Human MEF2A cDNA was amplified by PCR from pMT2-MEF2A plasmid [3] using the following oligonucleotides, 5′ to 3′: sense CATGCCATGGCCACCATGGGGCGGAAGAAAATAC (NcoI site underlined) and antisense CCGCTCGAGGGTCACCCACGCGTCCATC (XhoI site underlined). A Kozak sequence is present between the NcoI site and the initiation methionine codons in the sense primer. The PCR product was cut with NcoI and XhoI and ligated in pET28a vector (Novagen) in frame with the C-terminal 6-Histidine tag. His-MEF2A was produced in E. coli BL21 upon isopropyl-1-thio-ß-d-galactopyranoside (IPTG) induction. Bacteria were lysed in 50 mM Tris pH 8 solution containing 300 mM NaCl, 0.5% NP40 and 1 mM PMSF and sonicated. Soluble proteins were then incubated with NiTA agarose beads (Qiagen) in the presence of 15 mM imidazole. After extensive washes, bound proteins were eluted with 250 mM imidazole and dialyzed against 20 mM Tris pH 7.5 solution containing 10% glycerol and 1 mM DTT, at 4 °C. The purified proteins were analysed by SDS-PAGE and Coomassie blue staining or by Western blot using an anti-Histidine (PentaHis, Qiagen) or an anti-MEF2 antibody (C21, SantaCruz). Immunoreactive proteins were detected by chemiluminescence (ECL, Amersham Biosciences). For the estimation of MEF2 diffusion time, the protein was labelled using the ‘fluorolink Cy5 monofunctional dye 5-pack’ (Amersham-Pharmacia).

The wild-type (WT) and the mutant (Mut) DNA used in the assays were derived from the mouse muscle creatine kinase (MCK) promoter: WT Oligo: 5′ TAMRA-CTCGCTCTAAAAATAACCCTGTC 3′ and its complementary strand without label; Mut. Oligo: 5′ TAMRA-CTCGCTCTAAggcTAACCCTGTC 3′ and its complementary strand. Nucleotides in bold represent the MEF2 binding site [3], mutated nucleotides are in lower case. These HPLC-purified 23-mer DNAdeoxyribonucleotides were purchased from Sigma-Genosys. Annealing was performed by heating to 80 °C a solution of 1:1 ratio of the complementary strands, followed by slow cooling to room temperature. The efficiency of the annealing procedure was verified by electrophoresis of the oligonucleotide duplexes on acrylamide gel. The TAMRA-labelled DNAs were purified through a Sephadex G-50 column (Amersham) for removing the free fluorochrome.

FCS study was performed on the Confocor 2 system (Carl Zeiss, Jena, Germany) using a 40 X water immersion C-Apochromat objective lens (N.A. = 1.2). The measurements were carried out at room temperature in eight-well Lab-Tek I chambered coverglass (Nalge Nunc International). The 543-nm He–Ne laser beam was focussed into 20-μl solutions at 180 μm over the cover glass. The fluorescence emission was collected through a pinhole and a 560-nm-long pass filter. Photon counts were detected by an Avalanche PhotoDiode (APD) at 20 MHz for 30 s. For each sample, FCS measurements were repeated 10 times. The data evaluation was performed using the Zeiss FCS Fit software. Most of the intensity autocorrelation curves were fitted using a free diffusion model with three components: the fluorochrome alone, the labelled DNA and the fluorescent DNA–protein complex. Preliminary studies enabled us to fix the diffusion time values of the first two components and the percentage of the first component (fluorochrome). Moreover, a calibration step with 4 nM Rhodamine 6G (R6G) made it possible to evaluate the size of the confocal volume (≈0.4 fl).

Interaction assays were performed at room temperature in the following binding buffer (25 mM Hepes pH 7.9, 50 mM KCl, 0.5 mM EDTA pH 8, 5 mM MgCl₂, 5 mM DTT, 10% glycerol, 0.002% NP40). 5 to 400 nM of MEF2A proteins (typically: 100 nM) were mixed with 5 nM DNA and 0.5 μg poly (dI)(dC) in 20 μl of binding buffer.

Diffusion constants were compared to theoretical values calculated assuming globular or rod-like macromolecules. The diffusion constants of the globular molecules were calculated using:

The theoretical diffusion coefficient D of the oligonucleotide was calculated using [16]:

Given that MEF2 proteins are assembled as homo- and heterodimers and that dimerisation is a prerequisite for DNA-binding [9,11], and given the short size of the oligonucleotide, we have considered the binding of one protein dimer to a single binding site. The DNA-binding reaction may be represented by:

DNA∗ and Complex∗ are both fluorescent ( ^* ), and their concentrations may be evaluated using FCS. Therefore, the binding constant K can be evaluated according to the following equation:

Moreover, the complex fraction, as expressed by Eq. (5), can be rearranged as Eq. (6). Therefore, this form was used for the estimation of the dissociation constant (KD), from non-linear curve fitting (Y; f([free dimer]).

Excel (Microsoft, USA) and Origin (OriginLab, USA) software were used for data evaluation.

MEF2A protein was produced from E. coli upon IPTG induction. Bacteria lysates were then purified on nickel affinity column by the mean of a 6-histidine tag. The His tag was positioned at the carboxy terminus of the protein in order to minimize the possible interference with the N-terminal DNA-binding and dimerization domains (Fig. 1A). Coomassie blue staining of purified MEF2A showed the presence of the expected 62-kDa full length protein (Fig. 1B). Two minor bands were co-purified with His-MEF2A, one migrating around 50 kDa and the other close to 80 kDa. Like full length MEF2A, the 50-kDa protein was recognized by an anti-histidine and an anti-MEF2 antibody directed against the C-terminus of the protein (Fig. 1C). Therefore, the 50-kDa protein is likely to represent a truncated MEF2 protein. The identity of the 80-kDa protein is unknown. The DNA-binding ability of the purified MEF2A was verified by electrophoretic mobility shift assays using the MEF2 binding site derived from the muscle creatine kinase promoter that was used in subsequent FCS analyses (see sequence in materials and methods). These assays indicated that His-MEF2A was capable of binding to the MEF2 binding site, whereas His-MEF2A binding was not observed with a mutated DNA probe (data not shown). Thus, purified His-MEF2A appeared to be functional for DNA-binding and the protein was used in FCS experiments to characterize the interaction between MEF2 and its cognate DNA sequence.

3.2.1 Diffusion parameters of free molecules

The diffusion coefficients of the different fluorescent molecules used in the assays were first estimated individually. Measurements were performed with about 4 nM sample solutions in PBS containing or not 10% glycerol. Autocorrelation curves are shown in Fig. 2 and the diffusion coefficients are listed in Table 1. As expected, mean diffusion times through the confocal volume increased both with molecular weight (Fig. 2, compare 2 to 3) and medium viscosity (compare 1 to 2) and were correlated with a decrease in diffusion coefficients (Table 1). The addition of glycerol led to a 1.2 fold increase of the diffusion times due to the change in viscosity. The medium viscosity was evaluated to 1.3 by comparing the diffusion measurements of TAMRA and Cy5 with theoretical data assuming a globular model (Eq. (1)). In our subsequent interaction assays, Rhodamine 6G, TAMRA and TAMRA–DNA were analysed at the beginning of each experiment. Rhodamine 6G analysis was used to estimate the confocal volume. On average, at 543 nm, the confocal volume was close to 0.4 fl with an axial axis of 0.23 μm, and a longitudinal axis of 2.7 μm. At the longest 633-nm wavelength used for Cy5 excitation, the axial axis was larger, about 0.3 μm. Corrections for the differences in volume were used to standardize the results. This explains why Cy5-MEF2A (λexc=633nm) and TAMRA–DNA (λexc=453nm) have different diffusion times but equal diffusion coefficients (Table 1). Moreover, slightly different diffusion coefficients of Cy5-MEF2A and the DNA reflect the differences both in molecular weight and in shape: globular vs. rod-like (Eqs. (1) and (2)). In summary, our results illustrate that, as expected, diffusion coefficients strongly depend on viscosity, molecular mass and overall shape of the molecules. In our assays, preliminary analysis of TAMRA and TAMRA–DNA allowed us to set their diffusion times at fixed values for the analysis of interaction. A stepwise procedure improved the quality of the fits with additional components. The analysis of labelled MEF2A was performed in order to get an estimation of the protein diffusion (= 411 μs); however, in the interaction experiment, the label was on the DNA and not on MEF2A protein. The measured value is between the expected values for MEF2A dimer (451 μs) or monomer (357 μs), according to a globular model and under our experimental conditions at 633-nm excitation. However, due to the shape approximation involved in simple hydrodynamic models, FCS does not make it possible to discriminate, in our opinion, between monomer and dimer protein, only on the diffusion data. Our labelling process does not allow us to include additional data, such as concentration or count per molecule.

Table 1

Diffusion coefficients of the individual components

	Molecular mass	Diffusion time (μs)	Diffusion coefficient (cm2 s−1)
	(g/mol)	in PBS	in 10% glycerol	Measured in 10% glycerol	Theoretical (1.3 viscosity)
Rhodamine 6G	479	38.6 ± 2.6⁎	50.3 ± 2.7	2.9 ± 0.3×10−6	2.9 × 10−6
TAMRA	430	35.0 ± 2.6	50.1 ± 3.2	2.8 ± 0.3×10−6	2.8 × 10−6
Cy5	792	62.5 ± 1.9	80.1 ± 1.4	2.8 ± 0.2×10−6	2.6 × 10−6
Cy5-MEF2A	54 800 °	–	411 ± 63	5.5 ± 0.3×10−7	6.3 × 10−7○
					5.0 × 10−7○○
TAMRA–DNA	14 614	188 ± 29	220 ± 23	6.3 ± 0.5×10−7	7.0 × 10−7⁎⁎
(23 mer)

^⁎ Mean ± s.d.,

^○ monomer,

^○○ dimer,

^⁎⁎ without TAMRA, rod-like model.

3.2.2 Interaction analysis between MEF2A and TAMRA-labelled DNA

Interaction assays were first performed using MEF2A at a concentration of 100 nM and the wild type DNA (WT DNA) or the modified sequence (Mut. DNA). The normalised autocorrelation curves obtained with the WT DNA differ significantly from those obtained with the Mut. DNA or the DNA alone (Fig. 3A). The differences between the curves were more pronounced using higher concentrations of MEF2A (Fig. 3B). In this assay, the curves were shifted to the right, indicating the presence of, at least, an additional slowly diffusing component. The Mut. and WT autocorrelation curves were similarly fitted using a two- or three-component free diffusion model. The diffusion times of the first two components (TAMRA alone and TAMRA–DNA) were fixed from the experiments using these components alone, performed at the beginning of each experiment (see Table 1, for example). A third component was only found in the WT DNA case. The diffusion time of this component was significantly higher than that of the free DNA (>400 μs) and corresponds to the DNA–MEF2A complex. In the case of the Mut. DNA, the fit was not improved using the three component model, or the third diffusion value was close to that of the free DNA or the associated percentage was not significant. In summary, a significant diffusion change resulting from the binding of MEF2A to DNA was only observed with a consensus MEF2 binding site, whereas a mutated DNA sequence did not allow the interaction with MEF2A.

Interaction assays of MEF2A and DNA were performed using various MEF2A concentrations (5 to 400 nM). The results were adjusted using a three-component model: the fluorochrome, the fluorescent DNA and the MEF2A–fluorescent DNA complex. The complex ratios were estimated in each experiment and expressed relatively to the total concentration of MEF2A–TAMRA–DNA and TAMRA–DNA, excluding the free TAMRA fluorochrome (Eq. (5) in § Material and methods). The titration curve (Y=f[free protein]) was then non-linearly adjusted using the Levenberg–Marquardt algorithm (Origin software) (Fig. 4) and the apparent equilibrium binding constant (KD) of MEF2A dimers was estimated to 60±4 nM (Eq. (6), χ2=0.003). The best fit was found using the simple model (Eq. (6)), and cooperative binding was not observed, in agreement with crystallography data [14,15]. A binding constant equal to 110 nM was reported for MEF2C [18,19], but a significantly higher affinity (KD=0.5 nM) was published for MEF2A [20]. This difference might be explained by the fact that, in this study, only the N-terminal region (amino acids 2–86) containing essentially the MADS-box and the MEF2 domain and another DNA length (60 mers) were used. Moreover, in our study, poly (dI)(dC) was added in the interaction buffer to prevent non-specific binding. Precise estimation of theoretical diffusion values for DNA/protein complex is difficult; however the diffusion time of the complex (>400 μs, typically 550 μs) is significantly higher than that of the free DNA and corresponds to the DNA–MEF2A complex with MEF2A most likely in the dimeric form. However, the presence of few DNA–multimer complex, corresponding to the second band showed by electrophoretic mobility shift assay by Berger et al. [20], or the contribution of the minor protein bands (Fig. 1) cannot be totally rejected. To gain further information, cross correlation [21–23] studies could be performed. Their main advantage would be to provide concentration and diffusion data both on the protein, the oligonucleotide and their interaction. However, exact standardization and the use of fluorochromes with shorter excitation wavelengths, such as FAM, for diminishing spectral crosstalk with Cy5, would be needed.