the steroid receptor family of transcription factors. We have chosen to study the well characterized rat glucocorticoid receptor DNA binding domain protein fragment (GRdbd). 1 This fragment, which contains amino acids 440 -525 of the rat GR, contains two zinc finger motifs and has strong DNA sequence binding specificity (13). The solution structure of the GRdbd has been determined by NMR, and the crystal structure of the GRdbd bound to DNA has been deduced (14 -20). The GRdbd is composed of two zinc finger-stabilized ␣-helices. The first helix is responsible for specific DNA contacts (14), whereas the second is positioned away from the DNA at right angles to the first helix and making no DNA contacts at a target site. The GRdbd is a monomer in solution and has been shown to bind to DNA both as a monomer and cooperatively as a dimer (15,21). This paper addresses not only the question of whether GRdbd, as a model of the steroid receptor family of transcription factors, may employ an intersegment transfer mechanism as part of its target site search strategy but the mechanistic details of this transfer as well. These questions touch directly upon the ability of regulatory proteins to mediate their effects by rapidly locating specific binding sites in the genome.

SEQUENCE III
Gel Mobility Shift Assay-The conditions described by Alroy and Freedman (13) were adapted for these studies. Briefly, complementary strands of oligonucleotide DNA were allowed to slowly anneal for 2 h in TEK buffer from 90°C to room temperature. Annealed oligonucleotides were labeled with 30 Ci of [␣-32 P]dCTP (3000 Ci/mmol) and Klenow enzyme. Free counts were removed with a Sephadex G-25 spin column, followed by ethanol precipitation in 2.5 M ammonium acetate. The radioactive pellet was washed once with 70% ethanol, dried, and resuspended to 5 ng/l.
For a standard 20-l binding experiment, purified GRdbd (0.89 ng) and labeled DNA (2.0 ng) were incubated at a 1:1 molar ratio (4.5 nM each) unless otherwise noted, for 30 min at room temperature in GRdbd binding buffer (20 mM Tris-HCl, pH 7.9, 1 mM EDTA, 0.05% Nonidet P-40, 10% glycerol, 50 mM KCl, and 1 mM DTT). Dissociation experiments were conducted by dilution of GRdbd/DNA 100-fold in binding buffer containing 10-fold molar excess of the full binding site oligonu-cleotide to prevent reassociation of the GRdbd. Protein-DNA complexes were resolved on a precooled 10% nondenaturing polyacrylamide gel (75:1 acryl:bis) and run at 400 V for 2.5 h in 0.5 ϫ TBE buffer (40 mM Tris base, 1 mM EDTA, pH 8.0, 40 mM borate) at 0 -4°C. There was a delay time of ϳ15 s between dilution of complexes and the time the complexes entered the gel for the nominal zero time point. Gels were fixed for 10 min in 7% acetic acid followed by 10 min in 100% methanol (to prevent cracking). Dried gels were exposed to a phosphorimager screen for 2-18 h, and bound and free DNA were quantified on a Molecular Dynamics PhosphorImager, model 300E (Sunnyvale, CA).
Sheared E. coli DNA-A 50 ml overnight bacterial culture (DH5) in nutrient broth (22) was centrifuged at 3300 rpm (2000 ϫ g) for 10 min to form a pellet. Cells were resuspended in 13.3 ml of TE. To this was added 10% SDS to a final concentration of 0.5%, proteinase K to 200 g/ml, and RNase to 8 g/ml. The cells were incubated at 37°C for 1 h, and then NaCl was added to 0.7 M and polyethylene glycol 8000 to 0.5% final concentration; the solution was then shaken at 65°C for 10 min. The mixture was then left at room temperature for 30 min. The resulting DNA was extracted once with an equal volume of chloroform and three times with phenol/chloroform (1:1) and precipitated with 2 volumes of EtOH at Ϫ20°C overnight. The DNA pellet was centrifuged for 15 min at 14,000 ϫ g, washed with 70% ethanol, dried, and resuspended in TE. The A 260 was determined, and the DNA was diluted to 5 mg/ml. The DNA was sheared by 20 passages through a 25 gauge needle; it ranged in size from 500 to 20,000 bp as determined by gel electrophoresis on a 0.8% agarose gel.

Establishment of Conditions for Determination of Dissociation Constants-The
GRdbd has been shown to exist as a monomer in solution, but it can bind to DNA cooperatively as a dimer (13,(23)(24)(25)(26)(27)(28). This is presented schematically in Fig. 1. An intersegment transfer mechanism can be distinguished from cycling via repeated dissociation and reassociation with DNA by assessing whether the dissociation of the bound GRdbd from DNA is influenced by the concentration of DNA in solution. An intersegment transfer mechanism predicts that increasing DNA will increase the rate of dissociation, whereas increasing DNA will have no effect if a cycling mechanism is used. Experimentally, we measure the loss of the receptor-labeled DNA complex (R 2 D) with time at different concentrations of unlabeled DNA. First, GRdbd is bound to labeled DNA to equilibrium, and then disequilibrium conditions are imposed so that reestablishment of R 2 D* is negligible once dissociation has occurred. Generally, this is accomplished by blocking reassociation with a large molar excess of unlabeled DNA. However, this strategy is invalid if DNA itself influences the apparent dissociation constant. We therefore have determined the association and dissociation constants for monomer (k 1 , k Ϫ1 ) and dimer (k 2 , k Ϫ2 ) and established conditions wherein we can measure the apparent dissociation constant in the presence of different concentrations of DNA.
In Fig. 2 are presented equilibrium binding experiments at GRdbd concentrations ranging from 5 to 250 nM (1-50 ng) and oligonucleotides containing either a full glucocorticoid response FIG. 1. Kinetic description of the binding of the GRdbd to DNA and the intersegment transfer mechanism. R, GRdbd protein; D*, labeled specific oligonucleotide; D, unlabeled DNA. Association constants and dissociation constants are indicated by a lowercase k. The equilibrium constants are K 1 for monomer binding and K 2 for dimer binding. The apparent off-rate of GRdbd dimer from D* is a function of k Ϫ2 and k 3 . In the intersegment transfer model, unlike the cycling model, k 3 is greater than zero. element site (right) or a half-site (left). The gel mobility shift assay was used to separate bound complex from free DNA, and each was quantified by phosphorimager analysis. A half-site oligonucleotide was used to measure the equilibrium binding constant K 1 for GRdbd monomer binding to DNA. We were able to measure nearly pure monomer complex at concentrations of GRdbd up to 50 nM (10 ng/20 l). Using the experimentally determined values of free DNA and GRdbd/DNA complex, the equilibrium binding constant The dimer equilibrium constant K 2 was then determined by measuring the overall equilibrium constant from experiments in which GRdbd was bound to a full, palindromic binding site and solving was determined for the equilibrium dimer binding constant K 2 . The observation that K 2 is greater than K 1 is indicative that dimer binding is cooperative. Data presented in Fig. 2 were used to determine a Hill coefficient (Fig. 3). A coefficient of 1.5 was determined for the binding of GRdbd to the full-site oligonucleotide. A Hill coefficient for dimer binding of between 1 and 2 is indicative of positive cooperativity (29).
To study dissociation of a binding protein from DNA, reassociation must be prevented. This is often accomplished by adding excess, unlabeled competitor DNA to inhibit reassociation with labeled DNA. However, this will give incorrect estimates of dissociation rates if the added DNA itself influences the apparent rate of dissociation, as would be predicted for an intersegment transfer mechanism. Gel shift experiments are often done in conditions of protein excess to maximize signal. However, this also means that high excesses of unlabeled DNA must be used to soak up the excess binding protein, thereby exacerbating the possibility that the apparent dissociation rates are being influenced by the added DNA. An alternative strategy to prevent reassociation is to greatly reduce the concentration of the bound complex by dilution, shifting the equilibrium to favor dissociation. Experimentally, this is limited by the ability to detect binding in extremely dilute solutions.
We have established that we can readily assess GRdbd binding following a 100-fold dilution of our standard binding condition. To test the extent to which reassociation occurs at this dilution, we mixed DNA and GRdbd at concentrations equivalent to the 100-fold diluted standard condition and showed that some binding was still detectable (about 15% of that seen with undiluted extract). Therefore, some reassociation can occur even after dilution. To prevent this small amount of reassociation, we examined adding a small amount of specific oligonucleotide to the dilution buffer. In a binding reaction in which GRdbd is bound to labeled DNA at a 1:1 molar ratio, addition of an equimolar amount of unlabeled specific oligonucleotide reduces binding by precisely one-half and a 10-fold excess reduces binding by 90%, a concentration at which nonspecific oligonucleotide does not compete (Fig. 4A). In Fig. 4B we show that addition of a 10-fold molar excess of specific oligonucleotide to the dilution buffer completely prevents any reassociation. As subsequent experiments will confirm, addition of this concentration of specific oligonucleotide is still 200-fold below levels at which effects on dissociation rates are observed. Therefore, in all subsequent experiments, dissociation is initiated by 100fold dilution into a 10-fold excess of specific oligonucleotide. Other nucleic acids or test compounds were added as indicated.
Dissociation Kinetics of GRdbd/DNA Complex-Using the dilution condition established above, we examined the dissociation of monomer and dimer GRdbd. As can be seen in Fig. 5, dissociation of monomer from DNA, unlike that of dimer, is quite rapid. Dissociation constants were calculated according to the integrated first-order kinetic equation where k d is the dissociation constant, t is time, R t is the complex remaining at time t, and R 0 is the initial bound GRdbd (taken as bound counts at time 0). In all of the succeeding figures, dissociation data is presented as ln(R t ) versus time, and the absolute value of the slope of the best fit line represents the apparent k d . As shown in Fig. 5, the dissociation rate of monomer GRdbd from the half-site (k Ϫ1 ) is 1.2 ϫ 10 Ϫ2 s Ϫ1 (ϳ88 s), and dimer GRdbd from the full site (k Ϫ2 ) is 2.2 ϫ 10 Ϫ4 s Ϫ1 (ϳ77 min). Thus, the half-life of the monomer complex (ln(2)/k) is about 60 s, and for the dimer complex the half-life approaches 60 min.
The association rate for monomer (k 1 ) was too rapid to be measured by gel shift analysis, and because the association rate for dimer (k 2 ) is dependent on the formation of monomer (the GRdbd is a monomer in solution), a direct measurement of association was not possible. However, using the relation between the equilibrium constant and the kinetic constants (K 1 ϭ k 1 /k Ϫ1 , and similarly, K 2 ϭ k 2 /k Ϫ2 ) we calculate the values for k 1 and k 2 as 2.1 ϫ 10 5 and 5.6 ϫ 10 4 M Ϫ1 s Ϫ1 for monomer and dimer, respectively. The equilibrium constant for dimer binding is ϳ17 times greater than that for monomer binding, indicative of strongly cooperative binding. This is despite the fact that association of monomer is 3-4-fold faster than that of dimer. Thus, the cooperative effect can be attributed to stabilization of the DNA-bound complex upon dimer formation that results in an almost 60-fold reduction in off-rate compared to monomer.
Effect of DNA on Dissociation of GRdbd Dimer from DNA-Having established the kinetics of this system, we then measured the effect of additional DNA on the dissociation rate of dimer. Because the intersegment transfer mechanism predicts that increasing concentrations of DNA should increase the apparent dissociation rate of dimer (in contrast to the cycling model), we tested the ability of sheared E. coli DNA and specific or nonspecific oligonucleotides to increase the apparent dissociation rate of GRdbd bound to a specific oligonucleotide. As seen in Fig. 6, increasing amounts of sheared E. coli DNA in the dissociation dilution mix increased the measured off-rate from 2.7 ϫ 10 Ϫ4 to 3.6 ϫ 10 Ϫ3 s Ϫ1 (from t1 ⁄2 ϭ 42.8 min to t1 ⁄2 ϭ 3.2 min) over the concentration range of 0 -3.0 ϫ 10 Ϫ4 Mbp (0 -200 g/ml). This is a 13-fold increase in the apparent dissociation rate of GRdbd dimer. The same effect can be demonstrated for specific and nonspecific oligonucleotides (data not shown). In-creasing concentrations of oligonucleotide DNA from 1.0 ϫ 10 Ϫ7 to 1.0 ϫ 10 Ϫ5 M increases the off-rate up to 5.8 ϫ 10 Ϫ3 s Ϫ1 (t1 ⁄2 ϭ 2.0 min) for specific oligonucleotide, and 6.0 ϫ 10 Ϫ3 s Ϫ1 (t1 ⁄2 ϭ 1.9 min) for nonspecific oligonucleotide. Note that the size of the oligonucleotide used here is 34 base pairs, so the final concentration of added oligonucleotide DNA falls between 3.4 ϫ 10 Ϫ6 and 3.4 ϫ 10 Ϫ4 Mbp (between 2.2 and 220 g/ml), an effective range similar to that obtained with sheared E. coli DNA. These data suggest that the bound GRdbd dimer can transfer between two separate DNA molecules. In addition, the similarity between specific and nonspecific oligonucleotide results suggests that intersegment transfer occurs via non-sitespecific binding of DNA. This observation has implications for the binding mechanism employed, an issue addressed further below and under "Discussion." GRdbd Monomers Are Capable of Intersegment Transfer-Although our initial expectation was that intersegment transfer would occur by a strand invasion mechanism whereby one monomer of a bound dimer would transiently bind two separate molecules of DNA, the failure of specific oligonucleotide to increase the apparent off-rate any differently than nonspecific oligonucleotide or bulk DNA led us to question whether a domain other than the site-specific DNA binding domain was mediating intersegment transfer and therefore whether a monomer was itself capable of intersegment transfer.
To test this hypothesis, the half-site oligonucleotide was used at concentrations determined to permit exclusively monomer formation (see Fig. 2). We observe indications that monomer complex can undergo intersegment transfer even at added DNA concentrations at the low end of those at which dimer dissociation was affected. As shown in Fig. 7, nonspecific DNA increased the monomer dissociation rate to the limit of our systems measurement (t1 ⁄2 Ϸ 20 s). At DNA concentrations greater than 1 ϫ 10 Ϫ6 Mbp (22 g/ml), dissociation was faster than could be measured. Similar results were obtained with specific oligonucleotide. These data suggest that monomers as well as dimers of GRdbd can transfer from one DNA strand to another.
Comparison of Various DNA-like Molecules on Dimer Dissociation Kinetics-To test the specificity of the transfer mecha- Reactions were allowed to reach equilibrium (30 min at room temperature). The added competitor was either a specific or nonspecific oligonucleotide, as indicated. Concentration of initial reactants were 8.0 nM GRdbd, 4.0 nM specific full-site oligonucleotide DNA, and competitor as indicated. The complexes were resolved on 10% polyacrylamide nondenaturing gels as described under "Experimental Procedures." The dimer complex was quantified by phosphorimager analysis and plotted as the percentage of dimer complex remaining. All values were corrected for background. B, 10 ϫ specific competitor will prevent reassociation of GRdbd with labeled DNA following 100-fold dilution from standard binding conditions. Lanes 1-4 are complexes formed by adding GRdbd to labeled DNA at conditions comparable to those following 100-fold dilution from standard binding conditions (4.5 ϫ 10 Ϫ11 M, 1:1 molar ratio), in either the presence or the absence of the indicated molar excess of specific competitor at room temperature for 30 min. The diffuse shadowing observed in the region of the monomer band was enhanced by the photographic reproduction and was inconsistently observed.

FIG. 5. Kinetics of dissociation for monomer and dimer bound
GRdbd. Prebound complexes were diluted 100-fold in binding buffer with a 10-fold molar excess of specific oligonucleotide to prevent reassociation and incubated for the indicated time at room temperature before being loaded onto a 10% polyacrylamide nondenaturing gel as described under "Experimental Procedures." The dimer, monomer, and free DNA bands are indicated. The complexes are quantified by phosphorimage analysis (shown below each gel), and plotted as a function of the natural log of the remaining complex (R t ) versus time in seconds. The dissociation constant is determined as the absolute value of the slope of the linear best fit. nism, we employed several compounds that would mimic the long chain polyanionic character of DNA and several control compounds as well (Fig. 8). Three compounds with polyanionic character, heparin, polyglutamate, and ribo-poly(A)⅐poly(U) could promote dissociation of GRdbd from DNA. However, dissociation was not affected by a neutral macromolecule, polyproline, nor the polyanion subunits, glutamate and heparin disaccharide. Two cations, arginine and spermidine, likewise had no effect on dissociation. These data indicate that the interaction of GRdbd that permits intersegment transfer is primarily electrostatic, requiring a long chain polyanion to promote the dis-sociation of GRdbd from DNA.
A counterion condensation layer is present around DNA and other long chain polyions of sufficient charge density. Displacement of this charge condensation layer provides an entropic gain for protein binding to DNA. Because the intersegment transfer reaction appears to be via nonspecific interactions with negatively charged polyanions, our results suggest that the counter-ion condensation layer around the DNA may play a role in direct intersegment transfer (see "Discussion"). We wished to test whether there was a density of negative charge necessary to promote the intersegment transfer mechanism. Using poly-Glu and a 1:1 random copolymer of Glu/Glu-OEt, we examined the ability of each compound to mediate intersegment transfer. The results shown in Fig. 9 indicate that the transfer reaction is severely inhibited by the reduction of the average charge of the glutamate backbone by 1 ⁄2. Poly-Glu increased the off-rate over 80-fold at the highest concentration tested, whereas the Glu-Glu-OEt copolymer increased it only 3-fold. These data indicate the importance of the density of negatively charged residues along the chain, and implicate the need for the formation of a counter-ion condensation layer. The residual effect of the poly-Glu/Glu-OEt may be attributable to the fact that because it is a random copolymer, there will be an occasional region of high charge density due to the clustering of Glu residues in the polymer. Nevertheless, it is clear that charge density is important for the transfer of GRdbd bound to DNA.

DISCUSSION
Cooperative Binding by GRdbd-The present study confirms earlier work indicating that GRdbd exhibits positive cooperativity in binding to DNA as a dimer. Härd et al. (23) estimated by fluorescent spectroscopy that the cooperative effect was about 10-fold for a somewhat larger GRdbd fragment (115 versus 85 amino acids) binding at high salt conditions (270 mM). More recently, this group updated its findings at lower salt conditions (187 mM), estimating a 25-50-fold cooperativity (17,18). In our work, the equilibrium constant for dimer binding K 2 exceeded that of monomer binding K 1 by 15-fold. Moreover, the kinetic work presented here indicates that the cooperative effect is driven by stabilization of the bound dimer (t1 ⁄2 ϭ 60 min for dimer dissociation versus t1 ⁄2 ϭ 60 s for monomer dissociation).
Search Mechanisms in Chromatin-Since the initial observations by Riggs et al. (2) in the early 1970s that E. coli lac repressor associated with DNA 100-fold faster than simple diffusion would allow (2), a great deal of theoretical and experimental work has been conducted on search mechanisms for binding sites in chromatin. The theoretical framework for diffusion-controlled association of protein with DNA has been detailed extensively (31)(32)(33)(34). Based on von Smoluchowski diffusion theory (35), these authors describe the macromolecular and micromolecular interactions that occur when DNA-binding proteins interact with the DNA chain. Briefly, as a DNA-binding protein diffuses through a solution, it undergoes collisions with the DNA chain where nonspecific electrostatic contacts are made (macromolecular collisions). Electrostatic stabilization (with counter-ion release) can then allow the protein to be localized near the DNA chain; this results in multiple short range contacts (micromolecular collisions). Thus, the protein can sample several binding orientations before diffusing back into the solution. If the protein has localized near a binding site, then it can bind this site with high affinity. If, however, there is no specific site, and no other mechanisms are proposed to allow protein movement, then the only method available for further search is to dissociate back into the medium and rebind at another location. Clearly, this mechanism cannot account for a number of instances in which association rates of DNAbinding proteins have been measured that are substantially faster than diffusion-controlled mechanisms allow (2, 3).
Because the protein initially binds DNA electrostatically, a sliding model was proposed to facilitate local searches (12,36).
This model was based on the assumption that nonspecifically bound protein is able to move freely along the DNA chain by displacing salt ions from in front and filling the gap behind with no change in the total number of displaced counter-ions (5). This mechanism has been shown experimentally with restriction endonucleases (9, 10) and lac repressor-operator (6,11). In this fashion, a protein conducts a rapid (on the order of 10 Ϫ9 cm 2 s Ϫ1 (12)) one-dimensional search of the DNA. However, this search only occurs over a relatively short distance along the DNA strand before the protein dissociates.
In addition to simple diffusion and sliding, a third mechanism, intersegment transfer, was proposed to facilitate the location of specific sites in the genome. This mechanism envisions that a protein bound to one DNA site transfers directly to a second site via an intermediate stage in which it is transiently bound to two molecules of DNA (5,8,34,37). This strategy reduces the dimensions of the search space and allows the protein to conduct a rapid search for a binding site by maintaining a close association with the DNA matrix. This mechanism also has the advantage that increasing DNA concentration increases the rate of search by enhancing the apparent off-rate of the protein, thus allowing a more rapid sampling of potential binding sites. Solutions to the kinetics equations for a model protein suggest that the rate of site sampling can be increased by 3 orders of magnitude. 2 GRdbd Employs Intersegment Transfer as a Search Mechanism-Glucocorticoid hormone must initiate a series of steps to convert its receptor to an active transcription factor, including dissociation of the receptor from heat shock proteins, translocation of the receptor to the nucleus, location of binding sites on a target promoter, and assembly of a transcription complex. Glucocorticoids can modify rates of transcription within minutes (38), necessitating that the receptor rapidly locate and bind target sites. Our data are not consistent with the concept that the receptor finds these sites by sampling through repeated rounds of association and dissociation. The intrinsic dissociation rate of GRdbd dimer from DNA is slow; however, the apparent dissociation rate is greatly enhanced by DNA, a result consistent with our interpretation that the GRdbd is employing intersegment transfer as a search mechanism. This enhancement was detected at DNA concentrations ranging from 2 to 200 g/ml. The concentration of DNA in the nucleus of a eukaryotic cell is estimated at ϳ10 -50 mg/ml. This presents a physical obstacle to diffusion-controlled search strategies and an aid to an intersegment transfer mechanism. Even acknowledging that a fraction of DNA may be packaged so that it is effectively inaccessible to receptor, intersegment transfer is a particularly effective mechanism for exploiting the circumstances posed by the eukaryotic nucleus in undertaking a search for a target site.
Analysis of the crystal structure of the GRdbd⅐DNA complex indicates that specific DNA contacts are made via an ␣-helical domain present at the C-terminal end of the first zinc coordination finger (14). The intersegment transfer mechanism requires that GRdbd transiently bind two separate molecules of DNA. It is a straightforward assumption that one monomer of a DNA-bound GRdbd dimer could release from its site and bind a second DNA followed by transfer of the other half of the dimer to the second DNA. However, this supposition is inconsistent with the finding that the dissociation from a specific target is enhanced equally by a nonspecific oligonucleotide and by one containing a specific target site. If transfer involves the DNA binding domain that mediates specific binding, one would anticipate that transfer would be less favored if specific contacts 2 J. R. Cann, unpublished data. could not be made. Indeed, transfer can be mediated by a variety of polyanions. Moreover, our data indicate that the dissociation of monomer from DNA is enhanced by DNA, suggesting that the monomer itself can transiently bind separate molecules of DNA. Therefore, the GRdbd monomer must posses a second domain capable of nonspecific interaction with DNA. There are two possibilities for candidates for this domain. The crystallography data (14) indicate that the ␣-helix at the Cterminal end of the second zinc finger makes no DNA contacts and that when the GRdbd is bound to a specific site, the second helix is positioned away from and at right angles to the first helix. A peptide comprising the second zinc finger can bind DNA (39), and helix-disrupting mutations in this helix abrogate function in the context of the intact receptor (40). Thus, this helix is potentially responsible for mediating transfer through nonspecific DNA binding. A second domain that could assist with or mediate nonspecific DNA binding is a highly basic region near the C terminus of the GRdbd that was not localized in the crystal structure. In preliminary studies, a GRdbd fragment lacking this region (courtesy of T. Kerppola) failed to bind DNA.
Specificity of GRdbd Intersegment Transfer-During a search for target sites in vivo, a glucocorticoid receptor would first sample and transfer from many nonspecific sites before locating a specific target site. Binding of GRdbd to nonspecific DNA was too low affinity to permit its quantification and therefore assess whether DNA could influence the apparent off-rate. Therefore, the experiments above were conducted by initially binding GRdbd to a palindromic target site. Thus, it is clear that DNA at moderate to high concentrations enhances the apparent off-rate of GRdbd from specific target sites, as well as from nonspecific sites. It is possible that assembly of the receptor into a transcription complex at its target site abrogates intersegment transfer, perhaps because the second DNA binding domain necessary for the transient, simultaneous binding to two DNA molecules is involved in protein-protein interactions in such a complex. Alternatively, cycling on and off a target site may be critical to receptor action. One can envision mechanisms whereby the receptor acts as a single-shot effector that must be removed and reloaded onto a target site before it is capable of again fulfilling its activation function. In this case, intersegment transfer may be critical not just to the search for target sites but to the transcription activation function of the receptor at the target site as well.
The DNA mimic experiments (Figs. 8 and 9) indicate that there is surprisingly little constraint on the structure of the molecule that will promote dissociation of GRdbd from DNA. RNA, heparin, and polyglutamate all increase the apparent off-rate. This result supports the hypothesis that the second interaction is primarily electrostatic, requiring a long chain polyanion. Decreasing charge density inhibits the ability of the polyanion to mediate the transfer (Fig. 9). Based on theoretical analyses (41,42), the amount of thermodynamically bound counterions around a long chain molecule (e.g. DNA) is dependent on the average charge separation along the chain. Release of counter-ions from the DNA by protein binding has been shown to create an entropic stabilizing force for nonspecific protein-DNA interactions (41,42). For double-helical B-DNA this charge separation is 1.7 Å (43), and results in a fractional charge neutralization () of 0.88. The structure of poly-Glu is not known, but if we assume an extended chain, then the average charge separation can be represented by the ␣-carbon of the side chain projected back to a central line along the polypeptide backbone, and the charge separation should be about 3.62 Å (30). This gives a linear charge density of ϳ 1.96 (43), and the fractional charge neutralization () is 1 Ϫ (1/1.96) ϭ 0.49, indicating that poly-Glu has the capacity to form a thermodynamically bound ion condensation layer. The substituted polypeptide (1:1) poly-Glu/Glu-OEt, in which half of the charged carboxyl groups have been replaced with an ester linked ethyl group, has an average charge separation of ϳ7.2 Å and thus should have no counter-ion condensation layer associated ( ϳ 1.0; ϭ 0) (43). This random copolymer retains some small ability to mediate transfer, probably through scattered regions of clustered Glu residues, where the local charge density is high enough to create regions of organized counterions.
Dissociation Measurements and Competitor DNA-Dissociation measurements with the GRdbd also point out a potential problem encountered by experiments designed to measure dissociation rates of DNA-binding proteins. Many researchers use a large excess of specific competitor DNA in their experiments to prevent reassociation of the protein with a labeled specific DNA fragment. Because we have shown that the concentration of competitor DNA added to a protein-DNA binding reaction can potentially affect the rate of apparent dissociation, experiments using excess competitor DNA to measure dissociation must experimentally determine the minimal amount of necessary competitor in order to avoid a systematic error. In our experiments, with the GRdbd at 1:1 molar ratio to DNA, we determined that 2-fold excess specific competitor was sufficient to disrupt the binding equilibrium, and 10-fold excess specific competitor was sufficient to prevent reassociation to labeled DNA (Fig. 4). This latter concentration of specific competitor DNA (50 nM for standard conditions and 1/100 of that for the diluted dissociation conditions) is below that seen to influence the off-rate by intersegment transfer (Ͼ10 Ϫ7 M or 2-3 ϫ 10 Ϫ6 Mbp). For example, if a typical experiment using 1 ng of bound 30-mer oligonucleotide was competed with 400-fold competitor, then a 20-l sample volume would contain 10 Ϫ6 M oligonucleotide (3 ϫ 10 Ϫ5 Mbp), a concentration sufficient to substantially influence the dissociation rate of GRdbd. If the binding reaction contained nonspecific competitor, such as poly(dI⅐dC), which is often used to reduce the level of nonspecific protein binding from crude protein preparations, then this also could influence the measured equilibrium and dissociation kinetics.
We have presented data that are compatible with the theory that the GRdbd is able to transfer directly from one DNA molecule to another using an intersegment transfer mechanism. We propose that the active full-length GR is able to sample potential target sites rapidly by a similar mechanism. At a target promoter, the association of GR with transcriptional coactivators may stabilize GR dissociation from the site by abrogating intersegment transfer, perhaps because the protein-protein interactions sterically hinder the nonspecific DNA binding site needed to conduct intersegment transfer. Alternatively, receptor could activate transcription by a single-shot mechanism so that rapid transfer from DNA would be required to remove receptor from a target site in order for it to be reloaded and again fired. Although the intersegment transfer mechanism must still be shown with the full-length receptor, we are attracted to the idea of the receptor swinging through the nuclear jungle from DNA to DNA in search of a specific binding site.