Both the N-terminal Loop and Wing W2 of the Forkhead Domain of Transcription Factor Foxo4 Are Important for DNA Binding*

FoxO4 belongs to the “O” subset of forkhead transcription factors, which participate in various cellular processes. The forkhead DNA binding domain (DBD) consists of three-helix bundle resting on a small antiparallel β-sheet from which two extended loops protrude and create two wing-like structures. The wing W2 of FoxO factors contains a 14-3-3 protein-binding motif that is phosphorylated by protein kinase B in response to insulin or growth factors. In this report, we investigated the role of the N-terminal loop (portion located upstream of first helix H1) and the C-terminal region (loop known as wing W2) of the forkhead domain of transcription factor FoxO4 in DNA binding. Although the deletion of either portion partly reduces the FoxO4-DBD binding to the DNA, the simultaneous deletion of both regions inhibits DNA binding significantly. Förster resonance energy transfer measurements and molecular dynamics simulations suggest that both studied N- and C-terminal regions of FoxO4-DBD directly interact with DNA. In the presence of the N-terminal loop the protein kinase B-induced phosphorylation of wing W2 by itself has negligible effect on DNA binding. On the other hand, in the absence of this loop the phosphorylation of wing W2 significantly inhibits the FoxO4-DBD binding to the DNA. The binding of the 14-3-3 protein efficiently reduces DNA-binding potential of phosphorylated FoxO4-DBD regardless of the presence of the N-terminal loop. Our results show that both N- and C-terminal regions of forkhead domain are important for stability of the FoxO4-DBD·DNA complex.

The winged helix or forkhead box (Fox) 2 transcription factors are a family of structurally related transcriptional activa-tors involved in embryogenesis, tumorigenesis, and the metabolism control (1)(2)(3)(4). Members of this family share a conserved 100-amino acid large DNA binding domain (DBD) containing three ␣-helices, three ␤-strands, and two wing-like loops (see Fig. 1A) (5)(6)(7)(8)(9)(10)(11). The highest degree of sequence conservation among DBD is found in three helices, H1, H2, and H3. Cocrystal structure of transcription factor HNF-3␥ showed 14 protein-DNA contacts distributed throughout the forkhead domain (5). The main DNA recognition site is ␣-helix H3, which makes contacts with the major groove of DNA. Other regions of forkhead domain that can make important interactions with DNA are both wings (5) or N-terminal extension upstream of helix H1 (12). Structural studies have shown that binding to the DNA causes only small structural changes in the DBD but induces significant bending of the target DNA (10,13). The molecular mechanism of Fox factors binding specificity for target DNA is still poorly understood. Because various forkhead proteins belonging to different subgroups share almost identical folds the variations in DBD topology cannot explain distinct DNA-binding specificities. Therefore other features of forkhead proteins, e.g. differences in charge distribution in the DNA binding interface, are likely to be responsible for DNA-binding specificities (9,14).
The DNA-binding potential of FoxO-DBD is controlled by multiple mechanisms. The most important factors seem to be the PKB-induced phosphorylation and the binding of the 14-3-3 protein (22)(23)(24)28). The PKB phosphorylation site and the 14-3-3 protein-binding motif are located in the basic region of wing W2 at the C terminus of FoxO-DBD (Fig. 1B). Structures of HNF-3␥ (FoxA3) and Genesis (FoxD3) complexes revealed that the cluster of basic residues in the wing W2 is involved in DNA binding (5,10). Moreover, the removal of wing W2 abolishes DNA binding of forkhead transcription factor HNF-3␣ (29). Sequence alignment between HNF-3␥, Genesis, and FoxO sequences suggests that analogue-basic residues forming the second FoxO PKB phosphorylation site and the 14-3-3 binding motif might participate in DNA binding as well (Fig. 1B). This similarity could explain the inhibitory effect of both the phosphorylation and the 14-3-3 protein on DNAbinding activity of FoxO proteins. However, the regulation of DNA binding among various FoxO proteins seems to differ significantly. It has been shown that the phosphorylation of the second PKB/14-3-3 binding motif in the wing W2 suppresses DNA binding of FoxO1 and FoxO6 factors (28,30). On the other hand, the PKB-induced phosphorylation of both DAF-16 (Caenorhabditis elegans FoxO homologue) and FoxO4 (fragment 11-213) does not by itself affect their binding to the target DNA (22,23). Binding of the 14-3-3 protein to phosphorylated FoxO4 and DAF-16 has been shown to be necessary for complete inhibition of their binding to the DNA.
To better understand these differences among FoxO-DBD, we investigated the role of N-terminal loop (portion located upstream of the first helix H1) and C-terminal region (loop known as wing W2) of forkhead domain of transcription factor FoxO4 in DNA binding. Although the deletion of either the Nor C-terminal portion of forkhead domain partly reduces the FoxO4-DBD binding to the DNA, the simultaneous deletion of both regions inhibits DNA binding significantly. Förster resonance energy transfer measurements and molecular dynamics simulations suggest that both N-and C-terminal regions of FoxO4-DBD studied directly interact with the DNA. In the presence of N-terminal part the PKB-induced phosphorylation of wing W2 by itself has negligible effect on DNA binding. On the other hand, in the absence of this loop the phosphorylation of wing W2 significantly inhibits the FoxO4-DBD binding to the DNA. The binding of the 14-3-3 protein efficiently reduces DNA-binding potential of phosphorylated FoxO4-DBD regardless of the presence of N-terminal region. Our results show that both N-and C-terminal regions of forkhead domain are important for the stability of FoxO4-DBD⅐DNA complex.

EXPERIMENTAL PROCEDURES
Expression, Purification, and Phosphorylation of FoxO4-DBD-Expression plasmid for FoxO4-DBD (FoxO4 82-207 ) was a kind gift of Dr. J. Weigelt. FoxO4-DBD mutants containing a single Cys residue at two different positions in the wing W2 (mutation Ser 183 3 Cys and Ser 193 3 Cys) were generated using the QuikChange kit (Stratagene). To generate C-terminally truncated FoxO4-DBD⌬C (FoxO4 82-183 ) a stop codon was introduced instead of Gly 184 . To create N-terminally truncated FoxO4-DBD⌬N the DNA-encoding sequence FoxO4 98 -199 was ligated into pET-15b (Novagen) using the NdeI site. To generate doubly truncated FoxO4-DBD⌬N⌬C, a stop codon was introduced into a DNA of FoxO4-DBD⌬N at position 184. All mutations were confirmed by sequencing. All versions and mutants of FoxO4-DBD have been expressed and purified as described previously (9). Briefly, FoxO4-DBD was expressed as 6ϫHis tag fusion proteins by isopropyl 1-thio-␤-D-galactopyranoside induction for 12 h at 30°C and purified from Escherichia coli BL21(DE3) cells using Chelating-Sepharose Fast Flow (Amersham Biosciences) according to the standard protocol. Eluted protein was dialyzed against buffer 50 mM sodium citrate (pH 6.3), 2 mM EDTA, 1 mM DTT and purified using cation-exchange chromatography on an SP-Sepharose column (Amersham Biosciences). The protein was eluted using linear gradient of NaCl (50 -1000 mM). Fractions containing FoxO4-DBD were dialyzed against buffer containing 20 mM Tris (pH 7.5), 50 mM NaCl, 1 mM EDTA, 2 mM DTT, and 10% (v/w) glycerol. Purified FoxO4-DBD was then phosphorylated by incubation with 9 units of PKB␣ (Upstate Biotechnology) per milligram of recombinant protein for 2 h at 30°C in the presence of 10 mM magnesium acetate and 0.2 mM ATP. After the phosphorylation, the FoxO4-DBD was re-purified using cation-exchange chromatography as mentioned above. Eluted proteins were finally dialyzed against buffer containing 20 mM Tris (pH 7.5), 50 mM NaCl, 1 mM EDTA, 2 mM DTT, and 10% (w/v) glycerol. The completeness of the phosphorylation reaction was checked using the MALDI-TOF mass spectrometry.
Mass Spectrometric Analysis of FoxO4-DBD-Samples were first separated by 15% SDS-PAGE, and excised protein bands were digested with Lys-C endoprotease (Promega, Madison, WI) directly in gel (31). Resulting peptide mixtures were extracted by 30% acetonitrile and 0.3% acetic acid and subjected to MALDI-TOF mass spectrometer BIFLEX (Bruker-Franzen, Bremen, Germany) equipped with a nitrogen laser (337 nm) and gridless delayed extraction ion source. Ion acceleration voltage was 19 kV, and the reflectron voltage was set to 20 kV. Spectra were calibrated externally using the monoisotopic [MϩH] ϩ ion of FoxO4 peptides with known sequence. A saturated solution of ␣-cyano-4-hydroxycinnamic acid in 50% MeCN/0.3% acetic acid was used as the MALDI matrix. 1 l of the sample was loaded on the target, and the droplet was allowed to dry at ambient temperature, overlaid with 1 l of matrix solution, and allowed to co-crystallize also at ambient temperature.
Labeling of FoxO4-DBD Mutants by 1,5-IAEDANS-The Forkhead domain of FoxO4 (fragment 82-207) does not contain any cysteine residue. To prepare FoxO4-DBD specifically labeled with fluorescence probe at two different position in the wing W2 we inserted a single Cys residue at positions 183 (mutation Ser 183 3 Cys) and 193 (mutation Ser 193 3 Cys). Covalent modification of FoxO4-DBD containing single cysteine residue with thiol-reactive probe 1,5-IAEDANS was carried out as described elsewhere (32). Briefly, the protein (50 -70 M) in 50 mM Tris (pH 7.5), 100 mM NaCl, 1 mM EDTA and label were mixed at a molar ratio of 1:40 and incubated at 30°C for 2 h, and then at 4°C overnight in the dark. The free unreacted label was removed by gel filtration in buffer containing 50 mM Tris (pH 7.5), 100 mM NaCl, and 1 mM EDTA. The incor-poration stoichiometry was determined by comparing the peak protein absorbance at 280 nm with the absorbance of bound 1.5-IAEDANS measured at 336 nm using an extinction coefficient of 5700 M Ϫ1 cm Ϫ1 (Molecular Probes, Eugene, OR).
Steady-state Fluorescence Anisotropy DNA Binding Measurements-Steady-state fluorescence measurements were performed on a PerkinElmer Life Sciences LS50B fluorescence spectrometer at 22°C with 100 nM dsDNA labeled with fluorescein at the 5Ј terminus of both strands (S1FLC/S2FLC). Increasing amounts of protein were titrated into the cuvette. At each FoxO4-DBD concentration, the steady-state fluorescence anisotropy of fluorescein was recorded (excitation at 494 nm and emission at 520 nm). Anisotropy was calculated from the fluorescence intensities according to the relationship, r ϭ (I II Ϫ I Ќ )/(I II ϩ 2I Ќ ). The fraction of DNA bound (F B ) was calculated from Equation 1, where Q represents the quantum yield ratio of the bound to the free form; it was estimated by calculating the ratio of the intensities of the bound to the free fluorophore. Parameter r max is the anisotropy at saturation, r obs is the observed anisotropy for any FoxO4-DBD concentration, and r min is the minimum observed anisotropy of the free DNA. F B was plotted against the FoxO4-DBD protein concentration and fitted using Equation 2 to determine the K D for FoxO4-DBD⅐dsDNA complex formation (34), where K D is the equilibrium dissociation constant, [P 1 ] is the dsDNA-fluorescein concentration, and [P 2 ] is the FoxO4-DBD concentration. Non-linear data fitting was performed using the Origin 6.0 package (MicroCal Software Inc.).
Time-resolved Fluorescence Measurements-Fluorescence resonance energy transfer was observed between the AEDANS moiety covalently attached to Cys 183 or Cys 193 of FoxO4-DBD and fluorescein (FLC) moiety attached to the 5Ј terminus of one of the dsDNA strands. Fluorescence intensity decays were measured on the laser-based time-correlated single photon counting apparatus with multichannel plate photomultiple detection described in detail previously (35). Fluorescence decays have been acquired under the "magic angle" conditions when the measured intensity decay I(t) is independent of a rotational diffusion of the chromophore and provides unbiased information about fluorescence lifetimes. The samples were placed in a thermostatic holder, and all experiments were performed at 15°C in buffer containing 50 mM Tris-HCl (pH 7.5), 100 mM NaCl, and 1 mM EDTA. FoxO4-DBD concentration was 2.4 M, and dsDNA concentration was 6 M. Dansyl fluorescence was excited and collected at 315 and 370 nm, respectively. Fluorescence decays were processed using the singular value decomposition maximum entropy method, and mean lifetimes were calculated as described previously (35). The averaged efficiency of energy transfer E was calculated from the mean donor lifetime in the presence ( DA ) and in the absence of acceptor ( D ) in Equation 3.
The average distance between the donor-acceptor pair R was calculated from Equation 4, where R 0 is the critical Förster distance (the distance at which the energy transfer occurs with 50% efficiency). An R 0 value of 49 Å determined for AEDANS and fluorescein pair was used to calculate R (36,37).
Molecular Simulations-The MD simulations were performed using the GROMACS package 3.3 with the AMBER-94 parameter set exported to the GROMACS format (38). To prepare the starting structure the solution structure of FoxO4-DBD (sequence 92-181, PDB access code 1E17) was fitted onto the co-crystal structure of transcription factor HNF-3␥ with bound DNA (5,9). The missing wing W2 of FoxO4-DBD was modeled (up to residue 193) using the structures of transcription factors HNF-3 and Genesis as templates (5,10). Coordinates of dsDNA (S1/S2) containing the consensus FoxO binding sequence (5Ј-GCGTTGTTTACGC-3Ј) were generated using the FIBER program from the 3DNA package (39). The molecule of dsDNA was oriented in such a way that the motif 5Ј-TGTTT-3Ј was superimposed with the motifs 5Ј-TATTT-3Ј from the solution structure of Genesis⅐DNA complex and 5Ј-GACTT-3Ј from the crystal structure of HNF-3␥⅐DNA complex. FoxO4-DBD⅐DNA complex was then solvated with water in a periodic box using the TIP3P water model (40). All water molecules within 1.5 Å of any protein or DNA atom were removed. The system was electroneutralized by replacing water molecules with the highest electrostatic potential energy with Na ϩ ions. Next, the resulting system was energy-minimized using the steepest descent method. This was followed by a 50-ps equilibration MD run during which position restraints were applied on the complex. The system was then set completely free, and an MD simulation 40 ns long at 300 K temperature and 1 bar pressure was performed. Simulation of FoxO4-DBD alone (20 ns long) was performed using the same protocol. Both simulations were run with a 2-fs time step. Solute, solvent, and counterions were independently weakly coupled to a reference temperature bath (41) with a coupling constant of T ϭ 0.1 ps. The initial velocities were obtained from a Maxwellian distribution at the desired temperature. The pressure was controlled using the Berendsen algorithm with a coupling constant of P ϭ 0.5 ps. All bond lengths were constrained using LINCS (42). The particle mesh Ewald method was used to treat electrostatic interactions (43). The cut-off for van der Waals interactions was set to 9 Å. Trajectories were analyzed using the GROMACS software package (44). Cluster analysis was performed using the GROMOS method (45).

RESULTS
Characterization of Phosphorylated FoxO4-DBD-In this report, we studied a human FoxO4-DBD (sequence 82-207) containing forkhead domain and one PKB phosphorylation site Ser 193 located in the wing W2 (Fig. 1). FoxO4-DBD protein and its N-terminally truncated version FoxO4-DBD⌬N were phosphorylated by PKB, and the result of phosphorylation reaction was determined using MALDI-TOF mass Black circles indicate basic residues from the wing W2 that interact with DNA (5,10). Arrows indicate serine residues that were mutated to cysteines for the attachment of AEDANS moiety. Underlined FoxO4 sequence denotes 14-3-3 protein-binding motif. Stars indicate residues predicted by MD simulation to make contact with DNA.
spectrometry. Positive ion mass spectra were measured in the reflection mode to check the amino acid sequences of unphosphorylated and phosphorylated FoxO4-DBD tryptic peptides. The comparison of positive MALDI-TOF mass spectra of digested unphosphorylated and phosphorylated proteins clearly demonstrates the phosphorylated peptide of FoxO4-DBD. The detected peaks in positive ion mass spectra of phosphorylated pFoxO4-DBD having the protonized mass of 1599.6 (m/z) correspond to phosphorylated peptide APRRRAASMD-SSSK (the PKB phosphorylation/14-3-3 protein-binding motif is 189 RRAAS 193 ) and the same peptide with oxidized methionine in its sequence having the mass of 1615.7 (m/z). On the other hand, the unphosphorylated protein mass spectrum provided no peaks with the same values of m/z there. The identity and structure of both phosphorylated peptides (with or without methionine oxidation) was further corroborated by analysis of their post source decay spectra to authenticate Ser 193 as the phosphorylated amino acid (data not shown).
Effects of Truncations and Phosphorylation on DNA Binding Affinity of FoxO4-DBD-To assess the role of both N-terminal extension and C-terminal wing W2 in FoxO4-DBD binding to the DNA, we compared a DNA binding affinity of intact FoxO4-DBD and both N-and C-terminally truncated versions using the fluorescence anisotropy-based binding assay with a fluorescein-tagged dsDNA (Fig. 2). Intact FoxO4-DBD (sequence 82-207) binds to dsDNA with K D of 188 Ϯ 10 nM. The removal of wing W2 at the C terminus of forkhead domain caused a 2-fold increase in K D to 355 Ϯ 27 nM. A similar effect was also observed in the case of N-terminal truncation, where the removal of residues located upstream of helix H1 lead to a 2-fold increase in K D to 399 Ϯ 13 nM. Simultaneous deletion of both terminal portions, however, profoundly inhibited the DNA-binding potential of FoxO4-DBD (K D ϳ 1638 Ϯ 57 nM). Next, we studied the effect of PKB-induced phosphorylation of wing W2 at Ser 193 . Phosphorylation of intact FoxO4-DBD induced only a minor change of DNA binding affinity (K D ϭ 232 Ϯ 13 nM), whereas the phosphorylation of N-terminally truncated FoxO4-DBD⌬N caused significant inhibition of DNA binding affinity (K D ϳ 1430 Ϯ 62 nM). These data indicate that both N-and C-terminal regions participate in DNA binding and probably stabilize the protein⅐DNA complex. In addition, it seems that phosphorylation at Ser 193 by itself has negligible influence on DNA-binding properties of intact FoxO4-DBD, whereas phosphorylation of FoxO4-DBD⌬N reduces the DNA binding affinity to the same extent as observed in the case of doubly truncated FoxO4-DBD⌬N⌬C.
14-3-3 Protein Binding to Wing W2 Inhibits DNA Binding Potential of FoxO4-DBD-Phosphorylation of FoxO proteins by PKB creates two 14-3-3 recognition motifs and induces the binding of 14-3-3 proteins to FoxO factors (18,23,24,26). The exact role of the 14-3-3 protein in the regulation of FoxO transcription factors activity is still not fully understood. Interactions between 14-3-3 proteins and phosphorylated FoxO4 (fragment 11-213) or phosphorylated DAF-16 have been shown to result in inhibition of DNA binding (22,23). The 14-3-3 protein recognition sites border the forkhead DNA binding domain; one is located at the N terminus and a second one at the wing W2 of FoxO-DBD. Because the wing W2 of FoxO-DBD might participate in DNA binding, it has been speculated that the interaction between the 14-3-3 protein and the wing W2 contributes to the inhibition of FoxO DNA-binding properties (24). To test whether this is also true in the case of FoxO4 we investigated the effect of the 14-3-3 protein binding on DNA-binding properties of intact FoxO4-DBD. First, the association of 14-3-3 protein (human isoform zeta) with phosphorylated FoxO4-DBD has been checked using native Tris borate-EDTA (TBE) buffer PAGE electrophoresis. As can be seen from Fig. 3A, the phosphorylated pFoxO4-DBD forms a stable complex with the 14-3-3 protein compared with unphosphorylated FoxO4-DBD (e.g. compare lanes 4 -6 and 7-9). Formation of the stable 14-3-3 protein⅐pFoxO4-DBD complex was also confirmed using analytical gel filtration (data not shown). Next, we studied the effect of 14-3-3 protein binding on DNAbinding properties of FoxO4-DBD using fluorescence anisotropy-based binding assay (Fig. 3B). Titration data show that 14-3-3 protein can disrupt DNA complex of phosphorylated pFoxO4-DBD more efficiently compared with unphosphorylated FoxO4-DBD. The inhibitory effect of 14-3-3 protein, caused presumably by pFoxO4-DBD binding to the 14-3-3 protein, is consistent with similar observations made for C. elegans FoxO homologue DAF-16 and dpFoxO4   (22,23). Partial dissociation of unphosphorylated FoxO4-DBD⅐DNA complex is probably caused by weak interaction between the 14-3-3 protein and FoxO4-DBD (Fig. 3A, lanes 5 and 6).
Molecular Dynamics Simulations of FoxO4-DBD⅐DNA Complex-Structures of HNF-3␥⅐DNA and Genesis⅐DNA complexes demonstrated that the wing W2 at the C terminus of forkhead domains of these transcription factors makes important contacts with DNA and helps stabilize the protein⅐DNA complex (5, 10). On the other hand, N-terminal extensions of these two forkhead domains do not interact with DNA as has been observed for forkhead transcription factors E2F4 and DP2 (12). Because no experimental structure of FoxO⅐DNA complex is available, it is unclear whether the same holds true for FoxO factors. Therefore, we performed molecular dynamics simulations of both free FoxO4-DBD (20 ns long) and its DNA-bound form (40 ns long) to both characterize conformational behavior of N-and C-terminal regions of FoxO4 forkhead domain and to generate a realistic model of FoxO4-DBD⅐DNA complex. The starting structure for MD simulations was prepared using the solution structure of FoxO4-DBD, and both the missing wing W2 and the position of canonical B-form DNA duplex were modeled using the HNF-3␥⅐DNA and Genesis⅐DNA structures as templates (5,9,10) , and 400 pmol). B, fluorescence anisotropy binding assay. 14-3-3 protein binding to phosphorylated pFoxO4-DBD inhibits DNA binding more efficiently compared with unphosphorylated FoxO4-DBD. Fluorescence anisotropy of dsDNA probe labeled with fluorescein was used to follow FoxO4-DBD⅐DNA complex dissociation. Concentration of FoxO4-DBD and dsDNA was 1400 nM and 100 nM, respectively. superimposition of these representative conformations with solution structure of FoxO4-DBD is shown in Fig. 5A (Fig. 5, B and C). All these residues are highly conserved among FoxO transcription factors (Fig. 1A). Basic residues from wing W2 (Arg 188 , Arg 189 , and Arg 190 ) make contacts with phosphate groups of DNA backbone (Thr-17, Ala-18, and Ala-19 of strand S2). HNF-3␥ and Genesis transcription factors make very similar contacts to the DNA, with exception of the N-terminal loop upstream of ␣-helix H1 (5,10). This region, however, was found to interact with DNA in co-crystal structures of winged helix proteins E2F4 and DP2 (12). Simulation also suggests that FoxO4-DBD deforms DNA molecule by inducing a 19°b end.
Investigation of the Wing W2 Location Using FRET-FRET measurements between AEDANS moiety attached at two different positions within the wing W2 and a fluorescein moiety attached at 5Ј-end of either the sense or the antisense strand of dsDNA were used to investigate interaction between the wing W2 at the C terminus of forkhead domain and DNA. To specifically label the wing W2 for FRET measurements, we have constructed two FoxO4-DBD mutants containing a single Cys residue located at two different positions within the wing W2: Cys 183 (this mutant was named FoxO4-DBDc183) and Cys 193 (FoxO4-DBDc193). These two sites border the cluster of positively charged residues (Arg 188 -190 ) that are potentially involved in DNA binding (Fig. 1B). The Cys residues were selectively labeled with the extrinsic fluorophore 1,5-IAEDANS as energy donor. Labeling experiments revealed that both FoxO4-DBD mutants can be completely modified by 1,5-IAEDANS. The stoichiometries of AEDANS/mol of these two FoxO4-DBD mutants were found to be ϳ1. The DNA binding abilities of both unlabeled and labeled FoxO4-DBD mutants were checked using the electrophoretic mobility shift assay, and no significant differences were found (data not shown).
The presence of the energy transfer between AEDANS moiety attached to a cysteine residue of FoxO4-DBD and fluorescein moiety bound to the 5Ј-end of dsDNA was tested using the steady-state fluorescence. Fluorescence emission spectra of AEDANS-labeled FoxO4-DBD in the absence (both strands of dsDNA were unlabeled) and presence of fluorescein-labeled dsDNA (either the sense or antisense strand of dsDNA was labeled) are shown in Fig. 6. When excited at 336 nm, the AEDANS group has a fluorescence emission maximum near 488 nm. In the presence of fluorescein-labeled dsDNA the donor steady-state fluorescence intensities of both FoxO4-DBD mutants were quenched by ϳ30 -40% compared with proteins with bound unlabeled dsDNA, indicating the presence of FRET. The sensitized fluorescence emission of fluorescein (induced by the nonradiative energy transfer from AEDANS group) was observed as a peak at 521 nm.
Time-resolved fluorescence intensities of AEDANS-labeled FoxO4-DBD mutants exhibited complex fluorescence decays. The time-resolved emission data were analyzed by the singular value decomposition maximum entropy method and mean excited-state lifetimes ( mean ) were calculated as described previously (35). Lifetimes of AEDANS in the absence and in the presence of acceptor as well as efficiencies of the energy transfer are presented in Table 1. In the presence of fluorescein-labeled dsDNA for both FoxO4-DBD mutants the mean of donor was reduced as a result of the energy transfer (Fig. 7). Values of the FRET efficiency were used to calculate the average distances between the donor-acceptor pair. Distances between AEDANS moiety attached to Cys 183 or Cys 193 of FoxO4-DB and fluorescein attached to the 5Ј-end of either sense or antisense strand of dsDNA were calculated to be in range from 56.4 to 62.3 Å (Table 1).
To compare results of molecular dynamics simulation with FRET measurements the time evolution of distances between atoms corresponding to AEDANS and fluorescein attachment sites were calculated (Fig. 8A). Distances between residue Ser 183 and 5Ј-ends of both strands were relatively constant throughout the production run with average distance of 33.5 Ϯ 1.5 Å for pair Ser 183 -G1 (strand S1) and 36.5 Ϯ 2.9 Å for Ser 183 -G14 (strand S2). On the other hand, distances between residue Ser 193 and 5Ј-ends of both strands substantially fluctuated with average distance of 39.2 Ϯ 5.4 Å for pair Ser 193 -G1 (strand S1) and 22 Ϯ 6.3 Å for Ser 193 -G14 (strand S2). These fluctuations can be explained by the high flexibility of the C terminus of wing W2 in agreement with high positional fluctuations of its C␣ atoms (Fig. 6B). If we consider both the most probable position of the fluorescein moiety at the 5Ј-end of DNA duplex (46,47) and the size of dansyl moiety, the three of four calculated distances are consistent with our FRET measurements ( Fig. 8B and Table 1). The fourth distance between AEDANS attached at residue 193 and fluorescein at 5Ј-end of S2 strand is shorter by ϳ8 Å compared with the distance determined by FRET.

DISCUSSION
The DNA binding domain of forkhead box transcription factors is a compact ␣/␤ structure containing a classic helix-turnhelix DNA binding motif within a three-helix bundle, braced against an antiparallel ␤-sheet (4,5,(7)(8)(9)(10)(11). In addition, an extended loop between ␤-strands and the loop C-terminal to the ␤-sheet create two wing-like structures (Fig. 1A). The main DNA binding portion of forkhead box DBD is the H3 recognition helix. Other regions of forkhead domain that can make important interactions with DNA are wings (5) or N-terminal extension upstream of helix H1 (12). The likely function of these regions is the stabilization of protein⅐DNA complex.
Interaction of N-terminal Loop and Wing W2 of FoxO4-DBD with DNA-In the FoxO subset of Fox proteins the wing W2 contains a 14-3-3 protein-binding motif that is phosphorylated by PKB in response to insulin or growth factors (18,23,24,26). Both PKB-induced phosphorylation and the 14-3-3 protein binding have been suggested to reduce the DNA-binding potential of FoxO factors, and the alteration of wing W2 structure can be one of the main reasons of this inhibition (22-24, 28, 30). Structures of forkhead transcription factors HNF-3␥ and Genesis with bound DNA revealed that wing W2 participates in DNA binding through a cluster of basic residues (5, 10). If the same is true for FoxO4-DBD then it could explain the inhibitory effect of the 14-3-3 protein on its DNA-binding properties. The cluster of basic residues at the C-terminal end of wing W2 is an integral part of 14-3-3 protein-binding motif, and upon the 14-3-3 protein binding it would be buried within the 14-3-3 protein ligand binding groove and unable to interact with DNA (48). To test whether the wing W2 of FoxO4-DBD is directly involved in DNA binding, we first prepared a C-terminally truncated version of FoxO4-DBD missing the wing W2. The truncation of wing W2 caused only a 2-fold decrease in DNA binding affinity (Fig. 2). This result contrasts with the observation that removal of basic residues from wing W2 of closely related FoxO1-DBD completely inhibits its binding to the DNA (28). However, the construct used in our experiments contains the N-terminal loop upstream of helix H1 (residues Arg 82 -Ala 96 in the FoxO4 sequence), whereas FoxO1-DBD used by Zhang et al. (28) does not. To create a comparable construct of FoxO4-DBD we removed this N-terminal extension and observed a decrease in DNA binding affinity of ϳ2-fold compared with intact DBD. Upon the additional removal of wing W2 the DNA-binding potential of FoxO4-DBD⌬N⌬C was significantly inhibited as in the case of C-terminally truncated FoxO1-DBD used by Zhang et al. (28). Therefore, it seems that both N-terminal extension and C-terminal wing W2 are important for the stability of FoxO-DBD⅐DNA complex.
Because no experimental structure of FoxO-DBD with bound DNA is available we used molecular dynamics simulations and FRET measurements to investigate the conformation of N-terminal loop and wing W2 of FoxO4-DBD and their interactions with DNA. Our results suggest that both regions make contacts with DNA in agreement with the binding experiments. MD simulations predict that three residues from the N-terminal loop (Arg 93 , Arg 94 , and Trp 97 ) interact with DNA (Fig. 5C). Similar contacts between positively charged residues from this region DBD and DNA have been observed in cocrystal structure of forkhead transcription factors E2F4 and DP2 (12). MD simulations together with FRET measurements indicate that basic residues within the wing W2 (Arg 188 and Arg 189 ) interact with DNA in a similar fashion as shown for transcription factors HNF-3␥ and Genesis (Figs. 5B, 5C, 8A, and 8B) (5,10).
Effect of Phosphorylation and 14-3-3 Protein Binding-Phosphorylation of the second PKB/14-3-3 binding motif at the wing W2 of DBD has been shown to suppress DNA binding of FoxO1 and FoxO6 factors (28,30). On the other hand, PKBmediated phosphorylation of both DAF-16 (C. elegans FoxO homologue) and FoxO4 (fragment 11-213) does not by itself affect their DNA binding abilities, and the association with the 14-3-3 protein has been shown to be necessary for complete inhibition of DNA binding (22,23). Our binding experiments revealed that PKB-induced phosphorylation of Ser 193 has a negligible effect on DNA binding affinity of intact FoxO4-DBD absence (open squares, unlabeled dsDNA S1/S2 was used) and in the presence (triangles and open circles) of fluoresceinlabeled dsDNA (S1FLC/S2 or S1/S2FLC). The calibration is 73 ps/channel. (Fig. 2) in agreement with our previous observations that PKBinduced phosphorylation had no significant effect on DNA binding of FoxO4   (23,27). A possible explanation for the observed differences between FoxO1-DBD and FoxO4-DBD might be again in the absence of the N-terminal loop in FoxO1-DBD construct used by Zhang et al. (28). When we phosphorylated N-terminally truncated FoxO4-DBD⌬N (comparable to the FoxO1-DBD construct used in Ref. 28), we have observed strong inhibition of DNA-binding potential (Fig. 2).
Our results indicate that the introduction of negative charge by itself cannot be the only factor inhibiting the DNA-binding potential of intact FoxO4-DBD. Phosphorylation of Ser 193 might inhibit DNA binding of FoxO4 indirectly through the binding of 14-3-3 protein. Titration data show that the 14-3-3 protein binding to phosphorylated wing W2 of FoxO4-DBD can efficiently dissociate the FoxO4-DBD⅐DNA complex, whereas in the absence of phosphorylation the inhibitory effect of 14-3-3 protein is significantly weaker (Fig. 3B). Because the removal of wing W2 does not have as strong an inhibitory effect on FoxO4-DBD binding to the DNA, the inhibition caused by 14-3-3 protein likely arises not only from the interaction between 14-3-3 and wing W2 but potentially also from interactions with other region(s) of the forkhead domain.
In conclusion, our results show that both N-terminal loop located upstream of helix H1 and C-terminal wing W2 are important for FoxO4-DBD binding to the DNA. Although the removal of either region results in partial reduction of DNA binding, the simultaneous deletion of both portions inhibits DNA binding significantly. The PKB-induced phosphorylation of wing W2 by itself has negligible effect on DNA binding, whereas the complex formation between 14-3-3 protein and phosphorylated FoxO4-DBD inhibits DNA binding efficiently.