The Structure and Unfolding of the Third Type III Domain from Human Fibronectin

Abstract

Fibronectin is a modular extracellular matrix protein that is essential for vertebrate development. The 3^rd type III domain (3FN3) in fibronectin interacts with other parts of fibronectin and with anastellin, a protein fragment that causes fibronectin aggregation. 3FN3 opens readily both as an isolated domain in solution and when part of fibronectin in stretched fibrils, and it was proposed that this opening is important for anastellin binding. We determined the structure of 3FN3 using nuclear magnetic resonance spectroscopy and we investigated its stability, folding and unfolding. Similar to most other FN3 domains, 3FN3 contains two antiparallel β-sheets that are composed of three (A, B and E) and four (C, D, F and G) β-strands, respectively, and are held together by a conserved hydrophobic interface. Cis-trans isomerization of P847 at the end of β-strand C leads to observable conformational heterogeneity in 3FN3, with a cis peptide bond present in almost one quarter of the molecules. The chemical stability of 3FN3 is relatively low, but the folding rate constant in the absence of denaturant is in the same range as those of other, more stable FN3 domains. Interestingly, the unfolding rate constant in the absence of denaturant is several orders of magnitude higher than the unfolding rate constants of other FN3 domains investigated to date. This unusually fast rate is comparable to the rate of 3FN3 binding to anastellin at saturating anastellin concentrations, consistent with the model that 3FN3 has to unfold in order to interact with anastellin.

Fibronectin is an essential extracellular matrix protein (reviewed in ^1-3) that plays an important role in cell adhesion and migration and affects cell proliferation, survival and differentiation. It is required for vertebrate development, and mice that lack fibronectin die during embryogenesis with shortened anterior-posterior axis, absent notochord and somites, and abnormal heart and vasculature ⁴. Aside from development, fibronectin serves an important function in tissue repair ^5-7.

Fibronectin is composed of 12 fibronectin type I (FN1), two fibronectin type II (FN2) and 15-17 fibronectin type III (FN3) homologous repeats, and is secreted as a ~500 kDa disulfide-linked dimer. Formation of fibrils from soluble fibronectin dimers is not spontaneous and generally occurs only in the presence of cells, which assemble fibronectin fibrils around themselves. The molecular basis of the conversion of soluble fibronectin into insoluble fibrillar aggregates is poorly understood, and the atomic resolution structure of the fibrils still remains to be elucidated ^{3, 8}.

Some parts of fibronectin have been studied extensively, but relatively little information is available about the 3^rd FN3 domain (3FN3) (Fig. 1). Together with the 2^nd FN3 domain, 3FN3 was proposed to maintain soluble fibronectin dimers in a compact conformation by ionic interactions with the 12^th-14^th FN3 domains ⁹ and the 2^nd-5^th FN1 domains ¹⁰. 3FN3 was also reported to mediate cell adhesion and migration through binding to a so far unidentified β1 integrin ¹¹. Finally, 3FN3 and several other FN3 domains were shown to interact with a small fragment of fibronectin called anastellin ^12,13.

Figure 1.

Structure-based sequence alignment of FN3 domains from human fibronectin. The sequences are colored by percent identity ⁵⁷. The sequence of 3FN3 and a diagram of its secondary structure are shown at the top. EDA and EDB stand for the alternatively spliced extra domain A and B, respectively. The 4^th-6^th, 11^th and 15^th FN3 domains are not included because their structures are not available.

Anastellin exhibits anti-tumor, anti-metastatic and anti-angiogenic properties ^{14, 15} and requires endogenous fibronectin for these in vivo activities ^{16, 17}. The binding of anastellin to fibronectin leads to conversion of the soluble fibronectin dimers to insoluble aggregates that are reminiscent of fibronectin fibrils deposited by cells ^{18, 19}. It is not known how anastellin interacts with its target FN3 domains and why this interaction causes aggregation of fibronectin. It was proposed that transient opening of the target FN3 domains is important for binding to anastellin ^{12, 13, 20}.

We employed nuclear magnetic resonance (NMR) spectroscopy, equilibrium denaturation and stopped-flow methods to characterize 3FN3 and its opening behavior. We determined the solution structure of 3FN3, and we identified a minor 3FN3 conformer that is populated under native conditions. We also measured the stability of 3FN3 and investigated its folding and unfolding kinetics, and we show that the unfolding rate constant in the absence of denaturant is significantly higher than the unfolding rate constants reported to date for other isolated FN3 domains. This unusually fast unfolding rate is in the same range as the rate of 3FN3 binding to anastellin at saturating anastellin concentrations, consistent with the model that 3FN3 has to unfold in order to interact with anastellin.

Experimental Procedures

Expression and purification of 3FN3

The pET15b expression vector encoding 3FN3 was provided to us by Dr. Tomoo Ohashi and Dr. Harold Erickson (Duke University) ^{12, 13}. The His-tagged fusion protein was expressed in Escherichia coli BL21-CodonPlus(DE3)-RIPL cells (Agilent) and purified by affinity chromatography on a HisTrap FF column (GE Healthcare). The N-terminal His-tag was cleaved off with thrombin, and 3FN3 was separated from thrombin and the His-tag by affinity chromatography on HiTrap Benzamidine FF (high sub) (GE Healthcare) and HisTrap FF columns respectively. 3FN3 was then further purified by size-exclusion chromatography on Superdex 75 resin (GE Healthcare). The recombinant 3FN3 protein consisted of six extraneous amino acids at the N-terminus, residues 808-905 of human fibronectin (P02751), and two extraneous amino acids at the C-terminus (gshmgtTTAP...PRSDgt).

NMR spectroscopy

Samples for NMR spectroscopy typically contained 0.6 mM unlabeled, uniformly ¹⁵N-labeled or uniformly ¹³C/¹⁵N-labeled 3FN3 in phosphate buffered saline (PBS) (pH 7.5) supplemented with 10% ²H₂O. Most NMR data were acquired at 25°C on a Varian 600 MHz NMR System that was equipped with a triple resonance probe or a ¹³C-enhanced salt-tolerant cold probe. NMR experiments that were used to assign ¹H, ¹³C and ¹⁵N chemical shifts and to generate restraints for structure calculations included 2-dimensional (2D) ¹H-¹⁵N heteronuclear single-quantum correlation (HSQC), ¹H-¹³C aliphatic and aromatic HSQC, double-quantum-filtered correlation spectroscopy (COSY), total correlation spectroscopy (TOCSY), nuclear Overhauser effect spectroscopy (NOESY), HBCB(CGCD)HD, ¹H-¹⁵N heteronuclear multiple-quantum correlation (HMQC) optimized for histidine imidazole rings, 3-dimensional (3D) HNCACB, CBCA(CO)NH, C(CO)NH, H(CCO)NH, HNCO, HN(CA)CO, CCH-TOCSY, HCCH-TOCSY, ¹⁵N-edited TOCSY, ¹⁵N-edited NOESY, ¹³C-edited aliphatic NOESY (with and without sensitivity enhancement) and ¹³C-edited aromatic NOESY. In addition, a 2D ¹³C/¹⁵N-filtered, ¹³C/¹⁵N-edited NOESY spectrum ²¹ was acquired on a Varian 800 MHz NMR System equipped with a salt-tolerant ¹³C-enhanced cold probe for a sample containing 0.9 mM ¹³C/¹⁵N-labeled 3FN3 and 0.9 mM unlabeled 3FN3. To evaluate backbone conformational flexibility, steady-state ¹H-¹⁵N nuclear Overhauser effect (NOE) was measured using 2D heteronuclear correlation experiments. All data were processed with NMRPipe ²² and analyzed with CcpNmr Analysis version 2.2.2 ²³. Φ and Ψ dihedral angle restraints were generated with TALOS-N ²⁴.

Calculation of 3FN3 structure

Structure calculations were performed using Aria 2.3 ²⁵ in conjunction with CNS 1.21 ²⁶, with peaks from the NOESY spectra and dihedral angle restraints from TALOS-N as input (Table 1). Manual assignments of the NOESY peaks were retained. ¹H-¹⁵N NOE values and TALOS-N predictions indicated that the N- and C-terminal tails were dynamic, and no medium or long-range connectivities were identified for residues preceding A810 or following T901 during manual inspection of the NOESY spectra. Aria calculations nevertheless tended to create restraints between the tails and the rest of the protein even when other, more likely assignments were available. To prevent the generation of such artefactual restraints for the tails, only chemical shifts for residues T809-P902 were included in the calculations. In addition, manually identified intraresidual and sequential NOESY peaks stemming from the dynamic residues outside of the T809-P902 region were omitted from the input.

Table 1.

Experimental restraints and structural statistics for 3FN3

Number of peaks in NOESY spectra
3D 15N-edited NOESY	1057
3D 13C-edited aliphatic NOESY (without sensitivity enhancement)	1722
3D 13C-edited aliphatic NOESY (sensitivity-enhanced)	906
3D 13C-edited aromatic NOESY	64
2D NOESY in H2O	1912
2D NOESY in D2O	1343
2D 13C/15N-filtered, 13C/15N-edited NOESY	25
Number of experimental restraints
NOE distance restraints in the last ARIA iteration
Intraresidual	873
Sequential	454
Medium range (2 ≤ |i-j| ≤5)	171
Long range (6 ≤ |i-j|)	758
Ambiguous	1179
Dihedral angle restraints (Φ and Ψ)	168
Hydrogen bond distance restraints	60
Number of experimental restraint violations
Distance (NOE and hydrogen bond) violations > 0.5 Å	0.2 ± 0.4
Dihedral angle violations > 5°	1.1 ± 0.9
Root mean square deviation (RMSD) from experimental restraints
Distance (NOE and hydrogen bond) restraints (Å)	0.040 ± 0.002
Dihedral angle restraints (°)	0.9 ± 0.1
RMSD from idealized geometry
Bonds (Å)	0.0053 ± 0.0001
Angles (°)	0.63 ± 0.01
Impropers (°)	1.75 ± 0.06
RMSD of residues 811-901 from mean coordinates
Backbone atoms (N, Ca, C’) (Å)	0.33 ± 0.08
Heavy atoms (Å)	0.73 ± 0.06
Distribution of Φ and Ψ dihedral angles of residues 811-901 in Ramachandran plot 56
Most favored regions (%)	83.6
Additional allowed regions (%)	16.1
Generously allowed regions (%)	0.1
Disallowed regions (%)	0.3

The calculation protocol employed spin-diffusion correction ²⁷, quadrupled number of cooling steps during simulated annealing ²⁸ and otherwise mostly default Aria parameters and options. Hydrogen bond connectivity in the β-sheets was established based on NOE patterns, delayed ¹H/²H exchange and the presence of the hydrogen bonds in preliminary structures. Distance restraints were then introduced in subsequent Aria runs to reinforce the expected hydrogen bonds. In the final run, a total of 80 structures were calculated and 25 of them were refined in explicit water. The statistics for the 25 refined structures are summarized in Table 1. Figures depicting the structures were prepared with the molecular graphics program Molmol 2K.1 ²⁹. The coordinates and chemical shifts were deposited in the Protein Data Bank (PDB ID 2N1K) and the Biomagnetic Resonance Bank (BMRB ID 25564).

Equilibrium denaturation studies of 3FN3

Samples for equilibrium denaturation studies contained 1 μM 3FN3 and 0-6 M guanidine hydrochloride (GuHCl) in PBS (pH 7.4). After mixing, the samples were allowed to equilibrate at room temperature for at least 2 hours, and their fluorescence at 350 nm was then measured in a JASCO FP-8200 spectrofluorometer at 24°C, using excitation at 280 nm. For each GuHCl concentration tested, fluorescence intensities from triplicate samples were averaged. The concentration of GuHCl at which half of the protein is unfolded, [GuHCl]_50%, and the proportionality constant between the free energy of unfolding and the GuHCl concentration, m_D-N, were obtained by fitting equation 11 from ³⁰ to the denaturation data. The fitting was done in Matlab version 2012a (MathWorks Inc., Natick, MA). The free energy of unfolding in the absence of denaturant was calculated as $\Delta {\mathrm{G}}_{\mathrm{D}-\mathrm{N}}^{{\mathrm{H}}_2\mathrm{O}}=m_{D-N}{\left[\text{GuHCl}\right]}_{50\%}$³⁰.

Folding and unfolding kinetics

The changes in fluorescence that accompany folding and unfolding of 3FN3 were monitored in an Applied Photophysics SX20 stopped-flow spectrometer at 23°C, using 280 nm for excitation and a 320 nm cutoff filter to detect emission. Unfolding was initiated by rapidly mixing one volume of 10.6 μM 3FN3 in PBS (pH 7.4) with 10 volumes of PBS containing 1.40-5.00 M GuHCl. To initiate refolding, one volume of 10.6 μM 3FN3 and 3.00 M GuHCl in PBS was mixed with 10 volumes of PBS containing 0-1.14 M GuHCl. Five 5 s fluorescence traces were collected and averaged for each final GuHCl concentration. A double exponential function was then fitted to the unfolding traces and a triple exponential to the refolding traces. A 3 minute trace that was recorded for folding in 0.60 M GuHCl was fitted with a quadruple exponential function. The fitting was performed with the Pro-Data Viewer 4.1.18 software provided with the stopped-flow instrument. The unfolding and refolding rate constants in the absence of denaturant, $k_u^{H_{2}O}$ and $k_f^{H_{2}O}$, the proportionality constants between ln(k_u) or ln(k_f) and GuHCl concentration, m_ku and m_kf, and a constant describing the curvature of the unfolding rate, c, were obtained by fitting equations 1 and 2 from ³¹ to the rate constants for the major unfolding and refolding phases, respectively. The fitting was carried out in Matlab.

Results and Discussion

3FN3 is very soluble and stable and its NMR spectra are of high quality (Fig. 2). We were therefore able to assign the majority of its ¹H, ¹³C and ¹⁵N chemical shifts and we determined its solution structure on the basis of 2D and 3D homo- and heteronuclear NMR experiments.

Figure 2.

2D ¹H-¹⁵N HSQC spectrum of 3FN3. The assignments are indicated. Crosspeaks from the minor species that do not adjoin their respective major peaks are highlighted in red and denoted by an apostrophe (’). Signals from T825 and R842ε are aliased from 97.76 ppm and 83.67 ppm, respectively.

The structure of 3FN3 is shown in Fig. 3. Similar to most other FN3 domains ^32-43, 3FN3 contains 7 β-strands (labeled A-G) that form a three-stranded antiparallel β-sheet (β-strands A, B and E) and a four-stranded antiparallel β-sheet (β-strands C, D, F and G); β-strands A and G each contain a bulge (Fig. 3A,B). The two β-sheets are held together by an extensive hydrophobic interface that includes several residues highly conserved among FN3 domains, in particular P814, W831, Y841, L870, Y876 and T898 (Fig. 3C,D). The hydroxyl groups of Y841 and T898 are buried, and the signals from the hydroxyl hydrogens are observable in several NMR spectra (Fig. 4). The loops between β-strands A and B, D and E, E and F, and F and G are well-ordered, but the inter-sheet loop connecting β-strands B and C and the intra-sheet loop connecting β-strands C and D exhibit increased mobility, as indicated by the ¹H-¹⁵N NOE data (Fig. 5).

Figure 3.

Solution structure of 3FN3. A. A ribbon model of 3FN3 (residues 811-901). The color changes smoothly from blue at the N terminus to red at the C terminus, and the individual β-strands are labeled. B. 25 superimposed backbone traces of 3FN3. The coloring scheme and orientation are the same as in panel A. C and D. The hydrophobic core in 3FN3. To illustrate packing at the interface between the two β-sheets, the side chains of buried residues are shown in a space filling representation and are colored as in panels A and B. The backbone traces are grey. The orientation in C is the same as in panels A and B, and the view in D is related by a 180° rotation around the vertical axis.

Figure 4.

NMR signals from buried hydroxyl groups in 3FN3. A and B. Strips from a 2D ¹³C/¹⁵N-filtered, ¹³C/¹⁵N-edited NOESY spectrum of 3FN3 (mixing time 200 ms) with crosspeaks between Y841 Hη (panel A) or T898 Hγ1 (panel B) and nearby aliphatic hydrogens. The assignments for all observed signals are indicated. The NOESY spectrum also contains crosspeaks between Y841 Hη and W831 Hε1, and between T898 Hγ1 and V874 HN (not shown). C and D. A close-up view of Y841 (panel C) and T898 (panel D) in the 3FN3 structure. The backbone is colored as in Fig. 3A-B, and the side chains of Y841 and T898 are orange. The side chains of residues that are near the hydroxyl hydrogens according to the NOESY data are shown in yellow color. T898 Hγ1 (panel D) forms a hydrogen bond (yellow dashed line) to the backbone carbonyl oxygen of Q871.

Figure 5.

Backbone amide dynamics in 3FN3: ¹H-¹⁵N NOE data. The secondary structure of 3FN3 is outlined on top. The ¹H-¹⁵N NOE values for D905, g906 and t907 are negative and are not included in the figure.

The 2D ¹H-¹⁵N HSQC spectrum contains a set of secondary, weaker crosspeaks that stem from a minor 3FN3 species (Fig. 2). The population of this species is 23±3% as judged from peak volumes. The residues whose backbone amide chemical shifts differ the most between the minor and major forms are clustered in the C-terminal half of the four-stranded β-sheet (Fig. 6A). There are a total of 11 prolines in 3FN3, and we therefore investigated whether proline cis-trans isomerization might be responsible for the heterogeneity. The majority of the prolines are located at the N-terminal end of the domain, and only two, P847 and P872, are in the part of 3FN3 for which the heterogeneity in NMR spectra is observed (Fig. 6A). The Cα, Cβ and Cγ chemical shifts of P872 in the minor species are virtually identical to those in the major form (Fig. 6D). In contrast, the chemical shifts of P847 in the minor species (Cα 62.48 ppm, Cβ 34.93 ppm, Cγ 24.59 ppm) are significantly different from those in the major form (Cα 62.70 ppm, Cβ 32.08 ppm, Cγ 27.75 ppm) (Fig. 6C), and they are indicative of the S846-P847 peptide bond in the minor species being in a cis conformation ⁴⁴. The largest backbone amide chemical shift differences between the two forms are observed for S848, which immediately follows P847, for T854 in β-strand D, which flanks P847 on one side, and for Y876 and N877 in β-strand F, which flank P847 on the other side (Fig. 6B). Hence, all our data point to the cis-trans isomerization of the S846-P847 peptide bond as the primary source of the conformational heterogeneity in 3FN3.

Figure 6.

Conformational heterogeneity in 3FN3. A. Backbone amide chemical shift heterogeneity mapped onto the 3D structure of 3FN3. The backbone amide nitrogen of each residue with a minor crosspeak in the 2D ¹H-¹⁵N HSQC spectrum (Fig. 2) is displayed as a white sphere. The volume of the sphere is proportional to the chemical shift difference between the crosspeaks of the major and minor form. The backbone traces are colored as in Fig. 3A-B. Proline side chains are highlighted in yellow and labeled to help visualize their positions relative to the residues whose chemical shifts differ between the two forms. P902, which is located in the dynamic C-terminal tail, is not shown. The view is related to the orientation in Fig. 3A-C by a 40° rotation around the vertical axis. B. A close-up view with the residues labeled. C and D. ¹³C chemical shifts of P847 (panel C) or P872 (panel D) as observed in the 3D C(CO)NH and HNCACB spectra. The left two strips in each panel are for the major conformer, and the right two strips are for the minor conformer. Positive contour levels are black and negative levels are grey. The chemical shift assignments for the minor conformer are denoted with an apostrophe (’) and are colored red, and the assignments for the major conformer are shown in black.

Cis-trans isomerization of peptidyl-prolyl bonds in the context of a native state was observed previously in other folded proteins and, in some cases, was shown to be functionally important and act as a molecular switch ^45-49. The functional significance of the cis-trans isomerization of P847 in 3FN3, however, is not clear at this point, and the heterogeneity may just be a consequence of the relatively low stability of 3FN3 (see below) and the inability to stabilize one isomer strongly over the other ^{48, 49}. Interestingly, a number of the FN3 domains in fibronectin contain a proline residue at the position equivalent to P847 (Fig. 1), but a population of molecules with a cis peptide bond has so far not been identified in any of the structures determined by NMR spectroscopy ⁴⁰ or X-ray crystallography ^{35, 37, 42, 43}.

3FN3 interacts with anastellin, and it was proposed that the rate of binding at saturating anastellin concentrations (~ 0.02 s⁻¹) is determined by the rate at which 3FN3 spontaneously opens and exposes an otherwise cryptic binding site ¹³. However, the rates of unfolding that have been reported so far for other FN3 domains under comparable conditions are significantly slower than 0.02 s⁻¹ (7.21 ± 0.26 × 10⁻⁵ s⁻¹ for the 3^rd FN3 domain from tenascin ⁵⁰ and 2.3 ± 1.2 × 10⁻⁴ s⁻¹ for the 10^th FN3 domain from fibronectin, 10FN3 ⁵¹). Experiments probing the accessibility of engineered buried cysteines in FN3 domains suggested that 3FN3 opens readily, but the opening and closing frequency was not measured ⁵². This lack of information about the opening rates and opening equilibrium prompted us to undertake a rigorous investigation of the unfolding and refolding behavior of 3FN3.

To evaluate the stability of 3FN3, we measured the intrinsic fluorescence of the single tryptophan, W831, as a function of GuHCl concentration. The aromatic ring of W831 is buried in the hydrophobic interface between the two β-sheets (Fig. 3) and becomes exposed to solvent when 3FN3 unfolds; this is accompanied by a large increase in fluorescence emission at 350 nm (Fig. 7). Fitting of the fluorescence denaturation curve yielded the concentration of GuHCl at which half of 3FN3 is unfolded, [GuHCl]_50%, of 1.51 ± 0.02 M, and the proportionality constant between the free energy of unfolding and GuHCl concentration, m_D-N, of 2.65 ± 0.17 kcal mol⁻¹ M⁻¹. The free energy of unfolding in the absence of denaturant, $\Delta {\mathrm{G}}_{\mathrm{D}-\mathrm{N}}^{{\mathrm{H}}_2\mathrm{O}}$, was then calculated from these values to be 4.00 ± 0.26 kcal mol⁻¹. This corresponds to about 1 out of every 880 molecules being unfolded under native conditions. No deviation from a two-state transition was apparent, suggesting that the stabilities of the major and minor conformer are comparable.

Figure 7.

Equilibrium denaturation of 3FN3 in GuHCl. The black curve represents the equation fitted to the data.

The kinetics of 3FN3 unfolding and refolding were also monitored using W831 fluorescence. Two unfolding and three refolding phases were resolved in 5 s fluorescence traces at all GuHCl concentrations investigated, and an additional slow refolding phase was detected in a 3 minute experiment (Fig. 8). Outside of the transition region, the amplitudes of the individual phases are relatively independent of GuHCl concentration. The natural logarithm of the rate constant for the major refolding phase (empty black circles in Fig. 8A) varies linearly with GuHCl concentration. A linear equation, $\text{ln}\left(k_f\right)=\text{ln}\left(k_f^{H_{2}O}\right)-m_{k_f}\left[\text{GuHCl}\right]$, was therefore fitted to the data, yielding the refolding rate constant in the absence of denaturant, $k_f^{H_{2}O}$, of 58 ± 3 s⁻¹, and the proportionality constant, m_kf, of 2.57 ± 0.08 M⁻¹. In contrast, the natural logarithm of the rate constant for the major unfolding phase (filled black circles in Fig. 8A) displays non-linear dependence on GuHCl concentration, and a second degree polynomial, $\text{ln}\left(k_u\right)=\text{ln}\left(k_u^{H_{2}O}\right)+m_{k_u}\left[\text{GuHCl}\right]-c{\left[\text{GuHCl}\right]}^2$, was used to fit the data. The fitting yielded the unfolding rate constant in the absence of denaturant, $k_u^{H_{2}O}$, of 0.057 ± 0.010 s⁻¹, the proportionality constant, m_ku, of 2.1 ± 0.1 M⁻¹, and the curvature, c, of 0.15 ± 0.02 M⁻². The rate of 3FN3 folding is within the range observed for other FN3 domains but the rate of unfolding is several orders of magnitude higher than those reported previously ^{50, 51} and is comparable to the rate of anastellin binding to 3FN3 at saturating anastellin concentrations ¹³.

Figure 8.

Kinetics of 3FN3 folding and unfolding. A. The natural logarithm of the rate constants (upper panel) and the amplitudes (lower panel) are shown as a function of GuHCl concentration for two unfolding (black and red filled circles) and three refolding phases (black and red empty circles and black dashes). The slowest refolding phase that was only detected in a 3 minute experiment is not included in the plot. The solid black lines represent the equations fitted to the data for the major unfolding and refolding phases. B. A triple exponential function (left panel) or a quadruple exponential function (right panel) was fitted to a 3 minute refolding trace in 0.60 M GuHCl. The residuals are shown in the lower panels. The rate constants and amplitudes for the quadruple exponential function are k₁ = 17.5 ± 2.2 s⁻¹, a₁ = 0.38 ± 0.09 (20%), k₂ = 6.7 ± 0.3 s⁻¹, a₂ = 1.06 ± 0.09 (55%), k₃ = 0.28 ± 0.06 s⁻¹, a₃ = 0.07 ± 0.01 (4%), k₄ = 0.047 ± 0.002 s⁻¹, a₄ = 0.40 ± 0.01 (21%).

Interestingly, we also observed a minor unfolding and a minor refolding phase whose rate constants were slightly higher than those for the major phase (filled and empty red circles in Fig. 8A). The slopes and the positions of the minima in the ln(k) versus [GuHCl] plot for the minor and the major phases are very similar, and, consequently, the plot for these phases has a “nested chevron” appearance (Fig. 8A). We attribute the minor unfolding and refolding phase to unfolding and refolding of the minor 3FN3 conformer that contains a cis S846-P847 peptide bond. This interpretation is consistent with the stabilities of the minor and major conformer being indistinguishable in equilibrium denaturation experiments, and with the structure of the minor conformer overall closely resembling that of the major species (~70% of backbone amide chemical shifts are identical in the two forms). The minor kinetic phase accounts for 25 ± 5% of the unfolding amplitude, in excellent agreement with the population of the minor conformer that was estimated from the NMR data.

In addition to the major phase and the fast minor phase described above, we identified two slow minor refolding phases (Fig. 8B and black dashes in Fig. 8A). In 0.6 M GuHCl, these slow phases represent ~25% of the total refolding amplitude (Fig. 8B). We attribute the slow phases to processes in which proline cis-trans isomerization is the rate-limiting step, i.e. to refolding of denatured 3FN3 in which some of the prolines are in the non-native cis conformation ⁵³.

The kinetic data are in general consistent with a two-state folding transition, as demonstrated using several well established tests ^{31, 50, 54}. We do not observe a significant burst phase or deviation from linearity in the ln(k_f) versus [GuHCl] plot for the major refolding phase. In addition, the equilibrium unfolding constant derived from the kinetic data and adjusted for proline isomerization in the denatured and native state corresponds to free energy of unfolding, $\Delta {\mathrm{G}}_{\mathrm{D}-\mathrm{N}}^{{\mathrm{H}}_2\mathrm{O}}$, of 3.89 ± 0.12 kcal mol⁻¹. This value compares well with the free energy of unfolding from equilibrium experiments, 4.00 ± 0.26 kcal mol⁻¹. Finally, the m-value calculated from the kinetic data, m_k = RT(m_kf + m_ku), is 2.76 ± 0.08 kcal mol⁻¹ M⁻¹, in good agreement with the m-value obtained from equilibrium experiments, 2.65 ± 0.17 kcal mol⁻¹ M⁻¹.

We have inspected the sequence and structure of 3FN3 for features that might explain the relatively low stability and high unfolding rate. The structure of 3FN3 is very similar to that of other FN3 domains, and most residues in its hydrophobic core are identical to those in the markedly more stable 10FN3 or are conservatively substituted (Fig. 1 and 3). The only residue in the hydrophobic core that significantly differs from 10FN3 and from the FN3 consensus is P817. In 10FN3 and most other FN3 domains, this position is occupied by leucine (Fig. 1). In addition to hydrophobic core packing, electrostatic interactions are also known to play a role in protein stability ⁵⁵. 3FN3 is highly acidic, with 6 aspartates, 7 glutamates, 3 arginines and no lysines or histidines within the structured region (Fig. 9). It is therefore likely that unfavorable electrostatic interactions contribute to the lower stability and faster unfolding rate. The only residue that has so far been shown to destabilize 3FN3, D869, is in fact acidic ¹³. Most other FN3 domains contain a glycine or an asparagine at this position (Fig. 1), and a D869G mutation in 3FN3 increases the chemical stability of the domain and decreases the rate of binding to anastellin ¹³.

Figure 9.

Distribution of charged residues in 3FN3. The solvent accessible surface of 3FN3 is colored according to electrostatic potential, with areas of negative, positive, or neutral electrostatic potential depicted in red, blue, or white, respectively. In panel A, the orientation is the same as in Fig. 3A-C, and the view in panel B is related by a 180° rotation around the vertical axis.

Overall our results demonstrate that while the structure and folding behavior of 3FN3 are similar to those of other FN3 domains, 3FN3 differs from previously characterized FN3 domains in several aspects. Firstly, 3FN3 is conformationally heterogeneous due to a cis-trans peptide bond isomerization at the semi-conserved P847. Secondly, 3FN3 displays relatively low stability and significantly higher unfolding rate than other isolated FN3 domains investigated to date. Ohashi et al. proposed that when 3FN3 interacts with anastellin, the rate of binding at saturating anastellin concentrations (~ 0.02 s⁻¹) is determined by the rate at which 3FN3 spontaneously opens ¹³. We have now shown that the experimentally determined rate of 3FN3 unfolding (~ 0.06 s⁻¹) is comparable to the rate observed by Ohashi, consistent with the model that 3FN3 has to open in order to bind anastellin.

Abbreviations

2D

two-dimensional

3D

three-dimensional

10FN3

the tenth fibronectin type III domain from human fibronectin

3FN3

the third fibronectin type III domain from human fibronectin

COSY

correlation spectroscopy

EDA

extra domain A

EDB

extra domain B

FN1

fibronectin type I

FN2

fibronectin type II

FN3

fibronectin type III

GuHCl

guanidine hydrochloride

HMQC

heteronuclear multiple-quantum correlation

HSQC

heteronuclear single-quantum correlation

NMR

nuclear magnetic resonance

NOE

nuclear Overhauser effect

NOESY

nuclear Overhauser effect spectroscopy

PBS

phosphate-buffered saline

TOCSY

total correlation spectroscopy