A unique piezolyte mechanism of TMAO: Hydrophobic interactions under extreme pressure conditions Open Access

Around 16% of Earth’s species are found in the deep sea,¹ the potential birth place of life, where lack of light and increased pressure force these creatures to adapt. High pressure forces proteins to unfold and thereby disrupts vital functions, such as enzymatic reactions and sub-cellular transport.^2–8 It also destabilizes functional biocondensates, which are formed by liquid–liquid phase separation (LLPS), a process important for membrane-less compartmentalization in the cell.^9–13 The question is how are these organisms, called piezophiles because they have adapted to high pressure, able to counteract the high pressure conditions.

The adaptation can either occur through attuned piezophilic proteins or through the presence of small organic cosolutes referred to as osmolytes. It has been found that deep sea creatures have a high concentration of certain osmolytes, such as methylamines, in their cells that alongside of pressure increases linearly with depth.¹⁴ Based on this observation, these osmolytes, in particular trimethylamine N-oxide (TMAO), are thought to be specialized in their counteractions of pressure denaturation and have therefore been termed “piezolytes.”^14–16 TMAO has been found to increase the pressure stability of proteins and stabilize biomolecular condensates formed by LLPS under high pressure conditions,^9–13 but until now there are no indications that piezolytes are distinct from other osmolytes. In this regard, especially their effect on hydrophobic parts of proteins is of interest, since the hydrophobic interactions are important in LLPS, e.g., the elastine-like peptide (ELP), which exhibits pressure- and osmolyte-dependent LLPS, is made up of nearly 80% hydrophobic amino acids.¹² 

The effect of pressure on any chemical equilibrium is determined by the volume difference between reactants and products according to the Le-Chatelier’s principle. For example, in protein folding, the unfolded state has a smaller partial molar volume than the folded state, resulting in protein denaturation upon increasing pressure.⁴ On the other hand, the effect of cosolutes on a chemical equilibrium (e.g., protein folding or protein association in solution) is determined by the preferential binding of the cosolutes to the reactants and products.^17,18 Correspondingly, the dependence of the Gibbs energy, ΔG₂(p, c₃), of folding a protein on pressure, p, and cosolute concentration, c₃, at a fixed temperature, T, can be written as

where we used the standard convention and refer to water as component 1, protein as component 2, and the cosolute as component 3. In Eq. (1), ΔG₂ is the difference in Gibbs energy of the folded (F) and unfolded (U) states of the protein, $ΔV̄2=V̄2F−V̄2U$ the difference in partial molar volumes upon folding, $ΔΓ23=Γ23F−Γ23U$ the difference in cosolute preferential binding coefficient upon folding,¹⁷ R is the gas constant, and $a33=1+c3(G33−G31)−1$ an activity enhancement factor (⁠$>1$ for strong water-binding osmolytes), in which G₃₃ and G₃₁ are the Kirkwood–Buff integrals related to the components water and cosolute of the bulk solvent.

One hypothesis regarding how piezophiles have adapted to high pressure is based on the first part of the above expression: proteins from piezophiles have a smaller volume change upon unfolding, or the volume change upon unfolding is reduced due to the addition of piezolytes. However, the adaption of proteins has been disproven by Avagyan et al., who studied several proteins of piezophiles and non-piezophiles.¹⁹ Papini et al. measured the effects of protecting osmolytes and denaturants on $ΔV̄2$ of bovine ribonuclease and egg white lysozyme with differential scanning calorimetry and pressure perturbation calorimetry.²⁰ However, these authors found that there was nothing special about osmolytes that have been termed piezolytes, as all protecting osmolytes considered in their work had no direct effect on $ΔV̄2$ and the pressure denaturation of proteins. They therefore dismissed the term “piezolyte.” We note that, because $ΔV̄2$ is independent of c₃,²⁰ Eq. (1) implies that RTΔΓ₂₃a₃₃/c₃ does not depend on pressure.

Although TMAO cannot be distinguished as a piezolyte in the pressure stabilization of proteins, its hydration properties are pressure-dependent and, by contrast to other cosolutes, such as urea,²¹ determined by changes in electronic polarization at high pressure.^16,22,23 TMAO has a large dipole moment, which increases by nearly a factor of two when transferred from the gas phase to water.²³ Due to its zwitterionic nature, it is able to tightly bind water molecules with its negatively charged oxygen.²⁴ This property leads to the depletion of TMAO from many protein surfaces, with a corresponding increase in their thermal stability²⁵ and an indirect effect on pressure stabilization.²⁰ At ambient pressure, TMAO binds around three water molecules. At high hydrostatic pressure, the compression of the solvent, however, leads to electronic polarization effects that further enhance the dipole moment of TMAO.²³ Interestingly, at 10 kbar, it has been found that TMAO increases its water binding to a partial fourfold coordination of the oxygen atom.¹⁶ Although this change in TMAO–water interaction has an insignificant direct effect on the pressure stability of proteins,²⁰ it potentially affects hydrophobic interactions²⁶ and may therefore play a role in biomolecular LLPS and the pressure stability of functional biomolecular condensates, which are more susceptible to pressure than the folding of proteins.^11,12

To explore this question, we herein report a computer simulation study combined with free-energy calculations that quantify the effect of pressure-induced changes in TMAO polarization on the hydrophobic interaction of two nonpolar polyalanine-α-helices. All electrostatic interactions involving the polypeptide were removed from the model by setting the partial atomic charges on the solutes equal to zero. We therefore consider model hydrophobic solutes, not real biomolecules. This model system was chosen in order to isolate the potential role of piezolytic effects on hydrophobic interactions between solutes with extended nonpolar surfaces. The calculations reported below revealed that TMAO counterbalanced the pressure destabilization of hydrophobic interactions due to its enhanced dipole polarization at high pressure. We found that this pressure-stabilizing effect disappeared when ignoring the polarization of the TMAO dipole and found indications that this newly discovered effect was controlled by thermodynamic driving forces that should be generic for hydrophobic interactions in TMAO/water solutions under extreme pressure conditions.

We used the nonpolarizable, pressure-dependent TMAO force field (FF) of Hölzl et al.,²³ which, among other properties, reproduces the density of TMAO–water solutions at ambient and high pressures. This force field assigns adjustable partial charges to TMAO, which models the increased dipole polarization at high pressure. To explore the implications of this polarization effect on the hydrophobic interaction of the two nonpolar α-helices under high-pressure conditions (2 kbar), we performed umbrella sampling molecular dynamics (MD) simulations using two different approaches: (1) MD simulations were performed that used the polarized TMAO model with an increased dipole moment, from hereon called high-dipole FF, (2) MD simulations were performed that used the non-polarized TMAO model, parameterized for ambient pressure conditions, from hereon called low-dipole FF. We further calculated the difference in the partial molar volumes in the associated and dissociated states of the two α-helices and the preferential binding coefficient, Γ₂₃, of the single helix with these two approaches. The technical details on the force fields and umbrella sampling simulations are provided in the supplementary material.

Ambient pressure. Figure 1(a) reports the Gibbs energy of association, w(r_min), corresponding to the minimum of the potential of mean force (PMF) obtained with umbrella sampling (see supplementary material Fig. S1 for the PMF), as a function of the TMAO concentration (c₃) in solution. These data are in qualitative agreement with previously reported calculations based on the free-energy perturbation method.²⁶ The Gibbs energy of association displays a non-monotonic dependence on c₃ and has a minimum at c₃ = 1M, where the hydrophobic association is strongest. As was previously reported,²⁶ preferential TMAO binding gives rise to two opposing thermodynamic driving forces that lead to such a non-monotonic dependence: An entropy loss associated with the accumulation of excess TMAO in the hydration shells of the two helices drives their association (i.e., drives a decrease of the overall nonpolar solvent-accessible surface area) at low bulk concentrations (c₃ < 1M). However, when the bulk TMAO concentration is further increased, this effect is overcompensated by cohesive van der Waals interactions of TMAO with the two nonpolar solutes, leading to their dissociation. The concentration (c₃) where the hydrophobic association is strongest thus depends on a delicate energy–entropy balance linked to the weak binding of TMAO to nonpolar surfaces. This energy–entropy balance changes under varying environmental conditions, such as pressure.

High pressure. In pure water, pressure shifted the helix–association equilibrium toward the dissociated state, as indicated by the increase of the Gibbs energy of association [Fig. 1(a) at c₃ = 0M]. This observation may be explained by a mechanism where the corresponding increase in water-accessible nonpolar surface area leads to reduced water crowding.⁷ Interestingly, at 2 kbar, the Gibbs energy of association also depended non-monotonically on c₃, but the minimum at c₃ = 2M was deeper and shifted to a larger value of c₃ as compared to the data obtained at 1 bar. This observation is linked to the high-pressure polarization of the TMAO dipole that leads to stronger TMAO hydration and a correspondingly weaker preferential binding of TMAO to the two α-helical solutes, as will be demonstrated below. Therefore, a higher TMAO bulk concentration was required to reach the turning point where cohesive van der Waals interactions of TMAO with the two nonpolar solutes overcompensated the entropic driving force for their association. The shift of the Gibbs energy minimum caused by this mechanism led to the remarkable observation that at c₃ = 2M, the two data points in Fig. 1(a) almost merged, i.e., at this concentration, TMAO almost fully counteracted the pressure-induced destabilization of the hydrophobic interaction. Figure 2 summarizes the mechanism.

We repeated the high-pressure MD simulations using the low-dipole FF for TMAO. The corresponding data are shown in Fig. 1(b), together with the data obtained with the high-dipole FF. The TMAO stabilization effect observed with the high-dipole FF disappeared when the low-dipole FF was employed in the simulations at 2 kbar. A comparison of the data in Figs. 1(a) and 1(b) showed that the data obtained with the low-dipole FF at 2 kbar were shifted up along the vertical axis as compared to the data obtained with the low-dipole FF at 1 bar, while maintaining qualitatively the same non-monotic dependence on c₃. This indicated that the high-pressure polarization of the TMAO dipole was essential in achieving the pressure stability of hydrophobic contacts.

The observations made with the high- and low-dipole FFs, as an alternative to the above explanation based on preferential binding, could stem from their effects on the difference in partial molar volumes, $ΔV̄2$⁠, upon helix association. Figure 3(a) reports the data obtained for $ΔV̄2$ as a function of c₃ for the low-dipole FF and the high-dipole FF at 1 bar and 2 kbar. The low-pressure data indicated that $ΔV̄2$ decreased as c₃ increased, i.e., at 1 bar, TMAO increased the pressure stability of the hydrophobic contact between the two α-helices. By contrast, this dependence disappeared at 2 kbar independent of whether the high-dipole FF or low-dipole FF was used. Because $ΔV̄2$ did not depend on the TMAO dipole, the differences in the Gibbs energies of association [Fig. 1(b)] obtained at high pressure with the high- and low-dipole FFs originated from differences in preferential binding of TMAO to the α-helices.

The preferential binding coefficient (Γ₂₃) is shown in Fig. 3(b) and confirms this conclusion. When pressure was raised from 1 bar to 2 kbar, Γ₂₃ decreased. While this occurred with both the high- and low-dipole FFs, Γ₂₃ decreased more with the high-dipole FF, as expected based on the stronger TMAO–water binding of the high-dipole model. Note that the data in Fig. 3(b), furthermore, show that Γ₂₃ > 0, i.e., TMAO accumulated in the hydration shell of the α-helix and stabilized the hydrophobic contact interaction of the two helices through direct interactions, as indicated in Fig. 2. Positive preferential TMAO binding (Γ₂₃ > 0) has been observed at air–water and hydrophobic polymer–water interfaces,²⁸ and may be driven by solvent-crowding effects in bulk solution.^29,30

In summary, the present work showed that the term “piezolyte” aptly captures the effect of TMAO in terms of high-pressure stability of hydrophobic interactions. We proposed a stabilization mechanism that, by contrast to indirect depletion effects on protein stability, involves the role of weak van der Waals interactions. This direct interaction mechanism is based on two opposing thermodynamic driving forces: (1) an entropic force that strengthens the hydrophobic attraction between the extended nonpolar α-helices due to the natural tendency of adsorbed TMAO molecules to be assimilated by the bulk solution, (2) an enthalpic driving force that weakens the hydrophobic attraction between the extended nonpolar α-helices due to their cohesive van der Waals interaction with TMAO. The equilibrium of these two forces resulted in a minimum in the Gibbs energy of hydrophobic association upon changing the concentration of TMAO at low and high pressure. The high-pressure minimum occurred due to the electronic polarization of TMAO and was deeper and shifted to a higher concentration of TMAO, indicating that an increased concentration of TMAO at high pressure provided a stabilizing environment for hydrophobic interactions. We speculate that this mechanism may play a role in the observed increase of the muscle TMAO content in teleosts with the depth of the catch,¹⁴ but leave a detailed analysis to further studies.

We emphasize that the present work involves hydrophobic interactions of model nonpolar solutes. Because TMAO is depleted from many protein surfaces,³¹ the mechanism described herein will probably play a subordinate role in pressure denaturation of proteins but may have implications for protein–protein interactions driven by hydrophobic attraction between extended nonpolar surface patches. The piezolyte effect described herein resulted from nonspecific van der Waals interactions of TMAO with nonpolar surfaces and should therefore be generic for hydrophobic interactions under extreme pressure conditions.

See the supplementary material for the simulation methods including tables, supplemental text explaining the preferential binding coefficient, and supplemental figures showing the PMFs as a function of helix distance at different pressure values and TMAO concentrations. The simulation input files are available at https://tudatalib.ulb.tu-darmstadt.de/handle/tudatalib/3603.

The authors offer special thanks to Dr. Swaminath Bharadwaj for fruitful discussions. The MD simulations were performed on the Lichtenberg High-Performance Computer of the Technische Universität Darmstadt, Germany.