Three-Dimensional Domain Swapping Changes the Folding Mechanism of the Forkhead Domain of FoxP1

Abstract

The forkhead family of transcription factors (Fox) controls gene transcription during key processes such as regulation of metabolism, embryogenesis, and immunity. Structurally, Fox proteins feature a conserved DNA-binding domain known as forkhead. Interestingly, solved forkhead structures of members from the P subfamily (FoxP) show that they can oligomerize by three-dimensional domain swapping, whereby structural elements are exchanged between adjacent subunits, leading to an intertwined dimer. Recent evidence has largely stressed the biological relevance of domain swapping in FoxP, as several disease-causing mutations have been related to impairment of this process. Here, we explore the equilibrium folding and binding mechanism of the forkhead domain of wild-type FoxP1, and of two mutants that hinder DNA-binding (R53H) and domain swapping (A39P), using size-exclusion chromatography, circular dichroism, and hydrogen-deuterium exchange mass spectrometry. Our results show that domain swapping of FoxP1 occurs at micromolar protein concentrations within hours of incubation and is energetically favored, in contrast to classical domain-swapping proteins. Also, DNA-binding mutations do not significantly affect domain swapping. Remarkably, equilibrium unfolding of dimeric FoxP1 follows a three-state N₂ ↔ 2I ↔ 2U folding mechanism in which dimer dissociation into a monomeric intermediate precedes protein unfolding, in contrast to the typical two-state model described for most domain-swapping proteins, whereas the A39P mutant follows a two-state N ↔ U folding mechanism consistent with the second transition observed for dimeric FoxP1. Also, the free-energy change of the N ↔ U in A39P FoxP1 is ∼2 kcal⋅mol⁻¹ larger than the I ↔ U transition of both wild-type and R53H FoxP1. Finally, hydrogen-deuterium exchange mass spectrometry reveals that the intermediate strongly resembles the native state. Our results suggest that domain swapping in FoxP1 is at least partially linked to monomer folding stability and follows an unusual three-state folding mechanism, which might proceed via transient structural changes rather than requiring complete protein unfolding as do most domain-swapping proteins.

Introduction

The forkhead box family of transcription factors (Fox) is a large group of proteins that possess a structurally conserved DNA-binding domain of ∼100 residues known as the winged helix or forkhead domain (1, 2). The structure of the forkhead domain consists of a compact packing of three α-helices (H), three β-strands (S), and two loops or wings (W), arranged in the order of H1-S1-H2-H3-S2-W1-S3-W2 (3) (Fig. 1 A). From these structural elements, helix H3 and W2 interact with the major and minor grooves of DNA, respectively (Fig. 1) (2).

Figure 1.

Solution structure of the forkhead domain of FoxP1 and its domain-swapped dimer. (A) Cartoon representation of the solution structure of the monomeric mutant A39P/C61Y of human FoxP1 (PDB: 2KIU). The side chains of residues A39 (here substituted by proline) and R53 are shown as red sticks. The coloring of the secondary-structure elements follows the topology scheme indicated on the bottom, where // represents the hinge region that connects the swapped elements to the rest of the protein and also where A39 is located. (B) Comparison of the isolated monomer of FoxP1 and a homology model of the domain-swapped dimer of FoxP1, generated using the structure of the dimer of FoxP2 bound to DNA (PDB: 2A07) with the software MODELLER (50), with its polypeptide chains colored green and orange. The dashed line helps to visualize that the forkhead fold of the isolated subunit is maintained in the domain-swapped dimer, but is composed by structural elements from two polypeptide chains. The close-up on the right side denotes that the helix H2 has two more helical turns in the dimer (cyan) than in the monomeric form (red), a rearrangement that allows the exchange of the structural elements H3, S2, W1, S3, and W2 with an adjacent subunit. (C) Two different views, rotated 180°, of the structure of the domain-swapped dimer of FoxP1 bound to two DNA strands, generated using the structure of the domain-swapped dimer of FoxP2 as a template. The DNA structures are shown as spheres and the protein is shown in cartoon representation using the same color scheme as in (B). Rotation of the protein-DNA structure allows visualization of the position of hinge residue A39 (red spheres), which allows domain swapping, and helix H3 residue R53 (yellow spheres), which interacts with DNA. The structure representations were generated using the software VMD (51). To see this figure in color, go online.

These transcription factors constitute an ancient protein family that arose in early metazoans (4), with Fox-encoding genes arranging in clusters in basal bilaterians (5) while at the same time quickly expanding to acquire diverse specialized functions (6). In humans, the Fox family consists of 50 members (7) that control gene transcription during important processes such as regulation of metabolism (8), embryogenesis (9), organogenesis (10), immunity (11), and speech and language development (12). Moreover, deregulation of Fox proteins can promote the development and progression of cancer, which makes them potential targets for cancer monitoring and therapeutic treatment (13).

Although most forkhead domains from Fox proteins have been characterized as monomers in solution, the four members of the FoxP subfamily have been described to form dimers (14, 15, 16) via an unusual oligomerization mechanism known as three-dimensional domain swapping (17). This mechanism is characterized by an exchange of identical structural elements between two adjacent partners through the recruitment of interactions that stabilize the monomeric state. To enable this fundamental feature, the interactions between the swapping elements and the rest of the protein must be broken in one monomer and replaced with similar interactions, but in an intermolecular fashion (18), thus implying that protein unfolding is necessary for domain swapping to occur.

Experimentally, several lines of evidence obtained using a variety of domain-swapping proteins have largely stressed the requirement of protein unfolding to reach the intertwined dimer. For p13suc1, a sample of monomer at 0.5 mM under native conditions took several months to equilibrate with its domain-swapped form, but reached monomer-dimer equilibrium rapidly if the protein was unfolded and refolded using the same protein concentration (19). Other extreme cases, such as cystatin, were reported to have a predicted half-time of ∼2000 years for the dimerization process in the absence of denaturant (20). For the HIV-inactivating protein cyanovirin-N, a determination of the energetic parameters for the interconversion between its monomeric and dimeric forms in kinetic experiments confirmed that the activation barrier for domain swapping is similar in magnitude to the unfolding barrier, thus confirming the need for overall unfolding (21). Also, equilibrium unfolding of mutants of p13suc1 and cyanovirin-N that predominantly exist as dimers at micromolar concentrations follows a two-state unfolding mechanism just as observed for their monomeric counterparts, which can be explained by maintenance of the same folding nucleus in both native states (22, 23). Therefore, domain swapping in proteins can be conceived as an alternative route along the folding pathway, where the branching point corresponds to the folding transition state (24). Although intermediate states have been detected during kinetic refolding experiments in some domain-swapping proteins, their formation appears to be detrimental for the domain-swapped form (25).

Domain swapping is not only an intriguing and plausible mechanism that could explain how proteins may have evolved to form oligomers based on the same principles that define protein folding (17, 18, 26), it also has a role in regulating protein function (27). For several proteins, such as RNase (28), the restriction endonuclease SgrAI (29), and the RelB member of the NF-κB family of transcription factors (30), domain swapping modulates their function. Moreover, domain swapping is also critical for fibrillogenesis in disease-causing proteins such as cystatin C (31), γ-crystallins (32), and PrP (33).

In the case of the forkhead domain of FoxP proteins, the exchanging elements in the domain-swapped dimer correspond to helix H3, strands S2 and S3, and wing W2, which folds into a helix in FoxP proteins (14) (Fig. 1 B). Interestingly, isolated monomer and domain-swapped dimeric species of FoxP2 were shown to reequilibrate as monomer/domain-swapped dimer mixtures after 24 h at room temperature or 10 h at 37°C, in dramatic contrast to the variety of domain-swapping proteins studied to date (14). The DNA-binding sites of the domain-swapped dimer are at opposite ends of the dimer structure, suggesting that FoxP dimers may bind to nonadjacent DNA sites from single or multiple DNA strands (Fig. 1 C) and participate in the regulation of higher-order protein-DNA complexes (14) through the establishment of long-range chromosomal interactions (34). Moreover, domain swapping in Fox proteins seems to be a unique adaptive structural feature of the P subfamily, caused by the replacement of a proline residue from an FPYF motif located in the loop connecting helices H2 and H3, which is conserved in all other Fox members, by alanine (Fig. 1) (14). This change allows the extension of the length of helix H2 and the consequent displacement of helix H3 (Fig. 1 B) (15).

Disruption of the biological function of transcription factors is often related to mutations that abolish DNA binding. In the case of FoxP2, mutation of the DNA-contacting residue R553 to H in the forkhead domain helix H3 (14) is linked to a severe congenital speech disorder (12). Domain swapping appears also to be functionally relevant, since various mutations associated with impairment of this mechanism lead to several transcriptional deregulations and diseases. In FoxP3, a triple mutant that impedes domain swapping,W348Q/M370T/A372P, leads to a reduced repression and activation of FoxP3-regulated genes and abolishes their suppressor activity in regulatory T cells (15). Moreover, mutations R347H and F373A, which are linked to a severe autoimmune syndrome known as IPEX (immune dysregulation, polyendocrinopathy, enteropathy, X-linked), have been described to induce their phenotypes by hindering the ability of FoxP3 to dimerize (15).

The key role of the monomer-dimer transition in the function of FoxP proteins prompted us to perform a detailed analysis of the molecular mechanism underlying their folding and binding. In this work, we biophysically characterized the folding and dimerization mechanism of the forkhead domain of wild-type FoxP1 and two mutants, A39P and R53H, which disrupt the ability to form domain-swapped dimers and bind to DNA, respectively. We first characterized the temperature and concentration dependence of the monomer-dimer equilibrium, showing that dimerization occurs at micromolar protein concentrations within a few hours of incubation at 37°C in the absence of denaturants and is strongly favored energetically for both wild-type FoxP1 and R53H, whereas all dimerization is abolished for A39P. Then, we performed equilibrium unfolding experiments at different protein concentrations, showing that dimeric FoxP1 follows a three-state folding mechanism with a monomeric intermediate (N₂ ↔ 2I ↔ 2U), whereas the A39P mutant is better described by a two-state folding mechanism (N ↔ U). A comparison of the free energy of unfolding of the monomeric intermediate of wild-type FoxP1 and R53H with the unfolding of the monomeric A39P mutant shows that the latter is more stable by ∼2.0 kcal⋅mol⁻¹. Finally, we employed hydrogen-deuterium exchange mass spectrometry (HDXMS), which probes the solvent accessibility of several protein regions, to conduct a local structural characterization of wild-type FoxP1 and the A39P mutant under native conditions and under conditions that favor the intermediate state, providing strong evidence of the native-like features of the latter. Our results suggest that, in contrast to other domain-swapping proteins, the forkhead domain of FoxP1 follows an unusually fast dimerization mechanism, where domain swapping occurs starting from largely folded ensembles and can be partially attributed to reduced monomer folding stability and transient structural changes.

Materials and Methods

Protein expression and purification

Codon-optimized DNA sequences encoding the forkhead domain of human FoxP1 (GenScript, Piscataway, NJ) and its mutants were cloned into a modified pET-28a vector containing a His₆-tag, a TEV cleavage site and an S-tag sequence in the 5′ end of the gene. Amino acid residues were numbered according to the sequence numbering in the deposited structure of the forkhead domain of FoxP1 (PDB: 2KIU). Plasmids containing the DNA sequence of FoxP1 were used to obtain the A39P and R53H mutants by PCR mutagenesis using the QuikChange Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA). Protein overexpression was induced in Escherichia coli BL21(DE3) with 0.5 mM of isopropyl-β-D-thiogalactoside at an optical density (measured at 600 nm) of ∼0.9. Proteins were purified as described in Stroud et al. (14). The pooled proteins were cleaved with TEV protease to remove the His₆-tag and then placed again into Ni²⁺ affinity columns to separate the cleaved S-tag fused protein from both TEV protease and uncleaved protein. The proteins were stored in 2 M guanidinium chloride (Gnd·HCl) and 20% glycerol at 4°C. The proteins were then refolded by dialysis against standard buffer (20 mM NaH₂PO₄ pH 5.5, 50 mM NaCl, and 25 mM MgCl₂) and subjected to size-exclusion chromatography (SEC) before experiments were performed.

Equilibrium unfolding and refolding experiments

Unfolding and refolding experiments were performed in either standard buffer or CD buffer (20 mM NaH₂PO₄ pH 5.5, 50 mM Na₂SO₄ and 25 mM MgSO₄) supplemented with Gnd·HCl (molecular biology grade; Thermo Fischer Scientific, Waltham, MA) at concentrations ranging from 0 to 6.5 M. For all unfolding experiments, the proteins were exposed to different Gnd·HCl concentrations for at least 24 h at 25°C using a water bath. Refolding was obtained by incubating the protein first in 6 M Gnd·HCl for 2 h at 37°C and then diluting the unfolded protein in buffer with several Gnd·HCl concentrations. Unfolding curves were obtained by using 3–43 μM of protein in terms of monomer concentration. All curves were generated in triplicates.

SEC

SEC experiments were performed on a Water Breeze HPLC system (Waters, Milford, MA) using a Superdex 75 10/30 column (GE Healthcare Biosciences, Pittsburgh, PA) as described by Chu et al. (16), at several different temperatures (17–37°C) for both wild-type and mutant proteins. The column was equilibrated with 60 mL of mobile phase (standard buffer) and calibrated using the Bio-Rad gel filtration standard (Bio-Rad Laboratories, Hercules, CA), which contained the following components: vitamin B-12, 1.35 kDa, 8.5 Å Stokes radius (Rs); horse myoglobin, 17 kDa, 19 Å Rs; and chicken ovalbumin, 44 kDa, 30.5 Å Rs. To determine the effect of Gnd·HCl on the hydrodynamic properties of FoxP1, the column was equilibrated with the mobile phase supplemented with the same Gnd·HCl concentration as that of the samples to be injected. The elution volume and Rs of wild-type and mutant proteins in all conditions were estimated, taking into account the Gnd·HCl-dependent variation of the column’s void volume and total solvent-accessible volume, using Blue dextran and dithiothreitol as elution controls, respectively.

Circular dichroism spectroscopy measurements

Far-UV determinations were made in a Jasco J-1500 spectropolarimeter, employing cells with path lengths of either 5 or 1 mm depending on the protein concentration being used. Samples were incubated in circular dichroism (CD) buffer before experiments were conducted. The final spectrums resulted from an accumulation of four scans (scan rate of 100 nm⋅min⁻¹) between 200 and 250 nm. The CD signal at 223 nm was used to follow protein unfolding and refolding processes.

HDXMS

HDXMS was performed using a Synapt G2Si system with H/DX technology (Waters) as previously described (35). Briefly, 5 μL of protein solution at an initial concentration of 11 μM was allowed to exchange at 25°C for 0–5 min in 55 μL of deuterated standard buffer alone, for native-state conditions, or supplemented with 2 M of Gnd·HCl to characterize the domain-swapping intermediate state. Then, reactions were quenched for 2 min at 1°C using an equal volume of a solution containing 2 M Gnd·HCl, 1% formic acid, pH 2.66. The quenched samples were injected onto a custom-built column containing pepsin-agarose (Thermo Fischer Scientific, Rockford, IL) and the resulting peptic peptides were separated by analytical chromatography at 1°C within a mobile phase containing 0.1% formic acid using a gradient of 7–95% acetonitrile in 7 min. The analytes were electrosprayed into a Synapt G2-Si quadrupole time-of-flight (TOF) mass spectrometer (Waters). The mass spectrometer was set to MS^E-ESI+ mode for initial peptide identification and to Mobility-TOF-ESI+ mode to collect H/DX data. The mass range was set to 200–2000 (m/z), scanning every 0.4 s. Infusion and scanning of leu-enkephalin (m/z = 556.277) every 30.0 s was used for continuous lock mass correction. Peptides were identified and tandem mass spectrometry fragments were scored using the PLGS 3.0 software (Waters). Peptides with a score of >7 were selected for analysis if they had a mass accuracy of at least 3 ppm and were present in at least two independent runs. Deuterium uptake was determined by calculating the shift in the centroids of the mass envelopes for each peptide compared with the undeuterated controls, using the DynamX 3.0 software (Waters). The amount of deuteration was corrected for back-exchange (∼34%) based on a full-deuteration control, using a peptide covering the unstructured, fully solvent-accessible S-tag region as reporter.

Data analysis

The fractions of monomer and dimer by SEC experiments at different temperatures were quantified from the elution profiles by fitting to a bi-Gaussian distribution. The data were then analyzed according to a simple dissociation model, as described in Eq. 1:

$${\text{N}}_2\leftrightarrow 2\text{M},$$ (1)

where D and M represent the dimer and monomer, respectively. The dissociation constant (K_D) and Gibbs free-energy change of dissociation (ΔG_D) could then be calculated from the slope of a plot of [M]² versus [D], according to Eqs. 2 and 3, respectively:

$$K_{\text{D}}={\left[\text{M}\right]}^2/\left[{\text{N}}_2\right]$$ (2)

$$\Delta {\text{G}}_{\text{D}}=\text{-}R\text{T}\text{ln}{K}_{\text{D}},$$ (3)

where R represents the gas constant (0.00198 kcal⋅mol⁻¹⋅K⁻¹) and T is the temperature (in Kelvin degrees).

Eq. 3 can be rewritten to determine the enthalpic contribution to the dimer dissociation process, as represented by the van’t Hoff equation:

$$\text{ln}\left(K_{\text{D}}\right)=\text{-}\Delta \text{H}·{\mathit{R}}^{\text{-}1}{\text{T}}^{\text{-}1}+\Delta \text{S}·R^{\text{-}1}.$$ (4)

Unfolding curves were analyzed according to a three-state N₂ ↔ 2I ↔ 2U model with a monomeric intermediate or a two-state N ↔ U model. Energetic parameters, such as the change in Gibbs free energy (ΔG) and m-value (defined as a measure of the change in solvent exposure due to the presence of the denaturant) for the unfolding transitions N₂ ↔ 2I and I ↔ U, were calculated by fitting the data as described in Baez and Babul (36). ΔG and m-values for the two-state model were calculated in a manner similar to that described in Greenfield (37).

Fitting procedures were performed using the software GraphPad Prism 6.0c (http://www.graphpad.com).

Results and Discussion

Domain swapping of FoxP1 is thermodynamically favored and unaffected by mutations on the DNA-binding surface

The in vitro dimerization properties of the forkhead domain of FoxP1 (hereinafter referred to as FoxP1) were analyzed in a previous work and a K_D of 2.7⋅10⁻⁵ M at 37°C was estimated (16). Moreover, the hinge region has been shown to constitute a critical component for the dimerization phenomenon, since the evolutionary-based mutation of an alanine residue located in this region by proline (A39P), a highly conserved residue in all other Fox members, leads to abolishment of dimer formation, as shown for FoxP2 (14). However, the effect of mutating residues linked to DNA binding on the ability of FoxP1 to domain swap, especially those located contiguous to the hinge region of FoxP in helix H3, has not been explored. Here, we chose to study the dimer-monomer equilibrium of wild-type FoxP1 and the helix H3 mutant R53H, a highly conserved residue that was previously associated with severe speech disorders in FoxP2 (12) and is known to completely abolish binding to DNA (38).

We used SEC under equilibrium conditions to explore the dissociation properties of wild-type FoxP1 and the R53H mutant, and determined their K_D values at different temperatures ranging from 17°C to 37°C (Fig. S1 in the Supporting Material). As shown in Fig. 2 A, the monomer-dimer equilibrium of wild-type FoxP1 exhibits a strong temperature dependence. Although the estimated K_D value at 37°C is in good agreement with previous reports (16), the K_D value for wild-type FoxP1 decreased by 1 order of magnitude when the temperature was lowered from 37°C to 17°C (Table 1). In all cases, the K_D values are 1–3 orders of magnitude lower than those observed for other models of dimerization via domain swapping, such as RNase A (39), p13suc1 (40), and cyanovirin-N (23), which usually require the addition of denaturants to facilitate this process. This shows that FoxP1 is more prone to oligomerization than other well-described domain-swapping proteins, and emphasizes the biological relevance of this mechanism. Using Eq. 2 and the estimated K_D values, we determined the free-energy change upon dissociation (ΔG_D) at all assayed temperatures (Table 1) and found that dimer dissociation was energetically unfavored, with ΔG_D = ∼7.2 kcal⋅mol⁻¹ at 25°C.

Figure 2.

Temperature dependence of dimer dissociation under equilibrium conditions for wild-type FoxP1 and the R53H mutant. (A) Different dimer concentrations of wild-type FoxP1 were incubated at 37°C (solid circles), 33°C (open circles), 30°C (solid squares), 25°C (open squares), 20°C (solid diamonds), and 17°C (open diamonds) until equilibrium was reached. Monomer (M) and dimer (D) fractions were quantified by fitting the elution profiles to bi-Gaussian distributions and posteriorly plotted according to Eq. 2, where K_D values were obtained from the slope using a linear fitting. (B) van’t Hoff plot for wild-type FoxP1 (solid circles) and the R53H mutant (open circles) at the specified temperatures. Linear fitting was done according to the van’t Hoff equation (Eq. 4) to estimate the ΔH_D for dimer dissociation in wild-type FoxP1 (23.1 ± 1.3 kcal⋅mol⁻¹) and the R53H mutant (22.7 ± 1.1 kcal⋅mol⁻¹).

Table 1.

Thermodynamic Parameters for Dimer Dissociation of Wild-Type FoxP1 and the R53H Mutant as a Function of Temperature Using SEC

Temperature (°C)	Wild-Type		R53H Mutant
KD (×10−6 M)	ΔGD (kcal⋅mol−1)	KD (×10−6 M)	ΔGD (kcal⋅mol−1)	ΔΔGD (Mutant-Wild-Type)
17	1.96 ± 0.24	7.55 ± 0.08	ND	ND	ND
20	3.10 ± 0.28	7.36 ± 0.05	ND	ND	ND
25	5.15 ± 0.54	7.18 ± 0.07	4.12 ± 0.25	7.32 ± 0.04	0.13
30	10.48 ± 0.15	6.88 ± 0.01	6.45 ± 0.32	7.17 ± 0.03	0.29
33	15.36 ± 1.18	6.72 ± 0.05	ND	ND	ND
37	27.40 ± 1.34	6.45 ± 0.03	18.30 ± 1.16	6.70 ± 0.04	0.25

ND, not determined.

We also determined the dissociation equilibrium of the DNA-binding mutant R53H at 25°C, 30°C, and 37°C (Fig. S2 A). The results obtained show a similar behavior compared with the wild-type protein (Fig. S2 B), and the calculated energetic parameters were slightly lower than those estimated for the wild-type protein, by ∼0.2 kcal⋅mol⁻¹ (Table 1), showing that association was not significantly altered by the R53H mutant.

Using van’t Hoff analysis, we observed a linear dependence for the estimated K_D values for both wild-type FoxP1 and the R53H mutant as a function of temperature (Fig. 2 B), suggesting that their van’t Hoff enthalpies of dissociation (ΔH_D) are constant in the temperature range of 17–37°C. The estimated ΔH_D values for the two proteins are strongly similar, being 23.1 ± 1.3 kcal⋅mol⁻¹ for wild-type FoxP1 and 22.7 ± 1.1 kcal⋅mol⁻¹ for R53H. Our results are consistent with an enthalpy-driven association process, which is in good agreement with other well-described domain-swapped dimers (40, 41, 42).

From these results, we can conclude that domain swapping in FoxP1 is thermodynamically favored and that the helix H3 DNA-binding mutation R53H does not significantly affect this mechanism.

Equilibrium unfolding of domain-swapped FoxP1 follows a three-state folding mechanism in which dimer dissociation precedes protein unfolding

In all domain-swapping proteins studied to date, monomer unfolding is required to facilitate the segment exchange that characterizes the native, intertwined dimer structure (19, 21, 43, 44). In many of these models, dimer unfolding is best described as a two-state (N₂ ↔ 2U) mechanism, with no thermodynamic intermediates shown under equilibrium conditions (22, 23).

The fact that FoxP1 dimerization can occur at micromolar protein concentrations (Table 1) within hours of incubation (Fig. S3), without the need for additional perturbations as in other domain-swapping proteins (19, 43), suggests that we need to revise our view of its folding mechanism. To that end, we explored equilibrium unfolding as a function of both Gnd·HCl and protein concentrations to determine whether subunit dissociation precedes protein unfolding.

We monitored secondary-structure changes due to unfolding by CD at 223 nm and observed two clear transitions at a protein concentration of 13 μM (Fig. 3 A). The first transition accounts for ∼15% of the overall difference between the signals of native and unfolded proteins and is observed from 0.1 to 1.8 M of Gnd·HCl, with its midpoint (Cm) at 0.53 M of Gnd·HCl. The second transition occurs from 2.7 to 5.5 M of Gnd·HCl, with its Cm at 4.1 M, and leads to total protein unfolding. Refolding curves obtained by diluting unfolded FoxP1 from 6 M Gnd·HCl show a high superposition with the unfolding curves (Fig. 3 A), indicating that the unfolding of FoxP1 is a fully reversible process under these conditions and suggesting the presence of an intermediate state.

Figure 3.

Equilibrium unfolding of wild-type, R53H, and A39P FoxP1. Protein samples were incubated at several Gnd·HCl concentrations at 25°C and changes in ellipticity at 223 nm were monitored by CD. (A–D) All data obtained for wild-type FoxP1 (A and B) and R53H (C) were fitted to a three-state folding model with a monomeric intermediate, whereas data for the A39P mutant (D) were fitted to a two-state folding model (solid lines). (A) Equilibrium unfolding (black circles) and refolding (open squares) of wild-type FoxP1 at a protein concentration of 13 μM. The decay of the native fraction as a function of the Gnd·HCl concentration shows two transitions. (Inset) CD spectra for native (solid line), incubated at 2 M of Gnd·HCl (dotted line), and refolded (dashed line) protein. The spectra for the native and refolded proteins are superimposed. (B) Equilibrium unfolding of wild-type FoxP1 at protein concentrations of 3 μM (triangles), 13 μM (black circles), and 43 μM (inverted triangles). Only the first transition changes upon an increase in the protein concentration. (Inset) Change in the Cm for the first transition as a function of the protein concentration. (C) Equilibrium unfolding of the DNA-binding mutant R53H at a protein concentration of 13 μM as a function of Gnd·HCl. (D) Equilibrium unfolding of the monomeric mutant A39P at a protein concentration of 3 μM as a function of Gnd·HCl concentration. (Inset) CD spectra for native (solid line), incubated at 2 M of Gnd·HCl (dotted line), and native wild-type (dashed line) FoxP1. The spectra for the native and 2 M Gnd·HCl samples are superimposed.

We then varied the protein concentration to determine whether the first transition corresponds to a dimer-monomer dissociation step. In Fig. 3 B, it is clear that increasing the protein concentration from 3 to 43 μM causes an increase in the Cm for the first transition toward higher Gnd·HCl concentrations (Cm[3 μM] = 0.38 M; Cm[43 μM] = 1.14 M), whereas the second transition is largely unaffected, consistent with the idea that the first transition corresponds to dimer dissociation.

These results indicate that the simplest model to describe the equilibrium unfolding of FoxP1 is a three-state folding mechanism with a monomeric intermediate N₂ ↔ 2I ↔ 2U, where I represents a monomeric intermediate with ∼85% of overall secondary-structure content compared with the native protein. Energetic parameters obtained by fitting the CD data to this model are shown in Table 2. According to these results, the total free-energy change between the native dimer and unfolded monomer (ΔG₁ + 2ΔG₂) is ∼23 kcal⋅mol⁻¹, with the change of free energy related to the dissociation event (N₂ ↔ 2I) being 8.3 ± 0.3 kcal⋅mol⁻¹ and contributing ∼36% of the total change, whereas the change in free energy due to unfolding of the intermediate (I ↔ U) is 7.4 ± 0.3 kcal⋅mol⁻¹. It is worth noting that the free energy of dissociation into the intermediate is in good agreement with the ΔG_D estimated from the K_D values obtained by SEC and van’t Hoff analyses (Table 1).

Table 2.

Thermodynamic Parameters Calculated from the Equilibrium Unfolding of Wild-Type, R53H, and A39P FoxP1 as a Function of the Concentration of Gnd⋅HCl at 25°C

	Wild-Typea			R53Ha			A39Pb
N2 ↔ 2I	I ↔ U	N2 ↔ 2U	N2 ↔ 2I	I ↔ U	N2 ↔ 2U	N ↔ U
ΔG0 (kcal⋅mol−1)	8.31 ± 0.33	7.36 ± 0.28	23.03 ± 0.52	9.21 ± 0.29	6.02 ± 0.38	21.25 ± 0.61	9.33 ± 0.91
-m (kcal⋅mol−1⋅M−1)	2.30 ± 0.08	1.88 ± 0.03	6.06 ± 0.09	2.20 ± 0.11	1.78 ± 0.11	5.76 ± 0.19	1.90 ± 0.22

^aProtein concentration = 13 μM.

^bProtein concentration = 3 μM.

An equilibrium unfolding analysis of the R53H mutant of FoxP1 at a protein concentration of 13 μM, which favors dimer formation (Fig. 3 C), further confirms the proposed three-state folding mechanism. A comparison of the unfolding free-energy change between wild-type FoxP1 and the R53H mutant shows that the dissociation event (N₂ ↔ 2I) for R53H increased by ∼1 kcal⋅mol⁻¹, whereas the free energy required to unfold the monomeric intermediate decreased by ∼1.3 kcal⋅mol⁻¹, resulting in an overall decrease in stability of ∼2 kcal⋅mol⁻¹ (Table 2). Moreover, the dimer fraction observed after incubation of the monomeric species at 37°C is higher for the R53H mutant than for the wild-type FoxP1 when starting from the same protein concentration (Fig. S3). The similarities between the R53H mutant and wild-type FoxP1, in terms of their folding mechanism and energetic parameters, provide further support for the proposed three-state folding mechanism and the thermodynamic favorability of the domain-swapping process for this protein.

The m-values calculated for the dissociation and unfolding processes of wild-type FoxP1 are 2.30 ± 0.08 and 1.88 ± 0.03 kcal⋅mol⁻¹⋅M⁻¹, respectively. Using the theoretical relationship between the m-value and the difference of accessible surface area (ΔASA) for the unfolding process proposed by Myers et al. (45), the m-value of the first event (N₂ ↔ 2I) represents ∼36% of the total ΔASA calculated for FoxP1 in a complete unfolding process. Moreover, the theoretical ΔASA value (∼4132 Å²) is in good agreement with the one estimated by comparing the domain-swapped dimer of FoxP1 in Fig. 1 and the solution structure of either the open domain-swapping monomer (2 × ASA_monomer − ASA_dimer = 4882 Å²) or the well-folded monomer (3512 Å²). This low ΔASA for the dissociation step is not only consistent with the small change in secondary structure associated with this event but also in strong contrast to the ΔASA estimated for other domain-swapping proteins, where the extent of surface exposure to the solvent is equivalent to what is exposed by the monomer upon unfolding (20, 22, 46).

To confirm that the first transition in the unfolding of dimeric FoxP1 is fully explained by dimer dissociation, we engineered the hinge region substitution A39P, which was previously characterized as leading to a monomeric form in FoxP (14, 16). Our results show that the A39P mutant is monomeric even at protein concentrations as high as 120 μM (Fig. S4) and undergoes a single transition as monitored by CD measurements (Fig. 3 D). The resulting data for A39P are best explained by a two-state monomer unfolding mechanism (N ↔ U), with its midpoint being similar to that of the second transition observed for wild-type and R53H FoxP1. Moreover, a comparison of the CD spectra of the A39P mutant in native conditions and in the presence of 2 M of Gnd·HCl (Fig. 3 D, inset) confirms the scarce loss of secondary structure under these denaturing conditions, in contrast to the observations for the wild-type protein (Fig. 3 A, inset), reinforcing the conclusion that the first transition observed for both wild-type and R53H FoxP1 (Fig. 3, A and C) corresponds to both dissociation and small structural changes that precede the unfolding process.

Previous work in other domain-swapping proteins has shown that mutational perturbations that decrease monomer stability lead to an increased propensity to oligomerize (23, 47), consistent with the notion that most domain-swapping proteins studied to date require some degree of monomer unfolding to rapidly reach the intertwined dimer (48). Concomitantly, the free energy of unfolding of A39P FoxP1 was 9.3 ± 0.9 kcal⋅mol⁻¹ (Table 2), which is 2 and 3 kcal⋅mol⁻¹ higher than the free energy of unfolding of the monomeric intermediate of wild-type FoxP1 and R53H mutant, respectively. These results further confirm the observations of Chu et al. (16), indicating that the Ala-Pro substitution in the hinge region results in stabilization of the monomer and impairment of dimer formation. These results suggest an association between dimer formation and monomer folding stability, perhaps implicating the flexibility of the hinge loop in domain swapping in FoxP1. Also, the m-value for the single unfolding transition of the A39P mutant is within error of those calculated from the I ↔ U transition in wild-type FoxP1 and the R53H mutant, hinting that only small structural changes come along with the Gnd·HCl-induced dissociation step of the intertwined dimer of FoxP1.

To further corroborate the proposed three-state folding mechanism, we monitored the effect of Gnd·HCl on the hydrodynamic properties of FoxP1 by SEC experiments, using the monomeric mutant A39P as a control for the hydrodynamic properties of the monomer. Under native conditions, the Stokes radius (Rs) values obtained for the monomer and dimer were 17.5 and 23.8 Å, respectively (Fig. 4 A), in good agreement with theoretically predicted Rs values based on their molecular weights (49) (18.4 and 23.8 Å for the monomer and dimer, respectively).

Figure 4.

Effect of Gnd·HCl on the hydrodynamic properties of wild-type FoxP1 and the monomeric mutant A39P. (A) The Rs for each protein was determined by SEC as a function of the Gnd·HCl concentration. The fractions of monomer and dimer for both wild-type FoxP1 (solid line) and the monomeric mutant A39P (dashed line) were posteriorly determined. (B) Changes in the Rs for the wild-type FoxP1 dimer (open circles) and its isolated monomer (solid circles), and for the monomeric mutant A39P (open squares). The dimer fraction as a function of the denaturant concentration (crosses) is also shown. The protein concentration was 13 μM in all conditions.

With increasing denaturant concentrations in the range of 0–1.7 M, where the N₂ ↔ 2I dissociation event occurs, no species other than monomers and dimers were observed. Interestingly, a linear increase in Rs up to ∼20% was observed for both the dimer (28.6 Å) and monomer (21.0 Å) species of wild-type FoxP1 in this concentration range (Fig. 4, A and B), suggesting that the protein somewhat expands during this transition. A priori, this behavior could be attributed to structural rearrangements in helices H2 and H3 (Fig. 1 B) that were previously observed in the solved monomer and dimer structures of FoxP1 (16), FoxP2 (14), and FoxP3 (15). Consequently, this change in Rs is, at least for the monomer species, highly consistent with the increase in the radius of gyration between the structure of the well-folded monomer (PDB: 2KIU, 14.0 Å) and the open monomer observed in the domain-swapped structure (Fig. 1; 16.5 Å, ∼18% larger). After complete dissociation is achieved (Fig. 4 A; Gnd·HCl 1.7 M), unfolding of the monomeric intermediate proceeds until a final Rs of ∼32 Å is reached, consistent with the theoretically calculated values for the hydrodynamic properties of an unfolded chain (49).

Altogether, both the effect of Gnd·HCl on the hydrodynamic properties of wild-type FoxP1, as ascertained by SEC experiments, and the changes in secondary-structure content upon increasing denaturant concentrations, as monitored by CD measurements, are compatible with a three-state folding mechanism in which dissociation and small changes in compactness and secondary structure occur before the monomer unfolds. Intrigued by these results and the plausibility of an open conformation en route to domain swapping in FoxP1, we focused on resolving the structural features of the monomeric intermediate.

Native-like features of the monomeric intermediate of FoxP1

The results presented above confirm the presence of a monomeric intermediate in the folding and binding mechanism of wild-type FoxP1. Both CD and SEC experiments suggest that reaching of this intermediate state is accompanied by a scarce loss of secondary structure (Fig. 3 A) and a small change in its hydrodynamic properties (Fig. 4) in comparison with the native state.

To describe the structural features of the native and intermediate states of FoxP1, we performed an HDXMS analysis on a freshly SEC-isolated monomeric form of wild-type FoxP1 under native conditions and in the presence of 2 M Gnd·HCl, where the intermediate is observed. Both samples were allowed to exchange in deuterated buffer for 5 min at 25°C and then posteriorly quenched, digested with immobilized pepsin, and analyzed by mass spectrometry, whereas the monomeric mutant A39P was incubated with 2 M Gnd·HCl and used as a control for the extent of exchange in the nonswapping monomer. We only analyzed peptides that were identified in both wild-type and A39P FoxP1 pepsin-digested samples, which covered ∼84% of the forkhead domain sequence (Table 3; Fig. S5). As shown in Fig. 5, the native monomer exhibits a low extent of exchange, ∼38% on average, suggesting that the protein is well folded in these buffer conditions. The largest deuterium uptake (∼70%) is observed in residues 3–12 (Fig. 5), spanning the N-terminus of the forkhead domain of FoxP1 (peptide 1; Table 3), which is in good agreement with the lack of structure in this region in the solved structures of FoxP1 and FoxP2 (14, 16). To further confirm that the HDXMS analysis ascertained the native state of the monomeric form of wild-type FoxP1, we compared the extent of exchange of all covered regions at 25°C with the total solvent-accessible surface area (SASA) from the monomeric FoxP1 structure (PDB: 2KIU), and obtained a correlation coefficient of 0.81, with most peptides being within the 95% confidence interval (Fig. S6).

Table 3.

List of Peptides Obtained for Wild-Type and A39P FoxP1 as Quantified by ESI-TOF Mass Spectrometry

No. of Peptides	Peptide Mass (m/z)	No. of Amides	Sequence	Region of FoxP1
1	1150.625	7	VRPPFTYASL	3–12
2	1379.006	13	IRQAILESPEKQLTL	13–27
3	1982.091	15	IRQAILESPEKQLTLNE	13–29
4	1341.762	10	AILESPEKQLTL	16–27
5	1044.557	7	LESPEKQLT	18–26
6	1287.642	9	ESPEKQLTLNE	19–29
7a	2535.348	20(19)	FA(P)YFRRNAATWKNAVRHNLSL	38–58
8	2007.11	16	RRNAATWKNAVRHNLSL	42–58
9	1337.743	10	WKNAVRHNLSL	48–58
10	1256.711	10	VRVENVKGAVW	63–73
11	1700.896	14	VRVENVKGAVWTVDE	63–77
12	1315.759	8	VEFQKRRPQK	79–87
13	1216.69	7	EFQKRRPQK	80–87

^aThe Ala to Pro mutation in this peptide changes the number of available amides.

Figure 5.

Comparison of amide exchanges between monomer wild-type FoxP1 under native conditions, its monomeric intermediate at 2 M Gnd·HCl, and the monomeric mutant A39P incubated in 2 M of Gnd·HCl. Protein samples were allowed to exchange for 5 min at 25°C in deuterated buffer and then quenched, pepsin digested, and analyzed by mass spectrometry to determine their extent of exchange. Monomeric wild-type FoxP1 under native conditions is shown in black bars. The monomeric mutant A39P that was incubated in 2 M of Gnd·HCl and the monomeric intermediate of wild-type FoxP1 are shown in white and dotted bars, respectively. Peptide numbering is indicated in Table 3. Data are shown as the percentage of deuterium uptake according to the maximum theoretical uptake for each peptide.

A comparison of the extent of exchange between monomeric wild-type FoxP1 under native conditions and its intermediate state at 2 M Gnd·HCl reveals that the latter exhibits only a small increase in deuteration for all peptides, ranging from 1% to 17% and being on average 8% (Fig. 5). The largest difference is observed for residues 19–29 (peptide 6; Table 1), corresponding to strand S1, which precedes the domain-swapping helix H2. This can be attributed to a loosening of the interactions among strands S1–S3, which are formed within the same polypeptide chain only in the monomeric state (Fig. 1 A). Consistently, strands S2 and S3, covered by peptide 11 (Table 1), show a 9% increase in deuterium uptake. Peptides 1–3 (Table 1), spanning residues 3–27, also show an increase in deuterium uptake of 12%, which can be attributed to a loosening of the first helical turn of helix H1 and the aforementioned loosening of strand S1. In contrast, peptides 7–9, 11, and 12, spanning regions H3, H4, and W2 (Table 1), show only small changes, suggesting that these segments are relatively unperturbed in this intermediate. All of these results are in good agreement with our previous unfolding experiments using CD, showing small secondary-structure changes (Fig. 3 A) and increases in hydrodynamic radius (Fig. 4) in the N₂ ↔ 2I transition. Together with the HDXMS analysis (Fig. 5), the high helical content of FoxP1 (Fig. 1), and the changes in secondary structure and hydrodynamic radius obtained from CD and SEC experiments (Figs. 3 and 4), our results suggest that the intermediate state could be interpreted as an open native-like conformation accompanied by a soft loosening of the helical structures.

Peptides obtained for the monomeric mutant A39P under the same denaturant conditions that were used to obtain the monomeric intermediate of FoxP1 (i.e., 2 M Gnd·HCl) show no significant differences when compared with the native wild-type monomer (Fig. 5), with the exception of residues 3–12, which also show an increase in deuterium exchange of 8% (Fig. 5). Taking into account the behavior of the monomeric mutant from our unfolding experiments (Fig. 3 D), our HDXMS analysis suggests that the secondary structure of this mutant is less perturbed by denaturants than the wild-type protein at 2 M Gnd·HCl, with the exception of strand S1, which, as mentioned previously, interacts with strand S3 within the same polypeptide chain in the monomeric state. In contrast, wild-type FoxP1 shows both a loosening of its secondary structure and an increase in the extent of exchange of strand S1 concomitantly with the larger decrease in secondary structure (Fig. 3).

Altogether, our results suggest that the intermediate state corresponds to an open, native-like ensemble with small secondary-structure and hydrodynamic changes.

Conclusions

In this work, we studied the folding and binding mechanism of the forkhead domain of FoxP1 and of two mutants that abolish DNA binding and dimer formation (Fig. 1). SEC experiments showed that both wild-type FoxP1 and the DNA-binding mutant R53H domain swap at micromolar protein concentrations (Fig. 2) within hours of incubation at 37°C, confirming the feasibility of dimer formation via domain swapping without the need for additional perturbations, and reinforcing its biological relevance as a plausible mechanism to enable long-range chromosomal interactions (34).

Equilibrium unfolding experiments assessed by CD showed that both wild-type FoxP1 (Fig. 3 A) and the R53H mutant (Fig. 3 C) follow a three-state folding mechanism in which dimer dissociation occurs before protein unfolding, and that the Arg-His substitution does not have any relevant effects on the dimerization process. During this transition, the surface that is being exposed to the solvent, as ascertained by the m-value, is smaller than the expected exposure for a fully unfolded protein. This is a strong difference from previous observations for most domain-swapping proteins, where the m-value was consistent with thorough protein unfolding (20, 22, 46). In contrast, the A39P mutant follows a two-state folding mechanism (Fig. 3 D), consistent with the second transition observed for the domain-swapped FoxP1 proteins. By comparing the free energy of unfolding of the monomeric intermediate of both wild-type FoxP1 and the R53H mutant with the folding stability of the monomeric mutant A39P, we can conclude that the propensity for dimerization in FoxP1 is at least partially linked to monomer stability, as seen in other domain-swapping proteins.

On the basis of the small secondary-structure changes and meager increase in hydrodynamic radius observed for the intermediate state using SEC (Fig. 4), we performed a detailed analysis of the native and intermediate states of wild-type FoxP1 using HDXMS (Fig. 5). These experiments revealed that this intermediate largely resembles the structural features of the native state of FoxP1, with the main exception of strands S1–S3, which only interact when the protein is properly folded and seem to become loose at 2 M Gnd·HCl, thus providing evidence, in combination with our SEC and CD experiments, that the intermediate ensemble could correspond to an open native-like conformation.

Given that the most striking feature that separates FoxP from all other members of the Fox protein family is the Pro-Ala substitution in the FPYF motif, the energetic compensations that allow domain swapping in FoxP proteins must be largely associated with the structural rearrangements that occur in this region, along with the establishment of additional interfaces in the dimeric form of this protein during the evolution of the FoxP subfamily, as seen for FoxP3 (15). Such rearrangements (e.g., as illustrated in Fig. 1 B) would allow rapid domain swapping through transient structural changes rather than overall protein unfolding.

Altogether, we conclude that dimeric FoxP1 follows an unusual folding and binding mechanism compared with other domain-swapping proteins (Fig. 6), where local structural rearrangements in the hinge region, such as those seen in solved crystal structures of FoxP proteins, rather than complete unfolding, could be sufficient to reach the native dimer starting from isolated monomers.

Figure 6.

Proposed folding mechanism of the domain-swapped form of FoxP1. Protein unfolding (U) of the native dimer (N₂) of FoxP1 is preceded by its dissociation into a native-like monomeric species (I). The graph shows the fraction of each species as a function of Gnd·HCl concentration, calculated based on the thermodynamic parameters obtained for FoxP1 at a protein concentration of 43 μM.

Author Contributions

Designed research, E.M., E.A.K., C.A.R.-S., and J.B.; Performed research, E.M., C.C., P.V., and J.R.; Analyzed data, E.M. and C.A.R.-S.; Wrote the manuscript, E.M., E.A.K., C.A.R.-S., and J.B.

Editor: H. Jane Dyson.