DNA binding by single HMG box model proteins

Abstract

The HMG1/2 family is a large group of proteins that share a conserved sequence of ∼80 amino acids rich in basic, aromatic and proline side chains, referred to as an HMG box. Previous studies show that HMG boxes can bind to DNA in a structure-specific manner. To define the basis for DNA recognition by HMG boxes, we characterize the interaction of two model HMG boxes, one a structure-specific box, rHMGb from the rat HMG1 protein, the other a sequence-specific box, Rox1 from yeast, with oligodeoxynucleotide substrates. Both proteins interact with single-stranded oligonucleotides in this study to form 1:1 complexes. The stoichiometry of binding of rHMGb to duplex or branched DNAs differs: for a 16mer duplex we find a weak 2:1 complex, while a 4:1 protein:DNA complex is detected with a four-way DNA junction of 16mers in the presence of Mg²⁺. In the case of the sequence-specific Rox1 protein we find tight 1:1 and 2:1 complexes with its cognate duplex sequence and again a 4:1 complex with four-way branched DNA. If the DNA branching is reduced to three arms, both proteins form 3:1 complexes. We believe that these multimeric complexes are relevant for HMG1/2 proteins in vivo, since Mg²⁺ is present in the nucleus and these proteins are expressed at a very high level.

INTRODUCTION

HMG proteins were identified as a highly abundant group of non-histone chromosomal proteins by the Johns’ group (1), characterized by their dissociation from chromatin at lower salt concentration than histones, rapid migration in acidic SDS gels and high content of charged residues. Subsequently, Jantzen et al. (2) characterized the rRNA transcriptional control element UBF and found that it was homologous to HMG1/2 proteins. By sequence alignment they could identify three repeats of an 80 amino acid ‘box’ related to two similar boxes in HMG1/2. Since then HMG1/2 box sequences have been identified in more than 150 proteins (3–6), including transcription factors and DNA repair proteins. Substrates for HMG1/2 proteins include single-stranded (ss)DNA (7), duplexes containing AT sequences (8), supercoiled DNA (9) and branched or cruciform DNAs (10,11). The molecular structures of HMG boxes from several members of the family have been determined, revealing a common flattened triangular fold formed by intersection of three helical arms (12–15).

A characteristic of many HMG box proteins is their ability to recognize architectural features in DNA, whether they are A·T runs that promote DNA bending (16), DNA reacted with cisplatin (17) or branched structures (11). Bianchi’s group (18) first reported that HMG boxes bind branched DNA and proposed that HMG boxes interact most favorably with a sharp angular distortion or kink in DNA. Structures of complexes formed between HMG box transcription factors and cognate AT-rich sites indeed reveal sharply bent duplexes, with amino acid side chains inserted between base pairs of the DNA (19,20). A similarly distorted structure has recently been reported for the complex between a cisplatin-reacted duplex and HMGa (21). This mode of interaction rationalizes the ability of HMG box proteins to favor ssDNA (8) and supercoiled DNA (22) as substrates rather than classical duplex DNA, which is more rigid.

The yeast Rox1 hypoxic repressor is a sequence-specific DNA binding protein in the SOX class of HMG proteins (23). It consists of 368 amino acids, the first 80 of which comprise the HMG domain, while the remainder comprise a repression domain. Rox1 binds to the double-stranded (ds)DNA sequence YYYATTGTTCTC with a K_d in the 20 nM range and bends the DNA at an angle of ∼90° (24,25). The internal residues of this binding site, YATTGTT, are common to the SOX family members, perhaps reflecting a common requirement for DNA bending (26,27). The structure of the HMG domain of the SOX family member SRY complexed with DNA has been determined by NMR (20) and the Rox1 sequence has been modeled onto this structure (23).

Interactions between HMG box proteins and DNA are thought to consist of different binding modes: a tight 1:1 complex, followed by a mode in which additional proteins associate to form structures that ultimately can be large enough to be visualized in electron micrographs of complexes with supercoiled DNA (22). We show here using oligomeric substrates that rHMGb and Rox1 interact with ssDNA to form a 1:1 complex, with a K_d in the micromolar range. Duplex and branched DNAs show more complex stoichiometry. We present evidence that both proteins form 4:1 complexes with four-way junctions, based on data from gel shift, fluorescence competition, crosslinking and ultracentrifugation experiments. This description of the interaction between HMG boxes and branched DNA differs in several respects from the current models emphasizing 1:1 complexation (28) and has implications for structural studies of protein–DNA complexes involving HMG boxes as well as the biological role of these proteins. Tight complexes involving multiple proteins and branched DNA have been reported previously, in the cases of HMGD (29), SRY (18) and the HMGa box (30). Here we have used an equilibrium assay to establish the stoichiometry of binding more rigorously, with additional evidence from crosslinking and sedimentation equilibrium experiments.

MATERIALS AND METHODS

Expression and purification of HMGb

The laboratory of S. J. Lippard (Department of Chemistry, Massachusetts Institute of Technology) provided plasmid pHB1 encoding the second HMG domain (HMGb, amino acids 86–165) of rHMG1. Protein was expressed and purified as described (3,31,32). The final product was applied to a CM column and dialyzed against water. Protein concentration was determined according to Pace et al. (33).

The coding sequence of the HMG domain of Rox1, residues 1–90, was inserted into the vector pET-24a (Novagen). The fusion added six histidine codons to the 3′-end of the coding sequence. The plasmid was expressed in strain BL21(D3) in 2 l of Luria broth and Rox1 expression was induced with 1 mM IPTG (34). Cells were broken in a French press, the lysate cleared by centrifugation and the protein bound to nickel beads (Qiagen) and eluted with 0.5 M imidazole (as recommended by the vendor). After dialysis, the protein was bound to a nickel column (Pharmacia) and eluted with several steps of increasing imidazole (0.1–0.5 M). Minor contaminants were removed by passage through a Sephacryl S-100 column (Pharmacia). The Rox1 protein was >90% pure.

DNA substrates

The oligonucleotides used in this study were synthesized and HPLC purified by CyberSyn Inc. (Lenni, PA) and used without further purification. Junction DNA J1 was composed of four oligonucleotides as described by Kallenbach et al. (35). The sequences of these four strands are: 1, 5′-CGCAATCCTGAGCACG-3′; 2, 5′-CGTGCTCACCGAATGC-3′; 3, 5′-GCATTCGGACTATGGC-3′; 4, 5′-GCCATAGTGGATTGCG-3′. Junction Y1 (36) was composed of three oligonucleotides: 1, 5′-CACCGCTCTGGTCCTC-3′; 2, 5′-GAGGACCAACAGCCTG-3′; 3, 5′-CAGGCTGTGAGCGGTG-3′. The duplex containing the Rox1 binding site was composed of two oligonucleotides described by Balasubramanian et al. (24). The sequences are: 1, 5′-GGGTTTTCAGCCCATTGTTCTCGAGCCAAACC-3′; 2, 5′-GGTTTGGCTCGAGAACAATGGGCTGAAAACCC-3′.

Electrophoretic mobility shift assays

The strands designated 1 in the junction DNA J1, Y1 and Rox1 target duplex were radioactively 5′-³²P-labeled using T4 polynucleotide kinase (Gibco BRL). After inactivation of the enzyme, the strands were desalted on Bio-Spin 6 columns (Bio-Rad). DNA was formed by mixing the labeled strand with its complementary strand(s) in 50 mM Tris–HCl, pH 7.5, 10 mM MgCl₂, heating at 90°C for 2 min, then cooling overnight. Solutions of annealed DNA were incubated with various concentrations of HMG proteins on ice for 30 min in 20 mM Tris–HCl, pH 7.5, 100 mM NaCl, 1 mM MgCl₂, 10% (w/v) glycerol, 0.5 mM dithiothreitol. The reaction mixtures were applied to polyacrylamide gels and electrophoresed in 20 mM Tris, 10 mM HAc, 0.1 mM MgCl₂ at 4°C. Gels were dried on Whatman 3MM paper and exposed to X-ray film (Kodak) or scanned with a PhosphorImager.

Preparation of ɛDNA and concentration determination

Etheno-modification of a 21mer oligonucleotide probe S21 (GAGGACCAAAAAAACAGCCTG) was carried out as described (37) and the product purified by ethanol precipitation. The concentration of the product S21 was determined using an extinction coefficient kit (Sigma). The extinction coefficient of the fully digested oligonucleotide was calculated as 1.647 × 10⁵ M^–1cm^–1 (38–40).

Fluorescence competition titration

Fluorescent ɛDNA (S21) was titrated in the presence or absence of non-fluorescent competitor DNAs using an Aviv ATFF 212 filter fluorometer, equipped with a Hamilton Microlab series 500 syringe system for automated volume dispensing under PC control. Titrations were carried out at fixed volume, at 25°C in 20 mM sodium cacodylate, 100 mM NaCl, 1 mM MgCl₂. To avoid diluting the DNA components, the titrant contained the same concentration of DNAs as the initial solution in the cuvette. All titrations were corrected for the protein blank. The excitation filter was set at 308 nm and the emission filter at 405 nm. No inner filter effect could be detected during the titrations. The instrument uses QC correction to suppress lamp fluctuations during experiments.

Crosslinking

The crosslinking experiments were carried out according to Grasser et al. (41). Briefly, 400 µl samples of 5 µM HMGb solutions containing different concentrations of J1 DNA (0, 0.25, 0.5, 1 and 5 µM) in 20 mM sodium cacodylate, 100 mM NaCl, 1 mM MgCl₂ were treated with 4 µl of a freshly prepared suberic acid bis(N-hydroxysuccinimide ester) solution in DMSO (20 mg/ml). Reactions were quenched by adding 40 µl of 1 M Tris–HCl, pH 8, at 10 or 30 min. After precipitation with an equal volume of 50% (w/v) trichloroacetic acid, the samples were washed with cold acetone and vacuum dried. The products were separated on 15% SDS gels (30:0.8 acrylamide:bisacrylamide) and stained with Coomassie brilliant blue R250. The separating gel had a final Tris concentration of 0.73 M.

Equilibrium ultracentrifugation

Sedimentation equilibrium analysis was performed on a Beckman XL-A analytical ultracentrifuge as described by Shu et al. (42). Samples were loaded at initial protein concentrations of 100, 50 and 25 µM and analyzed at rotor speeds of 12 000 and 15 000 r.p.m. at 20°C. Data sets were fitted to a single species model using the program NONLIN (43) using an overall partial volume of 0.70 cm³g^–1 and solvent density of 1.08 gcm^–1 (44).

RESULTS

Gel mobility experiments

One of the most widely used methods for analysis of protein–DNA interactions is the electrophoretic mobility shift assay (EMSA), which has the ability to detect multiple species in the limit of tight binding complexes (45,46). Figure 1 indicates that multiple species are formed in the interaction between HMGb and the DNA four-way junction J1. This is in contrast to comparable published experiments with HMGa and branched DNA substrates (28), although these differ in the binding medium and gel conditions. In order to make the conditions comparable, we repeated our experiment using 1× TBE buffer as described by Pöhler et al. (28). We again observed multiple discrete bands rather than the single bound species reported. To determine whether or not multiple species are a feature of sequence-specific boxes as well, we carried out the experiment shown in Figure 2, where the HMG box from the transcription factor Rox1 replaces HMGb. In this case the bands corresponding to intermediate species are more clearly defined and a set of four complexes is detected. Since the Rox1 system shows well-defined complexes in gels, we performed the analogous assay using a three-arm junction, Y1, instead of the four-arm junction, J1. Figure 3 summarizes the results. In this case we observe three complexes rather than four. If we assume that the target is the branch rather than the arms, since duplexes bind more weakly, then we might suppose that the proteins are not binding equivalently. In the case of the cognate sequence duplex of Rox1, we find two complexes, suggesting a 2:1 limiting stoichiometry (Fig. 4).

Figure 1.

Binding of HMGb to a four-way junction (J1). Lane 9 contains the free junction (0.05 µM). Lanes 1–8 have the same concentration of J1 as in lane 9, with increasing concentrations of HMGb. The HMGb concentrations in lanes 1–8 are 0.05, 0.5, 1, 2.5, 3.75, 5, 10 and 20 µM, respectively. The gel was 12% with 19:1 acrylamide:bisacrylamide, run as described in Materials and Methods.

Figure 2.

Binding of Rox1 to a four-way junction (J1). Lane 1 contains the labeled strand 1 of J1 alone, lane 2 the free junction (0.05 µM). Lanes 3–12 have the same concentration of J1 as lane 2, with increasing concentrations of Rox1 protein. The protein concentrations in lanes 3–12 are 0.05, 0.25, 0.5, 0.75, 1, 1.5, 2.5, 3.75, 5 and 7.5 µM, respectively. The gel was 6% with 19:1 acrylamide:bisacrylamide.

Figure 3.

Binding of Rox1 to a three-way junction (Y1). Lane 1 contains the labeled strand 1 of Y1 alone, lane 2 the free junction (0.5 µM). Lanes 3–12 have the same concentration of Y1 as lane 2, with increasing concentrations of Rox1 protein. The protein concentrations in lanes 3–12 are 0.25, 0.5, 0.75, 1, 1.25, 1.5, 1.75, 2, 3 and 4 µM, respectively. The gel was 6% with 29:1 acrylamide:bisacrylamide.

Figure 4.

Binding of Rox1 to its cognate sequence. Lane 1 contains labeled strand 1 of duplex with the Rox1 binding site, lane 2 the free duplex (0.025 µM). Lanes 3–12 contain the same concentration of duplex as lane 2, with increasing concentrations of Rox1 protein. The protein concentrations in lanes 3–12 are 0.025, 0.125, 0.25, 0.375, 0.5, 0.75, 1, 1.5, 2 and 2.5 µM, respectively. The gel was 6% with 29:1 acrylamide:bisacrylamide.

Fluorescence competition assays

Fluorescence titration analysis has also been widely used to investigate protein–DNA interactions (47). Binding can often be detected by monitoring changes in the protein fluorescence signal upon binding to DNA. However, at the wavelength used to excite tryptophan we found overlap from absorption bands of the DNA. On the other hand, etheno-modified DNA can be excited at a longer wavelength, 308 nm, providing a window to monitor the binding behavior of DNA binding proteins without interference from absorbance by the bases (48,49). We titrated the fluorescent probe S21 with HMGb in the absence or presence of J1. The shift of the titration curve in the presence of J1 indicates competition between S21 and J1 for binding the protein (Fig. 5A). A variety of models were tested to fit the differential binding isotherms, in which the number of sites, the affinity of each site and the cooperativity between sites were varied (50). We start with a simple 1:1 complexation model, then progress to multiple sites with a single affinity, multiple sites with different affinities and finally cooperative models as outlined in Wyman and Gill (50). In the case of S21, the isotherm is closely fitted by the simplest model with 1:1 stoichiometry between protein and DNA. No 1:1 model satisfactorily accounts for the competition between S21 and J1 (Fig. 5A). The corrected differential binding isotherm is hyperbolic, with a limiting stoichiometry reaching four proteins to one J1 molecule (Fig. 5B). Good fits thus require multiple proteins attaching to J1 (Fig. 5C). While an adequate fit results from assuming four equivalent sites each with a K_d of 50 nM, the best fit results from a model in which one protein binds more tightly (K_d1 = 10 nM) than the other three, which have K_d2 = 160 nM (Fig. 5C). A model that assumes two kinds of site with two proteins occupying each site also gives a good fit, as shown in Figure 5C. The statistics are not quite as good, however: the r² values are 0.05 for n = 4, K_d = 50 nM; 0.08 for n₁ = 2, K_d1 = 15 nM and n₂ = 2, K_d2 = 170 nM; 0.006 for n₁ = 1, K_d1 = 10 nM, n₂ = 3, K_d2 = 160 nM. Thus the competitive binding experiments, which are equilibrium assays, provide firm evidence for multiple protein sites per branched DNA molecule.

Figure 5.

Fluorescent competition titration of J1 and S21 for HMGb. (A) Titrations carried out at 0.7 µM S21, with or without 0.5 µM J1. (B) Differential binding isotherm, calculated from (A). (C) Different models used to fit the binding isotherm.

We characterized the interaction further using a crosslinking assay to monitor potential protein–protein contacts that might arise on forming different protein–DNA complexes. Figure 6 shows that in the presence of J1 there is increased intensity of a dimer band that is only faintly visible in the case of the free protein. A trimer band appears as additional J1 is added. This suggests that as HMGb binds, no more than three molecules are spatially proximal to each other. This might be taken to support the heterogeneous site model with n₁ = 1, n₂ = 3, however, there are several other possibilities that cannot be eliminated.

Figure 6.

Crosslinking of HMGb. Lane 1 contains a set of molecular size markers. Lane 12 contains HMGb alone without any crosslinking reagent. Lanes 2 and 7 contain 5 µM HMGb only, in the presence of crosslinker. Lanes 3–6 contain 5, 1, 0.5 and 0.25 µM J1, respectively, in addition to 5 µM HMGb. Lanes 8–11 have the same composition as lanes 3–6, respectively. Lanes 2–6 were treated for 10 min; lanes 7–11 were treated for 30 min.

Molecular mass determination

Finally, we performed an equilibrium ultracentrifugation experiment to further define the stoichiometry of the complex between HMGb and J1. Figure 7 shows the results. A 4:1 mixture of HMGb and J1 is seen to sediment as a single species at three different concentrations, with an apparent molecular weight of 53.7 kDa, consistent with formation of a complex containing four proteins and one J1 molecule.

Figure 7.

(Lower) Analytical ultracentrifugation results (12 000 r.p.m.) recorded at 20°C in 20 mM sodium cacodylate, 100 mM NaCl, 1 mM MgCl₂, 0.5 mM dithiothreitol, pH 7.0, at 100 µM protein concentration and 25 µM J1 concentration. The natural logarithm of the absorbance at 296 nm is plotted against the square of the radial position. (Upper) Deviations from the calculated values are plotted as residuals.

To summarize the results, we find that the best fit to the binding isotherm derived from the fluorescence competition experiments is a final complex with four proteins bound to one junction; the best fit to the data corresponds to a model with heterogeneous binding sites, one protein binding more tightly (K_d = 10 nM) and three more weakly (K_d = 160 nM) at 25°C.

DISCUSSION

There is considerable interest in understanding the molecular basis of the recognition of DNA by HMG box proteins (28,51). There is mounting evidence implicating HMG1/2 proteins in a variety of biological processes: mouse knock-outs lacking HMG1 are lethal (52), while HMG1 interacts with p53 (53), HOX proteins (54) and RAG1 (55) and mediates the high affinity interaction of progesterone receptor with DNA (56). Most of the regulatory proteins that have been studied interact with their target DNAs via a system of base sequence-specific signals, typified by transcription factor and repressor interactions with cognate DNA sequences (57). In the case of Rox1 the target is a duplex containing the binding sequence 5′-GAGAACAATAAG-3′, which is AT rich and therefore likely to bend (16).

The experiments described here indicate formation of 4:1 protein cruciform complexes that does not depend on the DNA sequence itself (data not shown). The sequence-specific Rox1 protein binds its cognate duplex more tightly than branched DNA, while still forming tetrameric complexes with J1. In contrast, the SRY box binds more tightly to a junction than to its cognate duplex (18).

EMSA of the interaction between HMG box proteins and target duplexes or branched DNAs has previously been interpreted to indicate that HMG proteins bind as monomers to an ‘open’ form of four-arm DNA junctions (28). The tertiary structure of branched junctions that is stabilized by Mg²⁺ or other ions is thought not to be competent in the interaction: this structure forms in the presence of Mg²⁺, which inhibits protein binding. In these experiments there are indications of non-1:1 stoichiometric complexes at higher protein:DNA ratios (see for example figure 1A of ref. 28) but not in the initial phase of binding. Our present experiments were undertaken to clarify several issues concerning DNA binding by HMG boxes. How many HMG boxes bind to a branch in DNA, and how tightly? Our initial study has focused on the HMG box from the oxidation-responsive yeast transcription factor Rox1, which is structurally related to SRY (23), and the second box from rat HMG1 chromosomal protein, designated HMGb, the solution structure of which has been determined by NMR (58). Figure 2 shows the binding of HMGb to the four-arm junction J1, an immobile junction formed by association between four complementary 16mer strands (35). The four strands form a stable tetramer in the presence of Mg²⁺ ions, characterized by low mobility in native PAGE, selective drug binding at the branch site and UV hypochromism consistent with pairing of all the bases present. This tetramer folds to form a 2-fold symmetrical structure in the presence of Mg²⁺ ions, formed by antiparallel stacking of the extended duplex arms (59,60). In the absence of Mg²⁺, a more open structure predominates, and this has been taken to represent the target of the HMGa–cruciform interaction according to Pöhler et al. (28).

The major finding of this study is that in the presence of concentrations of Mg²⁺ that do not inhibit HMG protein–DNA interaction, 1:1 HMGb–J1 complexes can be detected only at the lowest concentrations possible; multiple complexes form otherwise. Fluorescence competition between HMGb and single strands such as the fluorescent 21mer S21 or any of the individual single strands that make up the junction J1 indicates an authentic 1:1 interaction. The competition process obtained in the presence of J1 results in a sigmoidal curve (Fig. 5A); the corrected binding isotherm (Fig. 5B) is hyperbolic, quantitatively consistent with a non-cooperative association of up to four proteins with the junction. The HMG box from the transcription factor Rox1 shows a particularly well-defined series of four complexes with J1 that co-exist over 1–2.5 µM protein concentrations (Fig. 2). A similar result was reported by Payet and Travers (29) using the single box HMG-D protein with a 30mer four-way junction.

The stoichiometry of HMG complexation is supported by the ultracentrifugation analysis of a 4:1 mixture, which sediments as a single component in the experiments we carried out. A protein–protein crosslinking experiment also reveals three HMGb molecules within crosslinking distance (<12 Å from the crosslinking arm in the complex).

This conclusion is seemingly at variance with the gel mobility shift data of Pöhler et al. (28), which support a 1:1 HMGa–junction complex. The NMR structures of several complexes of HMG proteins with cognate duplexes also show 1:1 stoichiometry (19–21). On the other hand a tetrameric complex between RuvA protein and a Holliday junction has been determined in which the junction is in an open form (61). Our data are not inconsistent with a structure of this kind: we find an optimal fit to the differential binding isotherm at 25°C in Figure 5B with a model in which a high affinity process (K_d ≈ 10 nM) involving one HMGb molecule is accompanied by a second mode involving three protein molecules with weaker affinity (K_d ≈ 160 nM) (Fig. 5). These affinities are comparable to the 1:1 complex reported by Pöhler et al. (28) (K_d = 200 nM), although our binding reaction was carried out in the presence of 1 mM Mg²⁺ ions. Tetramer binding is consistent with the junction assuming an ‘open’ or unstacked conformation as they propose, rather than remaining folded in the 2-fold symmetrical state (62), but this is by no means proven. Attachment of three Rox1 box proteins to the three-arm junction Y1 supports binding to open states of branched DNAs, since three arm junctions form tertiary structures only in the presence of additional bases at the branch point (63). Other examples of multiple HMG protein interaction with branched DNAs have been reported that support our view that these complexes are real and may play a biological role (29,30). As noted, the best fits suggest heterogeneous binding sites with different occupancy. Thus the three sites in the three-arm complex are probably not equivalent.

The conclusion from this study is that in the presence of 1 mM Mg²⁺ the branched DNA J1 binds four HMG boxes, with one box interacting more tightly than the other three. The best fit shows one site with a K_d of 10 nM and three others with K_d = 160 nM (Fig. 5C). Mathematically the fit in this case is hard to distinguish from a 2:2 model, in which two tight sites and two looser sites are present. It is not difficult to imagine that under suitable conditions of protein and DNA concentrations a 1:1 complex can appear to be the dominant binding mode, as reported previously (28).

What structural models are consistent with multiple protein binding? Several groups have suggested that HMG proteins interact with an open form of branched DNAs (28,64), as do junction resolving enzymes (65). If the difference between tight and loose sites is real, one site might correspond to the branch point itself, while the remaining sites represent facilitated occupancy of sites on the arms. Webb and Thomas (64) suggest that in HMG1 the A box occupies the tightest site, while the lower affinity B box binds one of the open arms. This is consistent with the footprinting data they present. Studies by Reeves’ group (66) show formation of a 2:1 complex of bovine HMG1 with a four-armed branched junction, consistent with a dimeric binding mode, or four boxes per complex.

The number of HMG proteins in the nucleus is very large; several hundred thousand are present in the mammalian nucleus (1). If we anticipate the presence of four boxes per complex in vivo, then a ratio of HMG1 protein containing two boxes per 10–15 nucleosomes will form one dimeric HMG1 complex per 20–30 nucleosomes. The local concentration of proteins is high, supporting the idea that the relevant complex is 4:1 rather than 1:1 in terms of boxes. Finally, it should be emphasized that the biological role of many HMG box proteins, including HMG1/2, is still imperfectly understood, as is the basis of their common ability to bind branched DNAs. The complete protein includes a strongly acidic C-terminal tail sequence, in addition to its two box subdomains, which is thought to mediate protein–protein interactions, since many nuclear proteins are basic. How this tail influences DNA binding is a question that remains to be addressed.