We have used several biophysical techniques, including CD, fluorescence, DLS and NMR, to characterize the equilibrium unfolding behavior of wt A4 and two dissociable variants containing C321S and C321S/F283L mutations, respectively. The wt A4 dimer () undergoes large-scale unfolding with a concomitant loss of helical secondary structure, solvent-exposure of aromatic side chains and increase in hydrodynamic radius in a sharp transition centered around 4.8 M GuHCl (). This exceptionally high stability toward solvent denaturation is not surprising, given that the A4 dimer contains a total of seven disulfide bonds (three within each monomer and one between the subunits). However, further analysis using global fitting techniques reveals a second, subglobal, unfolding transition at moderate denaturant concentrations, which is accompanied by a 25% decrease in the CD signal due to α-helical secondary structure, tyrosine fluorescence changes consistent with increased solvent exposure and a ~20% increase in hydrodynamic radius ( and , respectively). The combined unfolding data can be fitted on the basis of a unimolecular three-state mechanism consisting of two coupled unfolding transitions with free energy increments of 2.3 and 12.4 kcal mol− 1
, respectively. While most globular proteins of small to moderate size exhibit two-state unfolding transitions, accumulation of equilibrium intermediates at moderate denaturant concentration is not without precedent.15
Although we have only limited structural information, the presence of a partially unfolded state may well be related to the fact that the disulfide bond connecting the subunits of the A4 dimer is located at the tip of an exposed pair of loops;5
disruption of the non-covalent interactions between monomers would result in a pair of domains connected through a flexible linker segment. This scenario is consistent with the observed increase in Rh
, and the changes in optical parameters, which can be attributed to perturbations in the monomer structure upon disruption of the dimer interface.
With increasing denaturant concentration, the C321S variant, which lacks a covalent linkage, undergoes a series of transitions involving both unfolding and dissociation of the native dimer. By observing this second-order transition both as a function of denaturant and protein concentration, we can not only characterize the conformational equilibrium, but also determine the dissociation constant of the native dimer.17
Solvent-denaturation thus represents a valuable alternative to conventional methods, such as size-exclusion chromatography and fluorescence anisotropy, for measuring dissociation constants for homodimeric proteins and provides additional information on the energetically linked unfolding reactions. Our sedimentation equilibrium data () clearly show that in the absence of denaturant, the predominant form of C321S A4 is a dimer over the whole range of protein concentrations used in our biophysical experiments (~5 to 100 μM). At the other extreme of the unfolding transition (~6 M GuHCl), the protein appears to be fully dissociated into monomers, based on its reduced hydrodynamic radius compared to wt A4 (). This leaves us with three possible folding/oligomerization mechanisms in which intermediate states are either absent (Scheme 1
), dimeric (Scheme 2
) or monomeric (Scheme 3
), depending on the relative strengths of the intra- vs.
inter-subunit interactions. Among these, a fully cooperative unfolding/dimer dissociation mechanism without intermediates (Scheme 1
) can readily be ruled out, since it fails to explain the observed biphasic unfolding behavior and protein concentration-dependent changes in slope ( and ). The distinction between Schemes 2
and Scheme 3
is more subtle, since both mechanisms predict populated equilibrium intermediates. We initially favored Scheme 3
, since it resulted in somewhat lower χ2
values in our global fitting of the optical data (). The DLS measurements provided striking additional evidence in favor of Scheme 3
, since this is the only mechanism that can account for the observed decrease in Rh
at low to moderate denaturant concentrations, where the native dimer dissociates into folded monomers. In contrast, accumulation of a partially unfolded dimeric intermediate, according to Scheme 2
, would give rise to an increase in Rh
, similar to that observed for wt A4 ().
On the basis of this model, global analysis of the GuHCl unfolding data allows a reliable determination of the dimer dissociation constant. The value obtained for C321S A4, Kd
= 90 ± 50 nM, is consistent with the value of 50 ± 30 nM obtained independently using fluorescence anisotropy measurements vs.
protein concentration (). The Kd
of full length dissociable FXI was recently reported to be ~70 nM from AUC SE studies of FXI G326C.24
Since this value is comparable with that measured in the present study for the isolated A4 dimer using GuHCl induced unfolding and fluorescence spectroscopy, FXI A4 must contain most, if not all of the dimer interface surface area between FXI monomers. This conclusion is consistent with the crystal structure of the full length FXI zymogen, which shows that all of the residues responsible for dimer formation reside within FXI A4.5
Qualitatively, if we compare the population plot for C321S A4 () with that of wt A4 (), it is apparent that the C321 inter-subunit disulfide bond conveys a dramatic increase in both stability and cooperativity. For the dissociable variant, the fully unfolded form begins to appear around 1.5 M GuHCl and builds up over a wide range of denaturant concentrations extending above 6 M and is characterized by a relatively low m-value (0.6 kcal mol− 1
; ). The wt protein, on the other hand, shows no detectable unfolded population at concentrations <4 M GuHCl, followed by a very sharp transition at concentrations >4 M GuHCl. The corresponding m-value (2.57 kcal mol− 1
) indicates a major increase in solvent-accessible surface area, as expected for a globular protein of the size of the A4 dimer.25
However, a more quantitative comparison of the equilibrium parameters is complicated by the accumulation of intermediate states and the fact that the unfolded states of the two proteins represent different oligomeric states. In the case of C321S A4, the dimeric interactions are lost in the first transition from native dimer to the monomeric intermediate, while in wt A4 inter-subunit contacts are at least partially preserved during the first unfolding transition. These additional interactions and buried surface area stabilize the intermediate for the covalently linked dimer and render its unfolding transition more cooperative relative to the dissociable variant. In , we express the denaturant-dependence of dimer dissociation for C321S A4 in terms of an m-value (m1
= 1.2 kcal mol-1 M− 1
), which is related to the amount of surface area that is buried in the dimer and becomes exposed to the solvent upon dissociation. The m-value for the second transition, m2
= 0.6 kcal mol− 1
reflects the exposure upon unfolding of residues buried within each monomer. Thus, the overall increase in solvation is proportional to mtot
= 2.4 kcal mol−1
. In the case of wt A4, the m-value for complete unfolding is mtot
= 3.4.kcal mol−1
. This discrepancy cannot be readily attributed to the denatured state, since the cross-linked dimer is expected to retain more, not less, residual structure after unfolding. Therefore, the difference in total m-values must be attributed to structural differences of the folded states causing an increase in solvent-accessible surface area for C321S A4 relative to the wild type. A major contribution may come from the two disulfide-linked loops around C321, which are in close contact in the crystal structure of FXI,5
but were found to be more flexible in an NMR structure of the A4 dimer determined recently in our laboratory (Samuel, Cheng, Riley, Walsh and Roder, submitted for publication) Thus, the loops may undergo local unfolding upon disruption of the inter-subunit disulfide bond, giving rise to a 30% increase in solvent-accessible surface area.
Physiological concentrations of FXI in the blood plasma are ~62.5 nM (in terms of monomer), but the protein concentration of FXI during protein synthesis in the ER of liver cells is expected to be higher than secretion levels in the plasma. The contribution of the inter-monomer disulfide bond involving C321 toward stabilizing the folded dimer is expected to be primarily due to the loss of translational degrees of freedom of the cross-linked unfolded state compared to the C321S variant whose subunits can freely diffuse upon unfolding. However, a quantitative analysis of this effect is complicated by the fact that both wt A4 and C321S A4 exhibit multi-state unfolding transitions involving structurally distinct intermediates. The presence of an inter-subunit disulfide per se
is probably not essential for function, since rabbit FXI, which lacks this covalent bond, functions normally.26
On the other hand, the formation of stable noncovalent interactions between monomers appears to be essential for secretion of fully active FXI. This is underscored by our observation that the F283L mutation stabilizes the monomer at the expense of the dimer (), which is consistent with the previous findings that the mutation leads to elevated levels of retained monomeric protein inside cultured human kidney cells expressing FXI F283L relative to cells secreting wt FXI.10
These reports indicated that intermonomer disulfide bond formation occurred to a lesser extent in FXI F283L cells relative to wt FXI cells, strongly suggesting that the F283L mutation partially prevents formation of disulfide-linked homodimeric FXI protein in this cell culture model. While the in vivo
secretory pathway of FXI in liver cells is not well understood and the intracellular localization of disulfide bond formation is not known, two separate reports from cell culture studies suggest that intermonomer disulfide bond formation is important for secretion of FXI at normal levels.10
The F283L mutation results in a four-fold increase in Kd
relative to C321S A4 (350 nM vs.
90 nM), but does not inhibit formation of a native-like dimeric structure, as indicated by the close similarity between the 15
N HSQC NMR spectra of the two proteins. The effect of this conservative amino acid change on dimer stability is surprising, since the side chain of F283 is not directly involved in the A4 dimer interface, although the adjacent L284 side chain participates in a hydrophobic inter-subunit contacts.5
As illustrated by , F283 is part of a cluster of aromatic side chains, along with Y278, F311 and Y351, which make up the bulk of the hydrophobic core between the 5-stranded and the 2-stranded β-sheets of A4.5
A likely scenario for explaining the effect of the F283L mutation in shifting the equilibrium from the native dimer toward the monomeric intermediate is that the Phe to Leu amino acid change results in some side chain rearrangements within the core of A4 that are energetically favorable in terms of monomer stability, but cause unfavorable structural perturbations at the dimer interface. This scenario is consistent with the NMR chemical shift analysis shown in , which showed significant chemical shift changes not only for residues at or near the site of mutation, but also residues in the C-terminal half of the protein located on the opposite side of the aromatic cluster ().
Figure 10 Ribbon diagram of a monomer of wt A4, based on the crystal structure of FXI.5 Side chains are shown for residues involved in an aromatic cluster, some of which are altered in FXI patient mutations (F283L, Y351S, as well as the adjacent G350A/E).
Meijers et al. hypothesized that the F283L mutation in FXI affects protein folding and interferes with efficient secretion of FXI F283L.10
The FXI F283L mutant shows much lower secretion levels compared to wild type (<10%) and the enzymatic activity in plasma is <10% of normal.11
When purified, however, FXIa F283L exhibits normal enzymatic activity in a specific clotting assay.10
A Phe at position 283 is highly conserved in FXI A4 domains from various species. A Phe is also found in the corresponding position of two other apple domains of human FXI, A1 and A3, but is replaced by a Met in A2. Since these homologous domains are monomeric, we can conclude that an aromatic or long aliphatic side chain is critical for the stability of the monomeric state of the apple domain. This is consistent with the fact that the F283 side chain is partially buried within a cluster of other aromatic side chains ().
It is interesting that a second member of this aromatic cluster, Y351, is the site of another patient mutation (Y351S), and displays a similar phenotype as FXI F283L.27
In addition, two other patient mutations have been found to target G350, which precedes Y351 in the the C-terminal strand of the β-sheet (). The G350E mutation resulted in a very low secretion,14
whereas the G350A mutant is secreted at ~44% of normal levels, but showed low catalytic activity.28
Thus, 4 patient mutations, from a total of 14 known point mutations within the A4 domain 29
affect residues participating in or adjacent to the aromatic cluster, indicating that this is a critical structural feature, both in terms of monomer stability and maintaining a favorable interface for homodimer formation.
In conclusion, our biophysical evidence that the folding/unfolding equilibrium of dissociable (C321S) variants of A4 feature a well populated monomeric intermediate state has implications not only for understanding the folding of this dimeric protein, but also provides a rationale for understanding functional effects of certain mutations in patients with clotting deficiencies. In particular, the phenotype of the F283L mutation, which results in retention of momomeric FXI in the cytosol of liver cells, is consistent with our observation that the mutation results in a structurally more stable monomeric intermediate while destabilizing the native dimeric form. We hypothesize that the Phe to Leu substitution perturbs the side chain packing within a central aromatic cluster that results in energetic stabilization of the monomer at the expense of interactions between the subunits of the dimer.