Quinary interactions with an unfolded state ensemble

Abstract

Anfinsen's thermodynamic hypothesis states that the native three‐dimensional fold of a protein represents the structure with the lowest Gibbs free energy. Changes in the free energy of denaturation can arise from changes to the folded state, the unfolded state, or both. It has been recently recognized that quinary interactions, transient contacts that take place only in cells, can modulate protein stability through interactions involving the folded state. Here we show that the cellular environment can also remodel the unfolded state ensemble.

Introduction

Protein folding has fascinated chemists for over a century.1, 2, 3, 4, 5 In 1961, Anfinsen proposed that the tertiary structure of a protein is encoded entirely by its amino acid sequence. His thermodynamic hypothesis states that “the three‐dimensional structure of a native protein in its normal physiological milieu… is the one in which the Gibbs free energy of the whole system is lowest.”6 The Gibbs free energy of unfolding ( $\Delta G_{\mathrm{U}}^{\mathrm{o}\prime }$) is the free energy of the unfolded ensemble (U) minus that of the folded state (F)

$$\Delta G_{\mathrm{U}}^{°\prime }=G_{\mathrm{U}}^{°\prime }-F_{\mathrm{F}}^{°\prime }$$

The equilibrium thermodynamics of protein stability has been studied extensively in buffer. Although such experiments yield valuable information, simple solutions do not capture the effects of quinary interactions,7, 8, 9, 10, 11, 12 the transient contacts between macromolecules that organize the cellular interior. Repulsive interactions stabilize proteins by minimizing the available volume, favoring the smaller folded state.13 Nonspecific attractive interactions and attractive interactions with the protein backbone favor the unfolded state because it exposes more reactive surface. The relative strengths of attractive and repulsive contacts and the intrinsic properties of the test protein determine the net effect of quinary interactions on protein stability: some proteins are stabilized in cells relative to buffer,14, 15, 16 others are destabilized.16, 17, 18, 19, 20

We chose the B1 domain of protein G (GB1, 6.2 kDa, pI 4.5)21 as a model system for understanding the effects of quinary interactions. GB1 undergoes a two‐state folding equilibrium22 and has been extensively characterized in buffer18, 19, 21, 22, 23, 24, 25, 26, 27 and cells.12, 18, 19, 28, 29, 30 Quinary interactions do not change the structure of folded GB118 and have a negligible effect on its stability at pH 7.4.12 Importantly, GB1 is foreign to Escherichia coli, minimizing the risk of specific interactions, which would complicate interpretations. We refer to the T2Q;K10H variant as the wild‐type protein, because the T2Q mutation prevents N‐terminal degradation,25 and the K10H mutation provides an intracellular pH probe.31

We recently used this system to demonstrate that electrostatic interactions involving surface residues of the folded state contribute to protein quinary interactions.11, 12 Here, we focus on quinary interactions involving the unfolded state ensemble. Direct characterization of the ensemble is not practical because only about one in 10⁵ GB1 molecules is unfolded under native conditions. However, we can gain information by studying GB1 stability in cells and in buffer with amide proton exchange.32

NMR‐detected amide proton exchange33, 34, 35 is a powerful method for quantifying equilibrium protein stability. For globular proteins, each backbone amide proton takes part in an equilibrium between the closed, folded state and the open, exposed state. In D₂O, amide protons in the open state exchange for deuterons

$$\text{closed}-\text{NH}\underset{k_{\text{cl}}}{\overset{k_{\text{op}}}{\rightleftharpoons }}\text{open}-\text{NH}\stackrel{k_{\mathrm{int}}}{\to }\text{open}-\text{ND}$$

where k _op and k _cl are the rates of opening and closing, respectively, and the intrinsic exchange rate, k _int, is the rate of exchange in an unstructured peptide. The reaction is irreversible because the experiment is performed in D₂O. In buffer, hydrogen exchange is monitored by acquiring serial 15N–¹H heteronuclear single quantum coherence spectra (HSQC), and the temporal decrease in crosspeak volume is used to quantify stability. In‐cell exchange data, however, must be acquired in lysates due to the poor quality of in‐cell NMR spectra.18 This procedure requires acquisition of discrete time points, which means that amide proton exchange must be quenched so the exchange does not occur between cell lysis and spectral acquisition. Lowering the pH of the lysate to 3.5 quenches the reaction, because amide proton exchange is base catalyzed.34, 35 Control experiments show that serial acquisition and the quenched lysate procedure yield the same result.18

Under our conditions, k _int is rate‐determining for GB118, 23 and the observed rate of exchange, k _obs, is proportional to the equilibrium constant of opening, K _{op .}

$$k_{\text{obs}}=\frac{k_{\text{op}}}{k_{\text{cl}}}{k}_{\mathrm{int}}=K_{\text{op}}{k}_{\mathrm{int}}$$

K _op can be used to quantify local protein stability,

$$\Delta G_{\text{op}}^{°\prime }=-RT\hspace{0.17em}\text{ln}\hspace{0.17em}{K}_{\text{op}}$$

where R is the universal gas constant, and T is the absolute temperature, which is 310 K for the work described here. Exchange rates were converted to free energies using k _int values from SPHERE (310 K, alanine oligopeptide basis, pH 7.4).36 Importantly, k _int is the same in cells and buffer.37

For any particular protein, individual residues that become exposed only upon global unfolding may not yield identical $\Delta G_{\text{op}}^{°\prime }$ values because the intrinsic rates of exchange are derived from model peptides, not the protein being studied. Importantly, residues exchanging by global unfolding yield $\Delta G_{\text{op}}^{°\prime }$ values within 1 kcal/mol of $\Delta G_{\mathrm{U}}^{°\prime }$ obtained from thermal denaturation.38 Residues that yield $\Delta G_{\text{op}}^{°\prime }$ values under the conditions described here are known to exchange only upon complete GB1 unfolding.18, 26 For these reasons, we report the mean $\Delta G_{\text{op}}^{°\prime }$ values as $\Delta G_{\mathrm{U}}^{°\prime }$, along with the standard deviation of the mean.

GB1 is one of only a few proteins whose unfolded ensemble has been studied. Specifically, examination of GB1 peptides shows that a non‐native hydrophobic staple,39 an α‐helical motif formed between two hydrophobic side chains at positions N′ and N4, is formed between V21 (N′) and A26 (N4, Supporting Information Fig. S1).40 The V21 side chain forms a stereospecific interaction with the methyl group of A26, and the staple is strengthened by a hydrogen bond between the amide proton of T25 and a side chain oxygen of D22.40 The staple is absent in the folded structure, but it stabilizes the unfolded state via a reverse hydrophobic effect because the hydrophobic valine is less exposed in the unfolded state than the folded state.39, 41

To assess whether quinary interactions affect the unfolded ensemble, we quantified the stability of GB1 and three variants. The V21A and V21I variants are expected to disrupt the staple, and thereby stabilize GB1, by removing a hydrophobic interaction and introducing a steric clash, respectively. The V21T mutation, which maintains the geometry critical to the hydrophobic interaction, is expected to stabilize the staple by introducing a hydrogen bond between the threonine side chain oxygen and amide proton of A26, thus destabilizing GB1.

The $\Delta G_{\text{op}}^{°\prime }$ values for individual residues in buffer are shown in Fig. 1. The values are well distributed across the secondary structure: β1 (Y3, K4), α1 (A26, K28, V29, K31, A34), β3 (T44, D46), and β4 (T51, F52, T53). Crosspeak overlap and large k _obs values, however, limit the number of quantifiable residues, especially in the β2 region.26

Figure 1.

Stabilities in buffer. $\Delta G_{\text{op}}^{°\prime }$ values for the wild‐type (green), V21A (red), and V21I (blue) proteins. V21T (purple) was not stable enough to quantify, and the arrow indicates that the $\Delta G_{\text{op}}^{°\prime }$ is less than the least stable quantifiable residue (V29) in the wild‐type protein. The dashed lines indicate the mean $\Delta G_{\text{op}}^{°\prime }$ value, which equals $\Delta G_{\mathrm{U}}^{\mathrm{o}\prime }$, for each variant. $\Delta G_{\mathrm{U}}^{\mathrm{o}\prime }$ values are shown on the right with the standard deviation of the mean. $\Delta G_{\text{op}}^{°\prime }$ values were quantified as described.18, 35

We first examined GB1 and its variants in buffer (Fig. 1). The unfolding free energy for wild‐type GB1 in 75 mM 4‐(2‐hydroxyethyl)piperazine‐1‐ethanesulfonic acid /75 mM bis–tris propane/75 mM citrate, pH 7.4 at 310 K has been reported (6.80 ± 0.07 kcal/mol).31 As expected, disrupting the staple by introducing V21A and V21I stabilizes GB1, yielding $\Delta G_{\mathrm{U}}^{°\prime }$ values of 8.92 ± 0.08 and 8.86 ± 0.08 kcal/mol, respectively. Also as expected, reinforcing the staple by adding a potential hydrogen bond (V21T) destabilizes GB1. In fact, the V21T variant was too destabilized to quantify, but we know its stability is less than 6.4 kcal/mol, which is the smallest quantifiable value under these conditions. The limit is determined by the largest k _obs we can quantify and the intrinsic properties of the protein (i.e., specific k _int values).18 As discussed below, these results are consistent with the idea that, in buffer, the unfolded state of GB1 contains the hydrophobic staple.

Each of the four proteins is less stable in cells than in buffer (Fig. 2). This general destabilization is expected because attractive quinary interactions are present in cells.16, 18, 19, 20, 29, 30, 31, 42, 43, 44, 45, 46 Wild‐type GB1 has a stability of 6.6 ± 0.1 kcal/mol in cells. The V21A and V21I variants have $\Delta G_{\mathrm{U}}^{°\prime }$ values of 7.0 ± 0.2 and 7.33 ± 0.05 kcal/mol, respectively. The V21T variant was again too unstable for quantification, but its stability must be less than 6.1 kcal/mol, the smallest value we could measure in cells under these conditions.12 As discussed next, these results suggest that quinary interactions with the unfolded state ensemble can modulate protein stability.

Figure 2.

Stabilities in cells. $\Delta G_{\text{op}}^{°\prime }$ values for wild‐type (green), V21A (red), and V21I (blue) in cells. The dashed lines indicate the mean $\Delta G_{\text{op}}^{°\prime }$ value, which equals $\Delta G_{\mathrm{U}}^{\mathrm{o}\prime }$. $\Delta G_{\mathrm{U}}^{\mathrm{o}\prime }$ values are shown on the right with the standard deviation of the mean. V21T (purple) was too destabilized to quantify. The arrows indicate that the $\Delta G_{\text{op}}^{°\prime }$ values are less than the least stable residue (V29) in the wild‐type protein. $\Delta G_{\text{op}}^{°\prime }$ values were quantified as described in the Supporting Information.

Changing surface residues that are not involved in secondary structure should not affect stability.47 The surface residue V21 lies in an exposed loop, and its amide does not participate in intramolecular hydrogen bonds in the folded state of GB1.21, 26 Contrary to this expectation, but consistent with disruption of the hydrophobic staple in the unfolded ensemble, changing position 21 changes GB1 stability in buffer (Fig. 3). The mutations V21A, which removes two methyl groups, and V21I, which adds a methyl group, stabilize GB1 in buffer by 2.11 ± 0.04 and 2.06 ± 0.04 kcal/mol, respectively. In cells, the V21A and V21I mutations increase the stability by only 0.4 ± 0.2 and 0.7 ± 0.1 kcal/mol, respectively, suggesting that the staple is absent from the denatured ensemble in cells.

Figure 3.

Mutations that increase GB1 stability in buffer have a small effect in cells. Changes in opening free energies, $\Delta \Delta G_{\text{op}}^{°\prime }$ ( $\Delta G_{\text{op},\text{var}}^{°\prime }-\Delta G_{\text{op},\text{wt}}^{°\prime }$) for GB1 V21A (top panel) and GB1 V21I (bottom panel) in buffer (blue) and in cells (green). Dashed lines indicate the $\Delta \Delta G_{\mathrm{U}}^{°\prime }$ value derived from the mean $\Delta \Delta G_{\text{op}}^{°\prime }$ values and their standard deviation.

As predicted, the V21T mutation appears to strengthen the staple, because it decreases GB1 stability in buffer (Fig. 1). The mutation destabilizes GB1 to such an extent that we cannot quantify its stability, but given our limit of detection, the decrease must be more than 0.8 kcal/mol.18 We hypothesize that the V21T change stabilizes the unfolded ensemble because it maintains the stereochemistry and strengthens the staple by contributing an additional hydrogen bond.

The D40K variant19 serves as a control to show that other surface mutations do not act in this manner. Changing this surface exposed residue has a minimal effect in buffer (ΔΔG°′ = 0.1 ± 0.1 kcal/mol) but a large effect in cells (–1.34 ± 0.07 kcal/mol), because of increased charge–charge interactions with the native state.12

We also considered the possibility that the amino acid changes altered the folded state, but analysis of 15N–¹H HSQC spectra shows that the mutations have a negligible effect on the spectra (Supporting Information Fig. S2). Furthermore, the structure of a folded globular protein will remain the same in buffer and in cells.11 In addition, the patterns in the opening free energies along the primary structure are similar for all proteins in buffer and in cells. All these observations suggest that the mutations do not change the tertiary structure of GB1.

However, there is precedence for the idea that amino acid changes can affect unfolded states. For instance, mutations that have no effect on native contacts in staphylococcal nuclease change the stability of the unfolded state ensemble by impacting non‐native interactions,48, 49, 50 and introducing a disulfide bond into cytochrome c stabilizes a compact unfolded state, destabilizing the protein.51

The key point is that the effects from changing the staple are much less impactful in cells than in buffer (Fig. 4). The V21A and V21I mutations disrupt the staple, whose strength depends on the stereospecific interaction between V21 and A26, and stabilize GB1 in buffer by more than 2 kcal/mol. However, they have a modest impact in cells, increasing GB1 stability by 0.4 and 0.7 kcal/mol, respectively. These data suggest that quinary interactions with the unfolded state ensemble attenuate the stability changes observed in buffer by disrupting the hydrophobic staple. In other words, these interactions in cells mitigate the effect of the hydrophobic staple in buffer by remodeling the unfolded ensemble. Consistent with this idea, Guzman et al. showed that the cellular environment has opposing effects on the stability of two proteins with similar sizes and surface properties: one is stabilized while the other is destabilized.16 The authors suggest that quinary interactions with the unfolded ensemble may be responsible for the differences in stabilization.

Figure 4.

Free energy diagram of the relative free energies for the folded (solid line) and unfolded states (dashed lines) in buffer (blue) and in cells (green). Folded states are arbitrarily assigned a common value.

The interactions that disrupt the staple in cells may involve interactions between exposed hydrophobic residues and the cellular proteins. Such interactions were shown to exist on the native state of ubiquitin.30 However, charge–charge and hydrogen bonding interactions must also be considered.12, 29, 31, 52

The cellular interior is a dynamic environment teeming with hundreds of thousands of different macromolecules. Quinary structure imposes the organization essential for governing the myriad of biochemical reactions that sustain life. While most biochemistry is studied in dilute, buffered solutions, we now know that quinary structure modulates protein stability through interactions with the folded state.12, 18, 19, 20, 29, 30, 31 Our observations here indicate that quinary interactions may influence the ensemble of unfolded states, reinforcing the idea that data from experiments in dilute solution yield important fundamental information about proteins, but may not capture the impact of chemical interactions that occur only in the dynamic, heterogeneous environment of the cell. The source of these interactions can only be uncovered when proteins are studied in their physiological environment.7, 8, 9, 11 This knowledge will yield insight into the energetics that control protein chemistry in cells, providing new possibilities for understanding the physiological roles of chaperones, the proteasome and disease‐causing protein mutations, particularly aggregation‐based diseases, as they all involve the properties of unfolded protein ensembles in cells.53, 54, 55

Materials and Methods

The pET11a plasmids harboring the T2Q and T2Q;K10H GB1 gene have been described.25, 31 Mutations were installed using a Quikchange kit (Agilent) with the following primers and their reverse complements (mutation bolded): V21A, 5′‐CC ACC GAA GCT GCT GAC GCT GCT ACC GCG‐3′; V21I, 5′‐CC ACC GAA GCT ATT GAC GCT GCT ACC GCG‐3′; and V21T, 5′‐CC ACC GAA GCT ATC GAC GCT GCT ACC GCG‐3′. Mutations were confirmed by DNA sequence analysis (Eton Bioscience, Research Triangle Park, NC).56

Proteins were expressed in E. coli BL21(DE3) cells and purified as described.18, 21 For experiments in dilute solution, 2.2 mg of purified protein was resuspended in 500 μL of 75 mM bis–tris propane/75 mM 4‐(2‐hydroxyethyl)piperazine‐1‐ethanesulfonic acid/75 mM citrate, 99.9% D₂O, pH 7.4, and 15N–¹H HSQC spectra were acquired serially. pH meter readings are uncorrected for the deuterium isotope effect.57

Conflicts of interest

The authors declare no competing interests.

Supporting information

Supporting Information