Propensity for C-terminal domain swapping correlates with increased regional flexibility in the C-terminus of RNase A


  • Katherine H. Miller,

    1. Biophysics Graduate Group, University of California, Berkeley, California 94720
    2. Institute for Quantitative Biosciences-Berkeley, University of California, Berkeley, California 94720-3220
    Search for more papers by this author
  • Susan Marqusee

    Corresponding author
    1. Biophysics Graduate Group, University of California, Berkeley, California 94720
    2. Institute for Quantitative Biosciences-Berkeley, University of California, Berkeley, California 94720-3220
    3. Department of Molecular and Cell Biology, University of California, Berkeley, California 94720
    • Biophysics Graduate Group, University of California, Berkeley, CA 94720
    Search for more papers by this author


Domain swapping is a type of oligomerization in which monomeric proteins exchange a structural element, resulting in oligomers whose subunits recapitulate the native, monomeric fold. It has been implicated as a potential mechanism for protein aggregation, which provides a strong impetus to understand the structural determinants and folding mechanisms that trigger domain swapping. Bovine pancreatic ribonuclease A (RNase A) is a well-studied protein known to domain swap under extreme conditions, such as lyophilization from acetic acid. The major domain-swapped dimer form of RNase A exchanges a β-strand at its C-terminus to form a C-terminal domain-swapped dimer. To study the mechanism by which C-terminal swapping occurs, we used a variant of RNase A containing a P114G mutation that readily domain swaps under physiological conditions. Using NMR and hydrogen–deuterium exchange, we find that the P114G variant has decreased protection from hydrogen exchange compared to the wild-type protein near the C-terminal hinge region. Our results suggest that domain swapping occurs via a local high-energy fluctuation at the C-terminus.


Domain swapping is a unique mechanism of protein dimerization and higher order oligomerization that results in the formation of subunits that recapitulate native, monomeric structure. It has been implicated as a means for evolution of quaternary structure and modification of enzymatic activity. It is most often studied in terms of its potential as a mechanism for protein aggregation and amyloid formation as both processes are self-complementing and often involve polar zipper-like interactions.1 For proteins that populate the dimer at equilibrium, there must be favorable enthalpic interactions stabilizing the dimer, most likely in the open interface.

In spite of its implications for protein evolution and aggregation, very little is known about the driving forces that cause proteins to domain swap. Although there are over 50 domain-swapped structures in the PDB,2 there are only a handful of investigations into the folding mechanisms that lead to domain swapping.3–9 Among these studies, there is debate as to whether domain swapping occurs via the unfolded state or some high energy, partially unfolded state. In the first case, domain swapping serves as an alternative folding pathway9; in the latter, the monomer may undergo a fluctuation to some non-native, partially unfolded form, M*, that is capable of swapping and forming a domain-swapped dimer (Fig. 1).

Figure 1.

Structural models for C-terminal domain swapping in RNase A. Monomeric RNase A (PDB ID 2AAS) may dimerize (PDB ID 1F0V) either via complete unfolding or some partial unfolding event that produces a hypothetical exchange-competent intermediate, M*.

Bovine pancreatic ribonuclease A (RNase A) is an example of a protein that exists primarily in its monomeric form but is capable of domain swapping. RNase A can swap at either its N- or C-termini to form dimers or higher order oligomers. To form domain-swapped structures, the protein is subjected to lyophilization from 50% acetic acid. Under these conditions, 80% of the dimers are C-terminal domain-swapped dimers, while 20% are N-terminal domain-swapped dimers. Since this process requires lyophilization, in which the protein goes through the solid state, it has been difficult to study the process in solution.3, 8 Thus, the mechanism of domain swapping in RNase A under more physiological conditions, particularly at the C-terminus, remains poorly understood.

Previous experiments have examined the role of the N-terminal hinge region on domain swapping in RNase A and related RNases, bovine seminal RNase (BS-RNase) and human pancreatic RNase (HP-RNase). While WT-RNase A does not domain swap under physiological conditions, BS-RNase readily forms N-terminal domain-swapped dimers in solution,10, 11 most likely due to differences in the hinge regions. The exact role of these sequence differences in affecting the swapping propensity remains unclear, as RNase A still does not readily domain swap even when its N-terminal hinge region residues are replaced with those of BS-RNase.10 Deletion of this hinge, however, can drive HP-RNase, a monomeric protein, to form N-terminal domain-swapped dimers.11 It is worth noting that the exchanged N-terminal region in these N-terminal-swapped dimers is the same helix that is cleaved by subtilisin in RNase A to form RNase S, demonstrating that, as seen in domain swapping, the two regions can associate noncovalently.12

The structural determinants that lead to the C-terminal dimer formation are both less obvious and potentially more interesting than those of the N-terminal dimer. There are no existing data to suggest that wild-type RNase A populates some partially unfolded intermediate in which the C-terminal β-strand is exposed and prone to exchange. The C-terminal β-sheet is one of the first elements of RNase A to fold.13 It is highly protected from hydrogen exchange and is protected from carboxypeptidase A cleavage in its thermally unfolded state.14–16 In fact, Laurents and coworkers3 have suggested, based on their experimental studies on the folding of RNase A, that the protein domain swaps via a previously identified folding intermediate, IN, and therefore swapping involves accessing the unfolded state. Computational studies have also suggested that significant unfolding is required for swapping from the C-terminus.6 Our recent results,17 however, demonstrate that under physiological conditions, an increase in the population of the unfolded state does not increase the propensity for domain swapping.

Resolving the mechanism of C-terminal dimer formation in RNase A is of particular interest as it applies to the link between domain swapping and amyloid formation. The open interface in the C-terminal dimer contains two stacked β-strands, reminiscent of the polar–zipper interactions present in amyloid-β spines, suggesting that the mechanism that leads to C-terminal dimer formation may be more similar to events that result in aggregation.18 In fact, an engineered version of RNase A with an amyloidogenic polyglutamine sequence in the C-terminal hinge region was the first well-defined example implicating domain swapping in amyloid formation.19

Recently, we showed that a cis-proline in the C-terminal hinge region (P114) acts as a conformational gatekeeper for domain swapping in RNase A.17 This result strongly suggests that isomerization at P114 is a key event in the mechanism of RNase A C-terminal domain swapping. Mutation of P114 to an amino acid that strongly prefers a trans conformation, such as glycine or alanine, resulted in variants of RNase A that readily domain swap under physiological conditions. These RNase A variants provide an excellent system for studying the mechanism of domain swapping under experimentally tractable conditions.

Our hypothesis was that differences in the dynamics of P114G and WT-RNase A might correlate with the change in swapping propensity of the P114G variant and elucidate some partially unfolded, exchange-competent intermediate state. Rather than directly comparing the P114G variant to WT, we used another variant, P93A, that has the same global stability as P114G but does not readily domain swap under physiological conditions.17 To probe for differences between these variants, we implemented NMR hydrogen–deuterium exchange (HX) techniques.

Increased HX rates in specific regions of a protein are interpreted as reflecting either an increased population or easier access to some “open” or exposed state for that amide site via the following standard model as described by Linderstrom-Lang20:

equation image

The more flexible a site in the protein is, either due to local fluctuations, partial or global unfolding events, the faster it will exchange. By comparing the HX behavior of the two RNase A variants, P114G and P93A, under identical conditions, we can make inferences about how local changes in conformational flexibility affect domain swapping propensity.


Protein stability under HX conditions

Two variants of BP-RNase A were evaluated for folding and stability using circular dichroism (CD) spectroscopy. In addition to the P114G or P93A mutation, both variants contained the H119A mutation. H119A is an active-site mutation that helps to avoid toxicity during expression in bacterial culture. The free energy of unfolding (ΔGunf) under conditions suitable for HX experiments (pHcorr = 6.7, 50 mM NaPhos, 25°C) was determined by chemical-induced equilibrium denaturation and fit using a two-state assumption and linear extrapolation.21 Under these conditions, the stability of the P93A variant is 6.8 ± 0.24 kcal/mol (Cm = 2.4 M, m-value = 2.85 kcal mol−1 m−1) and the stability of the P114G variant is 6.48 ± 0.14 kcal/mol (Cm =2.4 M, m-value = 2.65 kcal mol−1 m−1) (Fig. 2): both are within error of each other and similar to the previously reported value of 6.8 kcal/mol in protonated buffer at pH 8.0, 100 mM Tris at 25°C.17

NMR assignments of backbone amide protons

To determine site-specific HX rates, NMR assignments for the NH protons of each variant were needed. Direct comparisons and assignments based on the published HSQC spectra of RNase A were not feasible as the individual mutations at H119, P114, and P93 altered the HSQC spectra enough to prohibit unambiguous identification of the backbone amide peaks. Thus, HNCA spectra were collected so that correlations to carbon assignments could be used to identify amide crosspeaks. Starting with several peaks that could be unambiguously identified based on comparison with the spectrum of WT RNase A, the C-α peaks from the HNCA allowed us to bootstrap our way through the rest of the protein. This approach allowed us to confirm assignments for most (∼80%) of the amide proton peaks in the spectra, which were then used for the following analysis. The spectra for the two variants (P114G and P93A) overlaid fairly well, which facilitated comparison of the decay of the same site between each variant (Fig. 3). A list of peak assignments is included in Supporting Information Table S1.

Figure 2.

Denaturant melts for P114G and P93A variants of RNase A in deuterated 50 mM NaPhos pHcorr 6.7. The data for the P93A variant are shown as filled circles and the data for the P114G variant as open diamonds.

Figure 3.

Overlaid 1H-15N HSQC spectra of P114G (red) and P93A (black). A: Protonated buffer, pH 6.7, 25°C. B: Spectra taken ∼1 h after HX was initiated in deuterated buffer, pHcorr 6.7, 25°C.

Comparison of HX behavior between the RNase A variants P114G and P93A

HX experiments were carried out by monitoring the proton occupancy at individual sites for 48 h after initiation of exchange, with the initial time point taken ∼45 min after initiation of exchange (see Materials and Methods). Since the half-time for dimer formation under these conditions is ∼72 h (equilibrium dimer population 20%, 10% formation at 72 h, data not shown), data acquisition was terminated after 48 h to avoid complications due to presence of the dimer. Based on this experimental time window and our assumption of a detection limit of ∼15% proton occupancy, we estimate that the minimum and maximum experimentally accessible exchange rates are ∼4.8 × 10−4 min−1 and 1.5 × 10−2 min−1, respectively. The calculated exchange rates for the observable probes in our experiments agree with this assumption (Tables I and II). Under the conditions of our experiments, HX of wild-type RNase A has been shown to take place in the EX2 kinetic regime.15, 22 Therefore, if we assume that this holds true for our variants, our experimental time window corresponds to a measurable ΔGHX between 5.5 and 8.5 kcal/mol (pHcorr = 6.7, 25°C), depending on the kin of each specific residue.

Table I. HX Rate Constants (in min−1) for P114G and P93A Variants of RNase A in Deuterated 50 mM NaPhos, pH 6.7 or 7.6, at 25°C
ResiduepHcorr 6.7, 25°CpHcorr 7.6, 25°C
104.08E − 031.72E + 042.11E − 033.33E + 042.14E − 022.60E+041.10E − 025.05E + 04
111.03E − 031.51E + 05SlowSlow1.90E − 036.55E + 056.42E − 041.93E + 06
122.77E − 031.11E + 05SlowSlow1.91E − 035.54E + 059.04E − 041.17E + 06
303.80E − 042.72E + 055.00E − 042.06E + 052.49E − 033.28E + 052.99E − 032.73E + 05
46    SlowSlowSlowSlow
492.06E − 033.95E + 045.55E − 031.47E + 041.36E − 032.49E + 051.00E − 033.37E + 05
523.99E − 031.27E + 043.96E − 031.28E + 042.43E − 021.65E + 044.12E − 029.74E + 03
553.47E − 031.97E + 04SlowSlowSlowSlowSlowSlow
561.68E − 037.70E + 049.86E − 041.31E + 051.98E − 035.20E + 052.41E − 034.28E + 05
601.88E − 029.98E + 033.96E − 034.74E + 04  3.70E − 024.02E + 04
62    SlowSlow5.44E − 044.87E + 06
632.26E − 031.51E + 04SlowSlow5.12E − 045.29E + 05SlowSlow
655.16E − 035.63E + 041.85E − 031.57E + 051.24E − 021.86E + 051.24E − 021.86E + 05
69  6.50E − 032.14E + 044.49E − 022.46E + 042.95E − 023.75E + 04
724.30E − 031.41E + 054.95E − 041.23E + 066.40E − 037.54E + 053.11E − 031.55E + 06
981.76E − 034.77E + 04SlowSlow1.80E − 033.70E + 052.18E − 033.06E + 05
99  SlowSlow3.75E − 031.95E + 052.58E − 032.84E + 05
1002.48E − 034.45E + 047.58E − 041.46E + 052.54E − 033.44E + 052.05E − 034.27E + 05
107SlowSlowSlowSlow3.30E − 042.16E + 051.66E − 034.31E + 04
110  9.02E − 043.22E + 05  4.24E − 035.45E + 05
111  SlowSlow  1.54E − 033.78E + 05
116  SlowSlow  SlowSlow
1189.97E − 049.45E + 03SlowSlowSlowSlowSlowSlow
1192.26E − 036.23E + 04SlowSlow1.14E − 034.23E + 05SlowSlow
Table II. HX Rate Constants (in min−1) for P114G and P93A Variants of RNase A in Deuterated 50 mM NaPhos, pH 7.6, at 37°C
ResiduepHcorr 7.6, 37°C
11  1.81E − 022.03E + 05
12  2.12E − 021.47E + 05
302.89E − 021.05E + 053.54E − 026.82E + 04
461.18E − 021.86E + 053.13E − 035.59E + 05
491.73E − 027.02E + 049.04E − 031.10E + 05
543.11E − 031.03E + 052.55E − 039.96E + 04
555.61E − 033.58E + 053.51E − 034.55E + 05
562.57E − 021.49E + 052.37E − 021.29E + 05
572.37E − 032.04E + 052.12E − 031.81E + 05
62  2.47E − 033.16E + 06
631.24E − 028.10E + 042.54E − 033.14E + 05
72  2.11E − 026.73E + 05
73  2.60E − 031.14E + 06
741.64E − 021.90E + 052.69E − 039.20E + 05
751.74E − 025.16E + 052.99E − 032.39E + 06
791.08E − 023.47E + 052.27E − 031.31E + 06
982.37E − 021.04E + 053.07E − 026.41E + 04
991.31E − 022.07E + 052.44E − 028.84E + 04
1002.88E − 021.13E + 052.04E − 021.27E + 05
1062.66E − 032.75E + 051.90E − 033.16E + 05
1074.95E − 035.36E + 044.86E − 034.33E + 04
1082.16E − 031.32E + 059.39E − 042.40E + 05
1093.73E − 034.70E + 051.96E − 037.09E + 05
110  2.85E − 022.40E + 05
111  1.42E − 021.21E + 05
116  1.43E − 033.02E + 05
1183.12E − 038.92E + 041.78E − 031.24E + 05
1192.38E − 027.22E + 042.30E − 036.20E + 05

Initial HX experiments were carried out at 25°C, pHcorr = 6.7, 50 mM NaPhosphate. Under these conditions, a number of residues in both variants show little to no hydrogen exchange on the time scale of the experiment (Table I). This is not unexpected, given the time limitation of our experiment and the fact that many residues in RNase A have previously been shown to have protection factors greater than that expected by exchange through global unfolding, also known as superprotection.23 Since the C-terminal strand is among the very well-protected regions, we repeated these experiments at a slightly higher pH (pHcorr= 7.6) at both 25 and 37°C, which correspond to a larger kin, and therefore (under EX2 conditions) access to more stable protons. Under these conditions, exchange rates were obtained for the slow-exchanging protons (Table II). The measured exchange rates were converted to protection factors (P = kin/kobs) using published values for kin24 calculated using the program SPHERE.25

For most sites in RNase A, the calculated protection factors for the P114G and P93A variants differ by less than an order of magnitude of one another (Tables I and II, Fig. 4). Notably, however, four sites show measurable differences; for all four of these probes, the exchange in P114G is faster than that in P93A. These residues can be divided into two categories: (1) Amide protons 110, 111, and 116: these probes show measurable decay in P93A, but exchange too quickly to be measured in P114G (undetectable in the initial time point). (2) Amide proton 119: this probe shows measurable exchange in both variants, but has a protection factor that is an order of magnitude less in P114G than in P93A. Although at 25°C the NH of 119 is too slow in P93A but measurable in P114G, at 37°C, the exchange rate at residue 119 is measurable in both variants. Initial data at 25°C suggested one other site with a potential difference in the two variants. The amide proton of residue 63 is too slow to measure in P93A and shows measurable, albeit slow, exchange in P114G. Further studies carried out at 37°C, where the kinetics could be measured in both variants, suggest that the exchange rates are very similar, and this result was likely just a result of the limitations of our experimental time window.

Figure 4.

Protection Factor (PF) versus residue number for P114G and P93A variants of RNase A. All protection factors were measured in 50 mM NaPhos, pH 7.6. Data for P93A are shown in gray and P114G in blue. A: PF measured at 25°C. P114G residues 110, 111, and 116 are shown in red. The horizontal lines at log(PF) = 7 and log(PF) = 2.5 are upper and lower limits, respectively, for amides whose exchange was too slow or fast to measure. B: PF measured at 37°C.


The two RNase A variants, P114G and P93A, have the same global stability yet very different domain-swapping propensity

We compared the energetics and HX behavior of two variants of bovine pancreatic RNase A, P114G and P93A. In both variants, a native cis X-Pro bond is mutated to a residue that prefers a trans conformation; however, only mutation of P114 results in the population of domain-swapped dimers under physiological conditions. A crystallographic study of a P114G variant confirms that the 113–114 amide bond, in the hinge region near the C-terminal β-strand, adopts the trans conformation with this mutation.26 We hypothesized, therefore, that local conformational fluctuations may be the driving force behind the increased swapping propensity of the P114G variant, and that NMR-based HX could be used to probe for such dynamics.

The P93A variant was selected as a point of comparison in order to control for the effect of global stability on domain swapping propensity and conformational flexibility. Because the P114G and P93A mutations destabilize the protein equally, the difference in swapping propensity cannot be the result of a change in population of the unfolded state.17 Since the global stability of the two is also the same under the conditions of our HX conditions (deuterated buffer system, Fig. 2), any differences in their observed HX behavior can be interpreted as local changes in flexibility or dynamics that affect swapping propensity.

Differences in protection factors suggest differential local dynamics

As expected for two proteins that differ by only two amino acids and show the same global stability, the observed HX rates are similar for almost all of the amide protons measured. In addition, the overall HX profiles are similar to those found in previous studies on wild-type RNase A. All of the sites that we observe as too slow to measure in our experiment are known to have very high protection factors and/or exchange at an energy greater than or equal to that of global unfolding.14, 15 It is also worth noting that a recent HX study on the isolated C-terminal-swapped dimer shows similar protection factors as the monomer, so any differential effects seen here cannot be accounted for by the presence of a small amount of dimer.8

Four sites, however, show notable differences in HX protection factors between the two variants: the amide hydrogens of residues 110, 111, 116, and 119. All four are localized near the swapping strand in the C-terminal domain-swapped dimer (Fig. 5). For each of these sites, the protection factor is significantly lower in the P114G variant, suggesting an increase in accessibility of some open conformation with these sites available for exchange.

Figure 5.

Structure of RNase A P114G with HX fluctuation sites highlighted. The crystal structure of RNase A P114G (PDB ID 1KH8) is shown in gray. The C-terminal β-strand that is exchanged in the C-terminal dimer is shown in blue. All residues whose amide hydrogens were found to exchange at least an order of magnitude faster in P114G than the control P93A variant are shown in red, and cluster around the exchanged C-terminal arm.

Using the standard model as described above and the steady-state assumption, the rate of exchange for each hydrogen can be described by Eq. (1):

equation image(1)

There are two limiting mechanisms by which HX can occur in this model. In the EX2 limit, kcl >> kin and kobs = kinKop, and the observed rate constants can be interpreted as differences of the free energy of the “open” conformation at specific sites (ΔGop = −RTlnKop). In the other, EX1 limit, kin >> kcl, kobs = ku, and the observed rates of exchange can be interpreted as reporting on the kinetics of opening at a specific site.27

The observed HX rates in our study likely represent exchange in the EX2 regime. Studies on wild-type RNase A demonstrate that the protein remains in the EX2 regime at pH 6.5 up to 45°C.15, 22 Our pH-dependent data also suggest that we are working in the EX2 regime. Under the EX2 regime, the observed rate of exchange depends on kin, which in turn depends on temperature and pH, and therefore the pH dependence of HX rates is generally considered indicative that a site is exchanging in the EX2 regime. On an average, the rates of exchange we observe at pH 6.7 and 7.6 increase approximately 10-fold, as expected for EX2 behavior. Thus, our working assumption is that the differences in the exchange behavior of our variants represent thermodynamic differences that result in different populations of the “open” conformation.

When comparing protection factors in P114G and P93A, it is also important to consider potential differences in their ground, or closed, state. The crystal structure of P114G reveals small differences in hydrogen bonding patterns at the C-terminus compared to wild-type RNase A,26 which are localized to the area surrounding the hinge region, residues 110–116 (the hinge itself is only 112–115). We assume that these H-bond distances in the C-terminal region will be very similar in WT and P93A and that any affects due to the H119A variant would be the same in both proteins and thus not likely affect the differential HX behavior. The hydrogen bonds A109O–H119N and E111O–V116N are both longer in P114G than the WT protein, and, as expected, both amides (116 and 119) have lower protection factors in P114G than in P93A. Other hydrogen bonds in this region, however, show no differences in H-bond distances, such as the 116O–E111N, even though the protection factor for E111 is also reduced in P114G. In fact, in general there appear to be more hydrogen bonds in this region in P114G than in WT, such as C110O–N113N and E111O–G114N. Therefore, based on the crystal structures alone, one might expect this C-terminal hinge region to display less dynamic behavior in P114G, suggesting that the changes in protection factors reflect fluctuations from the native state and not just alterations in the ground-state structure.

Relationship to domain swapping

The lower protection factors in the C-terminus of P114G implicate the population of a high-energy open state that may contribute to domain swapping. An open state with an extended or unfolded C-terminus would be expected to show decreased protection factors (lower ΔGHX) for the entire C-terminal hinge and β-strand. In general, our data support this model. Unfortunately, many residues within this region exchange too fast in both proteins to serve as probes of differential behavior, and, with the exception of 118, all amides we can measure show decreased protection factors in the P114G variant. Residues within the hinge, 112–115, form an exposed loop in the native structure and therefore exchange too fast in both variants. Similarly, the final C-terminal residues (121–124) do not show strong HX protection, most likely due to fraying.15 Within the β-strand itself, residue 117 is a proline and lacks an amide hydrogen, so its HX behavior cannot be followed. Unfortunately, residue 120 could not be assigned in the P114G variant, so the HX behavior at this residue could not be evaluated. Within the C-terminal region, there are five residues whose amide hydrogen exchange rates can be followed: 110, 111, 116, 118, and 119.

Both of the residues preceding the hinge region (110, 111) show decreased protection in the P114G variant. Two out of the three amides in the C-terminal β-strand (116, 119) showed significantly less protection in P114G than in P93A. However, it is interesting that residue 118 and most of the β-sheet that this strand packs against are slow to exchange and do not show decreased protection factors. Some of these residues may be restricted by the presence of the nearby disulfides within the core of the protein (C58–C110).

Our results provide support for a mechanism in which P114G domain swaps under physiological conditions via some partially unfolded intermediate state, M*, rather than a globally unfolded state. Mutation of proline at residue 114 appears to allow access to this high-energy, partially unfolded intermediate, M*. Access of this intermediate is likely favored by a trans conformation at the 113–114 amide bond. Thus, the proline at residue 114 in WT RNase A may serve as a conformational gatekeeper by inhibiting access of this intermediate, and the WT protein does not domain swap under these conditions. It is unclear whether domain swapping under extreme conditions, such as lyophilization from acetic acid, follows this same pathway. In sum, our results provide a mechanistic explanation for how prolines could play a more general role as conformational gatekeepers in domain swapping by inhibiting local flexibility.

Materials and Methods

Plasmid construction and protein purification

A plasmid containing the gene for BP-RNase A was generously donated by D. Eisenberg (with permission from the Raines lab, University of Wisconsin). The WT BP-RNase A gene from this plasmid was PCR amplified and inserted into a pET22b(+) vector. Quikchange mutagenesis (Stratagene) was used to create the two variants of BP-RNase A: P114G/H119A and P93A/H119A.

All protein constructs were expressed in BL21-codon + cells on LB media. The BP-RNase A variants were isolated as previously described from inclusion bodies28 and subsequently purified via gel filtration chromatography using a Hi-Load Superdex-75 16/60 column on a GE Healthcare AKTA purification system in a 20 mM Tris pH 8.0, 0.5 M NaCl buffer.

Isotopically labeled versions of both variants were expressed using the method described by Marley et al.29 These samples were expressed by growing cells in minimal expression media M9 containing 13C-glucose and/or 15N-ammonium chloride.

Protein stability studies

The global stability of the P114G/H119A and P93A/H119A variants was determined using guanidinium chloride denaturation monitored by CD spectroscopy in deuterated 50 mM NaPhos buffer, pHcorr = 6.7. The deuterated GdmCl was prepared by repeat (2×) lyophilization from D2O. Samples were prepared in aqueous and high denaturant buffers and allowed to equilibrate overnight. Data were collected on an Aviv 410 spectrometer using a 1-cm pathlength cuvette. The concentration of denaturant was adjusted using a Microlab 500 series titrator, and each sample was allowed to equilibrate for 10 min between shots before measurement was taken by averaging data at 222 nm for 60 s at 25°C. The data were fit to determine the global stability of unfolding, ΔGunf, using a fixed m-value of 3.0, under the assumptions that the system is two state and that ΔGunf has a linear dependence on denaturant concentration.21

NMR spectroscopy and hydrogen exchange

All NMR experiments were performed on a Bruker 800-MHz spectrometer. Spectra were processed using the NMRPipe/NMRDraw package,30 and spectra were analyzed and integrated using CARA.31 To attain accurate assignments, HNCA spectra were collected on 13C/15N-labeled versions of each variant in 50 mM NaPhos, pH 6.7, at 25°C. Assignments from the HNCA data were then used to unambiguously identify most peaks in HSQC spectra used for hydrogen exchange experiments.

For each sample, hydrogen exchange was initiated by 1:30 dilution of the sample into deuterated buffer, followed by concentration at 25°C. Time points for exchange were marked as the end of each NMR experiment, and for each data set the earliest time point after initial exchange via dilution and collection of the first NMR experiment was usually 45–60 min. For data sets collected at 25°C, tandem experiments, each 75 min, were run for ∼48 h without removal of the sample from the magnet. For experiments at 37°C, time zero was marked as the time when the sample started incubation at 37°C, since only residues that showed no decay at 25°C were being evaluated, and data collection consisted of tandem experiments, each 37 min, for up to 24 h.

Hydrogen exchange data were extracted from integration of peaks in 2D 1H-15N HSQC spectra, using the batch integration function in CARA. These data were then fit in Sigmaplot using a three-parameter, single exponential equation: y = yo + a*exp(−bt).


The authors thank Jeff Pelton for invaluable assistance with NMR experiments and Rachel Bernstein for helpful discussions. This work was supported by a grant from the National Institutes of Health GM 50945 to SM.