An investigation of the affinities, specificity and kinetics involved in the interaction between the Yin Yang 1 transcription factor and DNA


  • Filip M. Golebiowski,

    1.  Department of Physical Biochemistry, Faculty of Biochemistry, Biophysics and Biotechnology, Jagiellonian University, Krakow, Poland
    Search for more papers by this author
  • Andrzej Górecki,

    1.  Department of Physical Biochemistry, Faculty of Biochemistry, Biophysics and Biotechnology, Jagiellonian University, Krakow, Poland
    Search for more papers by this author
  • Piotr Bonarek,

    1.  Department of Physical Biochemistry, Faculty of Biochemistry, Biophysics and Biotechnology, Jagiellonian University, Krakow, Poland
    Search for more papers by this author
  • Maria Rapala-Kozik,

    1.  Department of Analytical Biochemistry, Faculty of Biochemistry, Biophysics and Biotechnology, Jagiellonian University, Krakow, Poland
    Search for more papers by this author
  • Andrzej Kozik,

    1.  Department of Analytical Biochemistry, Faculty of Biochemistry, Biophysics and Biotechnology, Jagiellonian University, Krakow, Poland
    Search for more papers by this author
  • Marta Dziedzicka-Wasylewska

    1.  Department of Physical Biochemistry, Faculty of Biochemistry, Biophysics and Biotechnology, Jagiellonian University, Krakow, Poland
    Search for more papers by this author

A. Górecki, Department of Physical Biochemistry, Faculty of Biochemistry, Biophysics and Biotechnology, Jagiellonian University, Gronostajowa 7, 30-387 Krakow, Poland
Fax: + 48 126646902
Tel: +48 126646151


Human transcription factor Yin Yang 1 (YY1) is a four zinc-finger protein that regulates a large number of genes with various biological functions in processes such as development, carcinogenesis and B-cell maturation. The natural binding sites of YY1 are relatively unconserved and have a short core sequence (CCAT). We were interested in determining how YY1 recognizes its binding sites and achieves the necessary sequence selectivity in the cell. Using fluorescence anisotropy, we determined the equilibrium dissociation constants for selected naturally occurring YY1 binding sites that have various levels of similarity to the consensus sequence. We found that recombinant YY1 interacts with its specific binding sites with relatively low affinities from the high nanomolar to the low micromolar range. Using a fluorescence anisotropy competition assay, we determined the affinity of YY1 for non-specific DNA to be between 30 and 40 μm, which results in low specificity ratios of between 3 and 220. Additionally, surface plasmon resonance measurements showed rapid association and dissociation rates, suggesting that the binding strength is regulated through changes in both ka and kd. In conclusion, we propose that, in the cell, YY1 may achieve higher specificity by associating with co-regulators or as a part of multi-subunit complexes.






surface plasmon resonance


Yin Yang 1


Yin Yang 1 (YY1) is a ubiquitous transcription factor belonging to the C2H2-type zinc-finger family. Depending on the promoter, YY1 may function as an activator or repressor of transcription. YY1 is engaged in regulation of a broad spectrum of cellular processes and transcription of some viral gene products [1,2]. Importantly, it is also involved in development [3] and DNA repair through homologous recombination [4]. YY1 is a member of multi-subunit complexes, such as the human INO80, that is involved in chromatin remodelling [5]. Pho, a Drosophila homologue of YY1, is involved in the functioning of Polycomb group complexes [6]. In both cases, YY1 is assumed to confer DNA-binding properties to the whole complex. YY1 recognizes its binding sites through four zinc fingers that are located at the C-terminus, of which the second and third finger contribute the majority of the contacts [7]. The N-terminal domain of YY1 is probably loosely structured, and contains elements that are distinguished by their amino acid composition, such as the two acidic regions [8]. This fragment of the protein does not bind by itself to DNA, and probably does not participate in DNA recognition [9,10].

The consensus sequence for YY1 is degenerate, and has the relatively short core sequences CCAT, which is more common, and ACAT. These core sequences were determined either from alignment of the confirmed binding sites or through in vitro selection from a pool of random sequences [1]. The interaction between YY1 and DNA has been investigated using electrophoretic mobility shift assay, and, in one case, by filter binding. Some of these studies reported apparent KD values ranging from 3 nm to 4 μm [11–13]. Biophysical studies concerning the details of the interaction between YY1 and DNA are limited in number. They include solution of the crystal structure of ΔYY1 (referred to here as YY1-DBD) bound to the adeno-associated virus P5 promoter +1 site (AAV+1) [7], and thermodynamic studies of the interaction between these molecules, which provided a dissociation constant of ∼ 0.5 μm [10]. However, this high binding constant was reported for only a single binding site with a specific biological function in transcription initiation, and, consequently, the result was not discussed in relation to the general function of YY1. Because YY1 regulates genes through a relatively unconserved binding site, we considered it important to investigate the mechanism of this interaction in more detail, and to determine how YY1 distinguishes among sequences to regulate gene expression, how affinities are related to the consensus, and how kinetics contribute to this recognition. We used diverse binding sites involved in the regulation of various biological functions to draw general conclusions about the function of YY1.

In this paper, we used fluorescence anisotropy and surface plasmon resonance (SPR) to measure the affinities and kinetics of the interaction between human YY1 or the YY1-DBD fragment and selected oligonucleotides. We used four naturally occurring sequences that show various similarities to the consensus sequence and a non-binding control sequence (Table 1). Moreover, we estimated the affinity of YY1 for non-specific DNA to investigate how YY1 selects specific sequences in vivo.

Table 1.   Oligonucleotides used in this study. The conserved core sequence is in bold, and the sequences are shown for the top strand only. The consensus sequence is shown at the bottom for comparison. AAV, adeno-associated virus; IgH, immunoglobulin heavy-chain enhancer; SRE, serum response element; HIV LTR, HIV long terminal repeat.
Name of the regulatory fragmentSequenceReference
IgH enhancer μE1 site (IgH) TGATCGG CCAT CTTGACTCC 14
AAV P5 promoter +1 site (AAV+1) AGGGTCT CCAT TTTGAAGCG 15
AAV P5 promoter –60 site (AAV–60) TTTTGCG ACAT TTTGCGACA 15
c-fos promoter SRE site (c-fos) AGGATGT CCAT ATTAGGACA 16
Non-specific sequence (NS) GAGATTAACGAGCTCAGTCC This work
Consensus sequence CGG CCAT CTTGNCTG 18


Fluorescence anisotropy measurements of the interaction between YY1 and DNA

We wished to determine the binding strength between YY1 and selected DNA sequences and to assess its dependence on the similarity of a given sequence to the consensus. We also performed a more thorough comparison of the DNA-binding strength of YY1 with the YY1-DBD fragment. To make this comparison, we performed fluorescence anisotropy measurements of the binding to six fluorescently labelled oligonucleotides. On the basis of these measurements, we determined the binding constants for four YY1-binding sequences; the remaining two sequences were non-binding. Fluorescence anisotropy is an equilibrium method that is based on detection of the polarized light emitted from a fluorescent probe attached to the molecule being studied. The degree to which the light is depolarized depends on the mobility of the fluorescent probe. Upon complex formation, the mobility of the probe is decreased, which leads to a decrease in depolarization of the emitted light. This allows the observation of complex formation at increasing protein concentrations. One of the advantages of this method is that the solution conditions can be directly and easily controlled.

The proteins that were used in this study were expressed in Escherichia coli and then refolded and purified on a DNA affinity column. To confirm that all of the cysteines in the refolded YY1 were in the reduced state, we performed a reaction with methyl methanethiosulfonate, a sulfhydryl-reactive compound. After the reaction of methyl methanethiosulfonate with YY1-DBD, the mass of the polypeptide, determined by ESI-MS, changed from 14 053.9 Da to 14 421.8 Da (Fig. S1). The difference of 367.9 Da is equal to the change in mass predicted for eight modified cysteines (a change of 46 Da per residue).

The sequences used in this study were chosen on the basis of confirmed binding sites that have been shown to be functional in the genome. Moreover, they had different similarities to the consensus sequence (Table 1). The IgH sequence is the closest to the consensus, followed by AAV+1, the AAV P5 promoter –60 site (AAV–60) and the c-fos promoter serum response element (c-fos). Three of them have a canonical CCAT core motif, and AAV–60 has an alternative ACAT core. Figure 1 shows the results of the fluorescence anisotropy measurements for binding between YY1 or YY1-DBD and the DNA sequences. In Fig. 1A,C, the binding curves for the IgH, AAV–60 and HIV long terminal repeat (HIV) duplexes are shown, and the binding curves for the AAV+1, c-fos and non-specific (NS) duplexes are shown in Fig. 1B,D. For the IgH and AAV+1 probes, the fluorescence anisotropy of the probe increased from ∼ 0.14 to ∼ 0.18 in the complex with YY1-DBD, or to ∼ 0.19 with YY1. The data were collected from at least two independent experiments and fitted to the single binding site model. YY1 binds to the IgH sequence with a KD of 0.2 μm, and to the weakest c-fos sequence with a KD of 12 μm. YY1-DBD binds the same sequences with KD values of 0.2 and 4.8 μm, respectively (Table 2). From the strongest to the weakest, the sequences ordered by their affinity for YY1 are IgH > AAV+1 > AAV–60 > c-fos. The same order applies to YY1-DBD, which suggests that the full-length and truncated proteins have similar sequence selectivities. We did not observe significant differences in the binding strength between those two proteins for any of the tested sequences, indicating that the N-terminal domain does not take part in DNA recognition, in agreement with a previous report [7]. The HIV and NS probes did not bind to YY1, as demonstrated by the lack of change in anisotropy values over the range of protein concentrations tested (Fig. 1). Thus, the data for these duplexes were not fitted because the change in anisotropy was low and did not allow reliable analysis. In addition, we did not observe even weak binding to the HIV site that is known to indirectly recruit YY1 [17]. Use of the NS sequence as a negative control also showed that the refolded recombinant transcription factor that we used is able to distinguish between specific and non-specific sites in our fluorescence anisotropy assay.

Figure 1.

 Fluorescence anisotropy measurements of the interaction between YY1 or YY1-DBD and DNA. The plots show the representative binding curves for the interaction between YY1 or YY1-DBD and six oligonucleotides that were fluorescently labelled with 6-ROX. The protein concentration is on the x axis, and the ratio of the measured anisotropy to the initial anisotropy [r/r(0)] is on the y axis. The solid lines show the least-squares fits to the 1 : 1 model. (A) Interaction of YY1 with the IgH (open inverted triangle), AAV–60 (open square) and HIV probes (open diamond). (B) Interaction of YY1 with the AAV+1 (open square with cross), c-fos (open triangle) and NS probes (open right-pointing triangle). (C) Interaction of YY1-DBD with the IgH (open inverted triangle), AAV–60 (open square) and HIV probes (open diamond). (D) Interaction of YY1-DBD with the AAV+1 (open square with cross), c-fos (open triangle) and NS probes (open right-pointing triangle).

Table 2.   Dissociation constants for the interaction between YY1 or YY1-DBD and DNA, obtained by fluorescence anisotropy measurements. The specificity ratios for YY1 were calculated based on the affinities for non-specific DNA obtained in competition experiments. Values are means ± standard deviation. CT, calf thymus; NS, non-specific.
K Dm) K Dm)Specificity ratio
IgH0.2 ± 0.10.2 ± 0.2180155220
AAV+10.8 ± 0.20.58 ± 0.04625376
AAV–602.8 ± 0.72.1 ± 0.2171521
c-fos4.8 ± 0.912.0 ± 8.0334

The effect of salt concentration on binding of YY1 to DNA

Next we assessed the salt dependence of the YY1–DNA interaction and compared this dependence among the various sequences. Using fluorescence anisotropy, we measured the interaction between YY1 and four DNA sequences (IgH, AAV+1, AAV–60 and c-fos) at various salt concentrations. This allowed us to investigate the salt sensitivity of the YY1–DNA interaction and to compare the salt dependence of various duplexes. The chosen range of salt concentration (50–125 mm) provided anisotropy responses that could be monitored and fitted with confidence, with the exception of the data for the c-fos duplex, that had to be analysed with a fixed molar response to obtain a correct fit. This approach is valid as the final anisotropy of the bound probe did not change at different salt concentrations or with different probe sequences (Figs 1 and 2A). As an example, Fig. 2A shows the results of fluorescence anisotropy for binding of YY1 to the IgH probe at four salt concentrations. The dissociation constants obtained from these experiments are shown in Table 3. As expected for a DNA-binding protein, the KD increased with increasing salt concentration. Next, we plotted the dissociation constants against the salt concentrations on a double-logarithmic plot, as shown in Fig. 2B. According to the theoretical method developed by Record et al. [19], the slope of the linear fit to the plot is related to the number of ion pairs formed upon complex formation (m′) by the equation:

Figure 2.

 Salt dependence of the YY1–DNA interaction. (A) Fluorescence anisotropy measurements for interactions between YY1 and the IgH duplex at 50 mm (open triangle), 75 mm (open circle), 100 mm (open square) and 125 mm (open diamond) NaCl. The black lines represent non-linear least-squares fits to the data. (B) Double-logarithmic plot showing the salt dependence of the interactions with four of the tested sequences, IgH (open triangle), AAV+1 (open circle), AAV–60 (open square) and c-fos (open diamond). The black lines show linear fits to the data. The parameters obtained are presented in Table 3.

Table 3.   Salt dependence of the YY1–DNA interaction. The slope was obtained from a linear fit to the double-logarithmic KD versus [NaCl] plot, and provides a measure of the salt dependence. The m′ parameter, which is the number of ion pairs formed, was calculated as described by Record et al. [19]. ND, not determined.
NaCl (mm)IgHAAV+1AAV–60c-fos
K Dm)Slope m K Dm)Slope m K Dm)Slope m K Dm)Slope m
 500.08 ± 0.01−1.91−2.20.15 ± 0.01−1.87−2.10.55 ± 0.05−2−2.31.91 ± 0.03−2−2.3
 750.14 ± 0.010.31 ± 0.031.0 ± 0.13.9 ± 0.2
1000.27 ± 0.030.58 ± 0.081.8 ± 0.37.9 ± 0.2
1250.42 ± 0.050.8 ± 0.33.0 ± 2.0ND

where ψ is a constant equal to 0.88 for the B-form double-stranded DNA. We obtained similar m′ values for all of the sequences (e.g. 2.1 for AAV+1 and 2.3 for AAV–60 and c-fos; Table 3). Consequently, YY1 has the same salt dependence for the various sequences tested, which indicates the same general mode of recognition. This relationship becomes more complex in the presence of Mg2+ ions and anions. Thus, the m′ parameter determined here is lower than it would be if the calculations accounted for the presence of magnesium. Therefore, this formula should be treated as an estimate that allows quantitative comparison of the various sequences.

Interaction of YY1 with non-specific DNA: assessment of the interaction specificity

Next, we determined the strength of the non-specific binding and compared it to the strength of the specific interaction. This information is important because the KD values for the interaction with specific sequences that we obtained were relatively high and were unusual for a specific transcription factor that must find its targets in the genome. To determine the binding strength, we performed competitive fluorescence anisotropy measurements in the presence of non-specific DNA. Competitive binding measurements were necessary because we could not measure the low-affinity interactions with non-specific DNA directly by either fluorescence anisotropy or SPR. In this method, binding of a protein to the ‘visible’ (fluorescent) ligand is monitored in the presence of an ‘invisible’ competitor molecule, which is accounted for in the model. In our experiments, we used the fluorescence anisotropy assay to measure the interaction between the IgH duplex and YY1 in the presence of various amounts of either calf thymus DNA or unlabelled 20 bp NS or HIV duplexes. For the experiments with calf thymus DNA, we assume that the concentration of overlapping non-specific binding sites is equal to the concentration of base pairs [20]. The measurements were performed with calf thymus DNA concentrations of 0, 57, 114 and 228 μm or with unlabelled duplex concentrations of ∼ 10 and 20 μm. The binding curves were collectively analysed using a method developed by P. Kuzmic and implemented in dynafit software [34]. Figure 3 shows the results of the competitive binding experiments. As expected, the binding weakened with increasing concentrations of calf thymus DNA. The data were fitted with both KD and KDc (where ‘c’ stands for competitive) as unknown parameters. From this model, we determined the strength of the non-specific binding of YY1 to the NS and HIV duplexes to be 31 ± 6 and 46 ± 12 μm, respectively (Fig. 3B,C). For the experiments with calf thymus DNA, the KDc was determined to be 36 ± 5 μm (Fig. 3A). The values for the specific interaction with the IgH duplex were determined correctly by this analysis, and this further supports our conclusions about the values for the non-specific interaction. Using the KDc values, we then calculated the specificity ratios, which are the ratios of non-specific binding to the specific binding strength. This parameter reveals how much a given transcription factor prefers its specific binding site over a random, non-specific sequence. The specificity ratios varied between 3 and 4 for the c-fos duplex, and, depending on the competitor DNA used, between 155 and 220 for the IgH duplex (Table 2).

Figure 3.

 Determination of the non-specific binding constant using a fluorescence anisotropy competition assay. The figure shows binding curves obtained using fluorescence anisotropy for the interaction between YY1 and the IgH duplex at increasing concentrations of non-specific competitor DNA. (A) Calf thymus DNA was used as a competitor, with concentrations of base-pairs (binding sites) equal to 0 μm (open inverted triangle), 57 μm (open circle), 114 μm (open square) and 228 μm (open triangle). (B) Non-labelled NS duplex was used as a competitor at concentrations of 0 μm (open triangle), 10.5 and 11.5 μm (open circle), or 20.5 and 23 μm (open square). (C) Non-labelled HIV duplex was used as a competitor at concentrations of 0 μm (open triangle), 10.5 and 11.5 μm (open circle), or 20.5 and 23 μm (open square). The solid lines represent non-linear least-squares fits to the data. Values for the dissociation constants for IgH (KD) and for the non-specific DNA (KDc) are shown.

Determination of the kinetics of the YY1-DBD/DNA interaction using SPR

To obtain further insight into the mechanism of the YY1–DNA interaction, we performed SPR experiments that allowed us to determine kinetic rate constants for this interaction. We were interested in determining how the rate constants regulate the affinity of YY1 for various sequences. Because SPR is not an equilibrium method and the obtained parameters depend on various factors, we discuss only the relative binding parameters. Towards this end, we measured binding of YY1-DBD to the three strongest sequences (IgH, AAV+1 and AAV–60), and to HIV as a non-binding sequence, immobilized on the surface of a streptavidin-coated chip. The protein concentrations of the analytes were between 10 and 200 nm. The sensorgrams obtained for the interactions of YY1-DBD with the specific sequences were overlaid with fits to the 1 : 1 Langmuir binding model (Fig. 4A–C). The observed binding was specific because only a weak signal was observed for the surface with the HIV duplex (Fig. 4D). The obtained rate constants ka, kd and KD for these interactions are listed in Table 4. The association and dissociation rates were relatively rapid, and both varied between sequences. The association rate constants were 3.4 × 106 M−1 s−1 for IgH and 1 × 106 M−1 s−1 for AAV–60, and the dissociation rate constants were 0.15 s−1 and 0.32 s−1, respectively. Both the ka and kd changed by approximately threefold between the strongest and the weakest binding sequence. The resulting KD values were 0.044 μm for IgH and 0.3 μm for AAV–60, and, although they were smaller than those obtained by fluorescence anisotropy, the order of binding strength did not change, i.e. IgH > AAV+1 > AAV–60.

Figure 4.

 Surface plasmon resonance characteristics for the interaction between YY1-DBD and YY1 with DNA. Global fits to the 1 : 1 binding model are shown as solid red lines. (A) Sensorgrams for the binding of YY1-DBD to immobilized IgH; including data that were obtained after repeated injections of the analyte (analyte concentrations: 10, 20, 40, 40, 60, 80, 100, 100, 150 and 200 nm). (B) Binding of YY1-DBD to AAV+1 (analyte concentrations: 10, 20, 30, 40, 50, 60, 70, 80 and 100 nm). (C) Binding of YY1-DBD to AAV–60 (analyte concentrations: 10, 20, 40, 60, 80, 100, 150 and 200 nm). (D) YY1-DBD was injected onto the surface that contained the immobilized HIV duplex (analyte concentrations: 20, 40, 100 and 200 nm). (E) Salt dependence of the interaction of YY1-DBD with IgH at a 25 nm analyte concentration and 50, 75, 100 and 125 mm NaCl. (F) Interaction between YY1 and immobilized IgH (analyte concentrations: 0, 25, 50 75, 100, 200, 200, 400 and 800 nm).

Table 4.   Kinetic and thermodynamic parameters of the YY1–DNA interaction. The values were determined from the SPR measurements using biotinylated duplexes as the ligand and YY1-DBD as the analyte.
Sequence K Dm) k a (M−1 s−1) k d (s−1)
IgH0.0443.41 E+060.152
AAV+10.1152.17 E+060.249
AAV–600.3091.03 E+060.319

We also measured the interaction of full-length YY1 with the IgH, AAV+1 and AAV–60 duplexes. YY1 showed an interaction with all of the tested sequences. Unfortunately, it also presented a complex binding response that did not reach saturation, even when the association phase was increased to 5 min. These data did not reliably and repeatedly fit any standard model that was included with the software. An example sensorgram is shown in Fig. 4F for the interaction of YY1 with the IgH duplex. This binding behaviour cannot simply be explained by the presence of the N-terminal domain because the YY1 N-terminal domain (amino acids 1-295) did not interact with the IgH sequence when measured using SPR (data not shown).

Additionally, we determined the salt dependence for the YY1-DBD:IgH interaction. Sensorgrams were obtained for binding at various salt concentrations. Both the association and dissociation rates changed; however, the largest differences were observed for the dissociation rate constants, which varied between ∼ 0.01 s−1 in 50 mm NaCl and 0.3 s−1 in 125 mm NaCl (Fig. 4E).


Using fluorescence anisotropy, we determined the dissociation constants for binding of the human YY1 transcription factor to four specific binding sites and quantitatively established the dependence of binding strength on the similarity of a given sequence to the consensus. The affinities, determined under true-equilibrium conditions, were in the high nanomolar to low micromolar range. In addition, we determined the binding constant for the non-specific DNA to be between 31 and 46 μm, resulting in relatively low specificity ratios of between 3 and 220. The SPR measurements for the association of YY1-DBD with specific DNA sequences revealed rapid on and off rates for the association, with ka values between 3.4 × 106 and 1 × 106 M−1 s−1 and kd values between 0.15 and 0.32 s−1. In addition, the salt dependence of the YY1–DNA interaction was found to be similar for all of the binding duplexes tested and to be regulated mainly by the dissociation rate, which suggests a similar mode of interaction for different sequences.

The KD value reported here is in good agreement with that determined by calorimetric measurements reported by Houbaviy and Burley [10]. These researchers performed thermodynamic studies of the interaction between ΔYY1 and the AAV+1 sequence, and reported a KD of 0.56 ± 0.03 μm at 25°C that is comparable to the KD of 0.8 ± 0.2 μm reported here. This value is comparatively high: other zinc-finger transcription factors have affinities in the low nanomolar or sub-nanomolar range, such as Zif268, with a KD of 6.5 nm (by SPR) [21] or 0.01 nm (by electrophoretic mobility shift assay) [22]. Several groups have reported apparent KD values for the YY1–DNA interaction, obtained from electrophoretic mobility shift assay experiments, in the low nanomolar range (between 3 and 40 nm) [11,16], with one exception, in which an apparent KD value of 4 μm was reported [12]. In our opinion, the differences between these values cannot be explained solely by the lower salt concentration used in the binding reactions, which were between 30 and 50 mm, and instead, they may result from differences in the methods. Unlike electrophoretic mobility shift assay or filter binding, both fluorescence anisotropy and isothermal titration calorimetry allow for equilibrium measurements. Thus, we believe that the values reported in our study represent a true measurement of the binding strength. This assertion is also supported by the fact that the protein was able to distinguish between specific and non-specific (NS) sequences in the direct anisotropy assay.

The competitive binding experiments allowed us to determine the KD for the interaction of YY1 with non-specific DNA. Depending on the non-specific DNA used, we obtained dissociation constants between 31 and 46 μm. It is important to note that the non-specific binding affinity that was determined using calf thymus DNA depends on the assumed number of binding sites. Here, we have assumed the maximum number of non-specific binding sites to be constant and equal to the concentration of base pairs. Therefore, this model does not account for the reduction in the number of binding sites when a ligand covers more than 1 bp, as described by McGhee and von Hippel [22]. However, such effects will be most pronounced at high protein concentrations and for strongly binding ligands. Moreover, the results obtained using calf thymus DNA match those from the experiments in which we used defined, weakly binding sequences (NS and HIV) as competitors. For these experiments, the number of non-specific binding sites is equal to the concentration of the DNA duplex. This suggests that the model used for calculation of non-specific binding sites in calf thymus DNA was correct.

Other transcription factors, including zinc fingers, display higher affinities for non-specific sequences. For Zif268, the affinity is probably ∼ 300 nm for calf thymus DNA, which results in a high specificity ratio of 31 000 [23]. The two zinc-finger protein ADR1 yeast transcription factor binds to a random nucleotide on the SPR chip with affinities between 440 nm for the full-length protein and > 5 μm for the zinc-finger domain only, and it displays low specificity ratios between 7 and 12 [24]. For the trp repressor, the specificity ratio was ∼ 1000 [25]. The affinity of YY1 for calf thymus DNA is rather low but still resulted in low specificity ratios. A high specificity ratio is important if a given transcription factor is to recognize specific regulatory sites in the genome. It is possible that YY1 may achieve higher specificity through interaction with co-factors, and this possibility is discussed below.

Several other groups have performed SPR kinetic analyses for zinc-finger proteins [21,24,26]. For the two zinc-finger binding domain of ADR1, the ka value was reported to be 2.64 × 104 M−1 s−1, and the kd value was 1.5 × 10−2 s−1 [24]. For the well-studied three zinc-finger protein Zif268, the ka value was determined to be 3.1 × 104 M−1 s−1, and the kd value was 2 × 10−4 s−1 [21]. An extreme case is the seven zinc-finger protein Zap1, a yeast transcriptional activator, four of whose zinc fingers interact with DNA, with a rapid ka value of 5.7 × 107 M−1 s−1 and a moderate kd value of 3 × 10−2 s−1 [26]. In comparison, the association rate constants for YY1-DBD are relatively fast but not faster than the values obtained for other zinc-finger proteins. However, the kd value for YY1-DBD is ∼ 10–1000 times faster than those for other zinc-finger proteins. In addition, in YY1-DBD, both the ka and kd values change with different sequences, and both influence the affinity. The KD values determined by SPR were lower than those determined by fluorescence anisotropy, but the relative strength of the binding by the three measured sequences was retained. This may result from the differences in the two methods, i.e. immobilization of the DNA to the negatively charged carboxymethylated dextran surface of the chip. Compared to SPR, fluorescence anisotropy is a true-equilibrium technique, and the binding affinities determined here are in agreement with previous isothermal titration calorimetry results [10].

YY1 is frequently associated with other factors, and this association may influence the stability of the complex with DNA. Examples include the reported cooperative association with TFIIB at the AAV+1 site [12] and the cooperative binding of SRF and YY1 at the serum response element of the c-fos promoter [27]. In addition, YY1 was shown to be associated with the Late SV40 Factor (LSF) on the HIV long terminal repeat, but only as a co-factor and probably without contact with the DNA [17]. Additionally, on the basis of the proximity of the Pax5- and YY1-binding sites, it was proposed that these transcription factors may bind cooperatively to DNA and with the assistance of other factors to regulate IgH locus contraction [28].

The selected sequences differ in their biological functions and in the way that they regulate transcription. Binding of YY1 to the IgH site was shown to be required for heavy-chain gene rearrangement [28], while the AAV–60 sequence contains the alternative ACAT core sequence that was found at some natural sites as well as in an in vitro selection experiment. This site is thought to repress transcription, and binding of the adenovirus E1A protein turns YY1 into an activator [15,29]. The AAV+1 site functions as an initiator element, and is required to start transcription in the presence of YY1, TFIIB and RNA Polymerase II [12]. In addition, in a circular permutation assay, the c-fos site was shown to bend on binding to YY1 [30].

We suggest that the relatively weak binding and low specificity ratio of YY1 may be because YY1 evolved to function with other transcription factors and co-regulators, or, alternatively, as part of multi-subunit complexes. Such additional interactions may increase its affinity towards selected binding sites, and, as a result, change the DNA-binding landscape. This hypothesis is supported by the finding that global regulators have lower specificities, higher expression levels and more associated co-regulators in the bacterial regulatory networks of E. coli and Bacillus subtilis [31].

On the basis of our findings, we propose two mechanisms that involve cooperative interactions through which YY1 is able to recognize specific sites. In the first possible scenario, YY1 is associated with other proteins that together form an active complex that recognizes specific binding sites. In such a complex, the affinities and global sequence specificities, based on the affinities for different sites, may change in comparison with the ‘free’ protein. An example of such a complex involving YY1 is the human INO80, an ATP-dependent chromatin-remodelling complex. This complex was sufficiently stable to be isolated with YY1 and several other peptides from HEK293 cells, which may suggest that YY1 is an integral part of this complex even in the absence of DNA [4,5]. Another example is the Polycomb group repressive complex that is involved in development and is present in both Drosophila and mammals [6]. YY1 was shown in vivo to be important for the function of the complex; however, its components did not co-purify with YY1, probably due to the low stability of the complex [32]. The second possibility is that YY1 scans the genome, forms a transient complex at a specific site, and is then bound by other factors that stabilize it, possibly by decreasing the dissociation rate.

In summary, we have found that YY1 binds with low affinity, and the interaction is characterized by a low specificity ratio. Furthermore, kinetic dissection of the binding process indicates that it is dynamic, with rapid on and off rates. These results suggest a mechanism in which recognition of specific binding sites by YY1 is aided by additional co-regulators.

Experimental procedures


Reagents for the buffers used in the protein purification and binding experiments were purchased from Sigma-Aldrich (St Louis, MO, USA) or Bioshop (Burlington, Canada). PEG 20 000 was purchased from Serva (Heidelberg, Germany). Lyophilized calf thymus DNA (sodium salt) and BSA were purchased from Sigma-Aldrich. All oligonucleotides for the anisotropy and SPR experiments were synthesized and HPLC-purified by Metabion (Martinsried, Germany) or Genomed (Warsaw, Poland). The sensor chips and reagents used for SPR were from Biacore (Uppsala, Sweden).

Protein expression and purification

Proteins were expressed and purified essentially as described previously [33]. Briefly, human YY1 and the YY1-DBD fragment (residues 293–414) were expressed in E. coli BL21 and Rosetta 2 cells, respectively. The cells were lysed in buffer containing 6 m guanidinium chloride, and the recombinant proteins were purified under denaturing conditions using Ni-NTA resin; the resulting fractions were pooled and stored frozen until refolding. The YY1-DBD protein without a His6 tag was purified as described above, except that imidazole was excluded from all of the chromatographic steps except elution. Before refolding, the protein concentration was adjusted to 1 mg·mL−1, and the mixture was heated at 60°C for 30 min. Refolding was performed against three changes of binding buffer (25 mm Tris, 100 mm NaCl, 10 mm MgCl2, 5 mm dithiothreitol, 0.1 mm ZnCl2 and 0.02% w/v sodium azide, pH 8.5) at 4°C. The refolded proteins were then purified on a DNA affinity column using a short 1.5-column volume washing step and step-gradient elution, dialysed to remove excess salt and stored at -80°C. The protein purity was assessed by SDS/PAGE, and was found to be ≥ 95% pure. The correct mass of the purified proteins was confirmed on an ESI-Q-TOF mass spectrometer (Bruker Daltonik, Bremen, Germany). The sample homogeneity was confirmed by size-exclusion chromatography. The protein concentration was determined by the Bradford assay [34].

For the reaction with methyl methanethiosulfonate, 100 μL of YY1-DBD in binding buffer was mixed with 50 μL of 206 or 21 mm methyl methanethiosulfonate in TBS (25 mm Tris and 100 mm NaCl) and with 0.05% P20 detergent (Biacore). The reaction proceeded for 3 min at room temperature, and was stopped by addition of 50 μL of 2% TFA in 20% acetonitrile. The samples were then purified using C18 spin columns (Pierce, Rockford, IL, USA) and analysed on an ESI-Q-TOF mass spectrometer (Bruker).

Preparation of the DNA

The oligonucleotides used for the fluorescence measurements were labelled at the 5′ end of the antisense strand using 6-carboxy-X-rhodamine (6-ROX). The oligonucleotides used for SPR were labelled with biotin on the 5′ end of the same strand. The sequences of the oligonucleotides used in this study are shown in Table 1. Lyophilized oligonucleotides were dissolved in deionized water and hybridized by heating at 95 °C for 5 min, and were then slowly cooled for several hours and frozen in aliquots at a concentration of 100 μm. The DNA concentration used was as specified by the manufacturer. Lyophilized calf thymus DNA was thoroughly dissolved in deionized water, sonicated using a water/ice bath sonicator (Bioruptor; Diagenode, Liège, Belgium) and filtered through a 0.22 μm (4 mm in diameter) filter (Millipore, Bedford, MA, USA). The ratio for absorbance at 260 and 280 nm was 1.6. The size of the calf thymus DNA was analysed by agarose gel electrophoresis and was ∼ 250 bp (Fig. S2). All DNA samples were stored at −20 °C.

Fluorescence anisotropy measurements

Fluorescence anisotropy was measured using a Fluorolog 3-21 L-format spectrofluorometer (Horiba Jobin Yvon, Longjumeau, France) equipped with a thermojacket connected to a temperature-controlled water bath set at 25 °C. Measurements were performed by monitoring the emission from 6-ROX excited at 570 nm, using a detection monochromator set at 610 nm with slit widths of 9 nm for excitation and 7 nm for emission and an integration time of 0.5 s. The samples were measured in a 3 mm × 3 mm fluorescence microcuvette (Hellma, Müllheim, Germany). Each measurement was repeated three times, and the mean was calculated to give an individual data point. The samples were prepared in a 60 μL volume in binding buffer supplemented with fresh 5 mM dithiothreitol, 0.1 mM ZnCl2, 0.1 mg·mL−1 BSA and 0.04% w/v PEG 20 000. Samples with a constant amount of DNA (50 nm) and different protein concentrations were mixed by gentle vortexing, centrifuged for 5 min at 25 000 g at 4 °C, and kept on ice. Before measurement, the samples were incubated for at least 10 min at 25 °C to bring them to equilibrium. For the binding experiments at different salt concentrations, the samples were prepared in the same way, and the desired salt concentration was achieved by dilution with binding buffer without NaCl or by the addition of NaCl from a 5 m stock prepared in water.

The data were analysed using a non-linear least-squares algorithm implemented in the dynafit program (version 3.28) (BioKin, Watertown, MA, USA) [35], and were fitted to a single binding site model. The data from each series were fitted independently, and the reported KD values are the mean at least two independent measurements with between 30 and 74 data points. The error in KD is given as the standard deviation. For experiments at different salt concentrations, the reported error was determined using dynafit software.

For estimation of the strength of non-specific binding, we performed competitive binding measurements in the presence of calf thymus DNA or non-labelled NS or HIV duplexes [36]. This method allows determination of the KD of a ligand that cannot be followed fluorescently provided that its concentration is known. The method assumes that a visible ligand (A*) with a known dissociation constant (KD) competes for binding with an invisible ligand (A) with an unknown dissociation constant (KDc). The samples for the anisotropy assay were prepared in binding buffer, as described for other experiments, and were then mixed with a constant amount of competitor DNA. Calf thymus DNA was added from a 0.79 mg·mL−1 stock solution in binding buffer containing BSA and PEG 20 000. The NS and HIV duplexes were added from 0.41 or 0.46 mm stock solutions, respectively. The complete dataset contained fluorescence anisotropy measurements for binding between YY1 and the IgH probe performed at various concentrations of non-specific DNA or without the competitor. The data were analysed using a model implemented in the dynafit software, according to the reaction schemes: A* + B = A*B and A + B = AB. Both KD and KDc were simultaneously fitted. The concentration of non-specific overlapping binding sites in the calf thymus DNA was taken to be equal to the concentration of base pairs determined from the absorbance at 260 nm using a molar extinction coefficient of 6 600 m−1·cm−1 [37]. The error was determined using dynafit software.

Surface plasmon resonance

The SPR experiments were performed using a Biacore 3000 instrument (Biacore). The DNA was biotinylated on the 5′ end of the antisense strand and was immobilized on a CM5 sensor chip with streptavidin covalently attached to a carboxymethylated dextran (Biacore). Before immobilization, the chip surface was washed with the running buffer (10 mm Hepes and 150 mm NaCl, pH 7.4) until a stable baseline signal was obtained and activated according to the manufacturer’s instructions. For immobilization, the DNA stock solution was diluted to 0.05 ng·μL−1 in 0.5 m NaCl, and injected into the cells at a flow rate of 2 μL·min−1 for 3 min. The chip surface was then washed at 20 μL·min−1 with subsequent solutions of 0.5 m NaCl, 1 m NaCl and 0.1% SDS, for 3 min each, to remove non-specifically bound ligand. The amount of coupled DNA was adjusted to 90–190 response units, such amount was suitable for the kinetic analysis. For all binding experiments, the chip surface was washed with the running buffer (10 mm Tris, 100 mm NaCl, 10 mm MgCl2, 5 mm dithiothreitol and 0.1 mm ZnCl2, pH 7.9, with 0.05% P20 surfactant). A series of YY1-DBD solutions was injected over the sensor chip surface at a 30 μL·min−1 flow rate with 2 min for association and 2 min for dissociation. YY1 was injected with 1 min association and 2 min dissociation times. A flow cell without DNA was used as the control. After each analysis, the chip surface was regenerated for 1 min using a 1 m NaCl solution. The surface regeneration was confirmed by repeating the analyte injection (Fig. 4A,F). YY1 and YY1-DBD binding was not influenced by mass transfer effects. All measurements were performed at a constant temperature of 25 °C. For the measurements at different salt concentrations, the chip was washed with running buffer containing the selected salt concentration. The protein samples were prepared by dilution of a 13 μm protein stock in the selected running buffers. The collected data were analysed using BIAevaluation software version 4.1 (Biacore). Sensorgrams corrected for the reference were aligned, and a blank run was subtracted. The resulting dataset was analysed by simultaneous fitting of ka and kd to the 1 : 1 binding model.


We would like to thank Dr Sylwia Kedracka-Krok and Artur Pirog for performing the mass spectrometry verifications of the purified proteins. This research was supported by grant number 3128/P01/2006/31 from the Polish Ministry of Science and Higher Education. The Faculty of Biochemistry, Biophysics and Biotechnology of Jagiellonian University is the beneficiary of structural funds from the European Union (grant number POIG.02.01.00-12-064/08 –‘Molecular Biotechnology for Health’). The MS measurements were performed using equipment purchased using the European Regional Development Fund under the framework of the Polish Innovation Economy Operational Programme (contract number POIG.02.01.00-12-167/08, project Malopolska Centre of Biotechnology).