Introduction

Unlike the proteins that we generally encounter, which are denaturated by gentle heating, proteins in hyperthermophiles, microorganisms able to survive in temperatures above 80 °C, exhibit extremely high thermal stability. Their stabilization mechanism has attracted much attention1,2,3,4,5,6,7,8,9, as these proteins are mainly composed of general amino acids, indicating that the same weak molecular interactions seen in general proteins, such as hydrogen bonding, charge interactions, van der Waals (vdW) interactions and the hydrophobic effect contribute to their high thermal stability. Thus, a deep understanding of these stabilization mechanisms will enable us to rationally design not only new proteins but also novel artificial molecules with very high thermal stability relying only on molecular interactions much weaker than the covalent bond.

Although many efforts have been made in protein engineering using point mutation10,11,12 and fragment proteins13 to reveal this stabilization mechanism in the past two decades, our understanding has not yet reached the point where we can establish a general design principle applicable to a variety of molecular families. Artificial molecular self-assemblies generated with the aid of the hydrophobic effect have also been reported14,15,16,17,18,19,20,21,22,23,24. Previously we reported hexameric aggregates of gear-shaped amphiphilic molecules (GSAs) in aqueous methanol23 and in water24 under the control of the solvophobic effect and vdW interactions between the components25.

Here we report a highly thermally stable cube-shaped artificial molecular self-assembly composed of six GSAs in water. Even though the nanocube is assembled solely with the aid of vdW interactions, cation–π interactions and the hydrophobic effect as in proteins, one of the synthetized nanocube complexes shows higher thermal stability than the most stable thermophilic protein ever reported (PhCutA1)26, over 150 °C. Comparing the thermodynamic parameters in the GSA library enabled to evaluate the contribution of each molecular interaction to the stability, which emphasized that extremely high thermal stability is realized only when all the molecular interactions and the hydrophobic effect efficiently work together.

Results

Molecular design

Among the stabilizing factors in proteins, the hydrophobic effect much contributes to the stability27,28,29,30,31. The hydrophobic effect originates from the desolvation of water molecules upon folding and assembly, and the stabilization free energy linearly decreases with the decrease in the desolvation surface area (ΔSAS)32. Van der Waals forces are the weakest molecular interactions involved, but are not negligible when molecular surfaces tightly contact with each other. With such features in mind, in order to attain the hydrophobic contribution effectively and to maximize vdW interactions, we designed a large hydrophobic surface with tight meshing and high complementarity as in a mortise and a tenon, or as in Japanese Hozo, which is a traditional method used to assemble furniture without the use of nails and glue (Fig. 1a, b). Our molecular Hozo, BM GSA (Fig. 2), is based on a C2v symmetric hexaphenylbenzene (HPB) framework with three types of substituents on the periphery. Two hydrophilic N-methylpyridinium rings (cyan rings in Fig. 2) are introduced so as to endow water-solubility. The shape complementarity between the GSAs in the nanocube can be tuned by alkyl groups (methyl groups and deuterium atoms coloured in red and green, respectively, in Fig. 2). In the nanocube, the benzene ring coloured in magenta in Fig. 2 is placed between the two positively charged N-methylpyridinium rings to generate cation–π interactions (Fig. 1c, d), which are often seen in binding sites relating to positively charged species in biological systems33,34,35.

Fig. 1
figure 1

Design concept and structural features of the nanocube. a A photo of a mortise and a tenon, or Hozo, which is traditionally used in Japan for making furniture without the use of nails and glue. b A concept of molecular Hozo toward supramolecular structures constructed without directional molecular interactions. The hydrating water molecules on both faces of the GSA are omitted for clarity. c The nanocube contains a C3 axis and has point symmetry, belonging to the S6 point group. d Molecular structure of BM nanocube. e The nanocube has about a 1-nm-sized inner space, which can be filled with hydrophobic molecules (G)

Full size image
Fig. 2
figure 2

Chemical structures of gear-shaped amphiphiles. Gear-shaped amphiphiles (GSAs) are hexaarylbenzene derivatives possessing two hydrophilic N-methylpyridinium rings (cyan) and hydrophobic substituents on the periphery of a hexaphenylbenzene core. The GSAs have different hydrophobic groups. BM contains a benzene ring (magenta) and three methyl groups (red) as hydrophobic substituents. In the same way, PM contains a pyridine ring (yellow) and three methyl groups (red), HM a hydrogen and three methyl groups (red), and BD a benzene ring (magenta) and three deuterium atoms (green)

Full size image

Self-assembly of nanocubes in water

BM GSA was synthesized from a pentabrominated HPB derivative through the selective alternate lithiation36, 37 as the key reaction followed by cross-coupling reactions (Supplementary Methods). A 1H NMR spectrum of BM GSA in D2O showed a complicated spectral pattern (Fig. 3b), while that in CD3OD was perfectly consistent with that of the monomeric BM GSA (Fig. 3a), suggesting hydrophobic aggregation of BM GSAs in water. 1H DOSY measurement revealed that all the signals show the same diffusion coefficient (D = 1.51 × 10–10 m2 s–1) (Supplementary Fig. 4), indicating the formation of a single species with a hydrodynamic radius of ca. 1.1 nm, which is almost the same as the molecular size of the nanocube. Three chemically inequivalent methyl signals of the tolyl groups coloured in red in Fig. 2 (0.89–1.34 ppm in Fig. 3b) indicate that the GSAs in the aggregate have C1 symmetry (Fig. 1c), which is consistent with the crystal structure of PM nanocube with S6 point symmetry24 (Fig. 1c). A large up-field shift of some aromatic and the tolyl methyl protons arising from the shielding effect strongly suggests the molecular meshing between the GSAs in the nanocube composed of six BM GSAs.

Fig. 3
figure 3

1H NMR spectra of GSAs and nanocubes (500 MHz, 298 K, [GSA] = 1 mM). a BM GSA in CD3OD. b BM nanocube in D2O. c PM GSA in CD3OD. d PM nanocube in D2O. e HM GSA in CD3OD. f HM nanocube in D2O. g BD GSA in CD3OD. h BD nanocube in D2O. The assignment of the 1H NMR signals is indicated in Supplementary Figures 1–12

Full size image

Thermal stability of nanocubes

The thermal stability of the BM nanocube was investigated by variable temperature (VT) 1H NMR measurements. Upon increasing the temperature, only the tolyl signals for the nanocube and for the monomer appeared in an up-field region (Fig. 4 and Supplementary Fig. 22), indicating highly cooperative assembly/disassembly between the two states. In addition, the assembly/disassembly transition was perfectly reversible. The disassembly temperature (T1/2) of the BM nanocube, at which half of the nanocubes are disassembled into the monomers, is 403 K (Table 1), which is higher than the melting temperature (Tm) of most hyperthermophilic proteins known to us (because different definitions of Tm have been introduced in the previous reports, Tm and T1/2 are used separately in this paper). These results clearly indicate that it is possible to construct thermally stable self-assembled structures utilizing only weak molecular interactions and the hydrophobic effect.

Fig. 4
figure 4

Partial variable temperature 1H NMR spectra of BM nanocube showing the methyl signals of the tolyl groups (500 MHz, D2O, [BM GSA] = 1.1 mM). Blue and red solid circles indicate BM nanocube and BM monomer, respectively. The signal of residual DHO at 298 K was used as a reference. The BM nanocube and the BM monomer were quantified by the integrals of the methyl signals. VT 1H NMR spectra for the other nanocubes are shown in Supplementary Figures 23–29

Full size image
Table 1 Properties of nanocubes
Full size table

Next, in order to evaluate the contribution of each molecular interaction, thermal stability of the nanocubes assembled from GSAs with different substituents introduced on the HPB core (Fig. 2) was investigated (Table 1). When the benzene ring in BM GSA (coloured in magenta) is replaced with pyridine (BM→PM), whose electrostatic potential surface is slightly less negative than that of benzene, the T1/2 of PM nanocube (385 K) decreased by 18 K (Supplementary Fig. 24), indicating that the stability of the nanocubes is very sensitive to the strength of cation–π interactions. Indeed, the nanocube from HM GSA, which lacks the benzene ring, showed very broad 1H NMR signals (Fig. 2f), suggesting a less ordered structure, but the formation of the nanocube from HM GSA was confirmed by a similar diffusion constant to that of the BM nanocube (Supplementary Fig. 8). As expected, the T1/2 of HM nanocube significantly dropped to 313 K (Supplementary Fig. 26), which is due to the absence of the stabilization from the cation–π interactions, to electrostatic repulsion between the two pyridinium groups and to the decrease in shape complementarity. The replacement of the three methyl groups in the tolyl moieties in BM with deuteriums (BM→BD) also decreased T1/2 (338 K) by 65 K (Table 1 and Supplementary Fig. 28), which suggests the importance of shape complementarity.

The thermal stability of the nanocubes was also investigated by differential scanning calorimetry (DSC) measurements (Fig. 5). The transition temperature of PM nanocube to the monomers is well consistent with T1/2 determined by VT 1H NMR measurements. Because of the temperature limit of the instrument, the transition temperature of BM nanocube and a full disassembly profile for PM nanocube could not be obtained. As to HM and BD nanocubes, observation of no apparent peak prevented the determination of the thermodynamic parameters. Although the size of the nanocubes is as small as 2 nm, ΔSAS for the nanocubes ranges from 3400 to 4200 Å2 (Table 1) and is much larger than that for molecules of a similar size and is comparable to that of small proteins. This is due to the indented hydrophobic surface of GSAs. The stability of the nanocubes well correlates with ΔSAS (Table 1); the larger the ΔSAS, the higher the stability of the nanocube (Supplementary Fig. 50). However, an exception was found between BM and PM nanocubes. Although the desolvation surface areas for BM (4258.6 Å3) and PM (4220.8 Å3) nanocubes are comparable, T1/2 of BM nanocube is 18 K higher than that of PM nanocube, which suggests that BM nanocube is more stabilized by enthalpic contributions from vdW and cation–π interactions.

Fig. 5
figure 5

DSC profiles of nanocubes. Red, blue, green, and orange lines indicate PM, BM, BD, and HM nanocubes, respectively

Full size image

Stability curves of nanocubes

As heat capacity change (ΔCP) in hydrophobic unfolding and disassembly is positive, the change in the free energy for the disassembly, ΔGd (T), displays a parabola-shaped curve (stability curve)38 (Fig. 6a). When ΔCP is constant, ΔGd (T) is expressed by Eq. (1):

$${\mathrm{\Delta }}G_{\mathrm{d}}\left( T \right) = {\mathrm{\Delta }}H_{\mathrm{d}}\left( {T_{\left( {\Delta G = 0} \right)}} \right)\left( {1 - \frac{T}{{T_{\left( {\Delta G = 0} \right)}}}} \right) \\ + {\mathrm{\Delta }}C_{\mathrm{P}}\left( {T - T_{\left( {\Delta G = 0} \right)} - T{\rm{ln}}\left( {\frac{T}{{T_{\left( {\Delta G = 0} \right)}}}} \right)} \right),$$
(1)

where T(ΔG=0) is the temperature at which ΔGd = 0. ΔHd is the enthalpy change for the disassembly of the nanocube. Typical stability curves that are theoretically prepared based on Eq. (1) with constant ΔCp are shown in Fig. 6. The stability curve suggests that an ordered structure should be destabilized upon decreasing temperature, which is contrary to general intuition. Although Eq. (1) is a theoretical equation that is valid only when ΔCP is constant with temperature, the destabilisation of proteins at decreasing temperature was truly observed39, 40. Figure 6a, b shows the change of the stability curve for a constant T(ΔG=0) value with varying ΔCP and ΔHd at T(ΔG=0). In hydrophobic unfolding and disassembly, ΔCP linearly correlates with ΔSAS41, 42, so the stability curve of the nanocube with large ΔSAS is narrow, indicating a large change in ΔGd with temperature (Fig. 6a). When ΔHd at T(ΔG=0) is large, the stability curve shifts upward with no change of the shape (Fig. 6b). There are three strategies to increase the thermal stability of the nanocube (Fig. 6c): (1) by widening the stability curve (black to blue curves) (strategy I), (2) by shifting the stability curve upward (black to red curves) (strategy II) and (3) by shifting the stability curve toward the right (black to green curves) (strategy III). Thus, the extremely high thermal stability of BM nanocube can be discussed by comparing the stability curves of the four nanocubes. To create the stability curve, ΔCP, T(ΔG=0) and ΔHd at T(ΔG=0) must be determined. It is worth noting that while T(ΔG=0) = T1/2 for the folding of monomeric proteins, in the case of molecular assemblies like nanocubes expressed by Eq. (2), T(ΔG=0) is not equivalent to T1/2.

$$\left( {{\mathrm{GSA}}} \right)_6 \leftrightharpoons 6\,{\mathrm{GSA}}$$
(2)
Fig. 6
figure 6

Case studies of a stability curve for nanocube. The change in the free energy of disassembly with temperature is simulated based on Eq. (1) under the assumption that ΔCP is constant with temperature. a Stability curves with different ΔCP of disassembly with constant T(ΔG=0). b Stability curves with different ΔH(T(ΔG=0)) and constant T(ΔG=0). c Three strategies to increase thermal stability. The black line indicates the original stability curve. Lowering the curvature of the black line (blue line) (strategy I), moving the black line up (red line) (strategy II), and moving the black line to the right (green line) (strategy III) increase T(ΔG=0) to gain higher thermal stability. Although T(ΔG=0)≠T1/2 for the nanocubes, T1/2 also increases with the three strategies

Full size image

In addition, in the case of molecular self-assembly, T1/2 is a function of the concentration, so T1/2 is not an appropriate parameter for the discussion of thermal stability. Thus, T(ΔG=0), which is not dependent on the concentration, has often been used as Tm of protein assemblies instead of T1/2. However, in this paper T1/2 at [GSA monomer]total = 1.0 mM is defined as the disassembly temperature because most part of the nanocubes are disassembled into the monomers at T(ΔG=0). It intuitively seems unfair to compare T(ΔG=0) of the nanocubes with Tm of smaller aggregate of proteins (for molecular self-assemblies, T(ΔG=0) is higher than T1/2).

The ΔCP values for HM and BD nanocubes were determined from ΔH values in a certain temperature range (298–308 K), which were experimentally determined by dilution isothermal titration calorimetry (ITC) measurements43 (Supplementary Tables 3 and 4). On the other hand, the stability of BM and PM nanocubes is so high that no heat was detected when a concentrated aqueous solution of the nanocubes was added into pure water, which prevented us from determining the ΔCP values for these nanocubes. The experimentally determined ΔCP value for BD nanocube is well consistent with that estimated by the linear relationship between ΔCP and ΔSAS (ΔCP = 1.88 × ΔSAS)44. However, as mentioned above, because the structure of HM nanocube is less ordered, the ΔCP values determined by the experiment and estimated from the equation contradict each other. For BM and PM nanocubes, the ΔCP values were estimated from the equation (Table 1). T(ΔG=0) and ΔHd at T(ΔG=0) were determined so that the stability curve would fit well to the experimentally obtained ΔGd values. The stability curves of BM and PM nanocubes are narrower but higher than those for HM and BD nanocubes (Fig. 7a), indicating that BM and PM nanocubes were stabilized mainly by enthalpic contributions (strategy II in Fig. 6c). The stability curve of BM nanocube lies slightly higher than that of PM nanocube, reflecting higher enthalpic stability of BM, which would originate from stronger cation–π interactions between the benzene and the pyridinium rings. Considering that the ΔCP values for HM and BD nanocubes are smaller than those for BM and PM nanocubes, the ΔGd values for HM and BD nanocubes are less sensitive to temperature, which corresponds to strategy I (Fig. 6c). Indeed, even though the stability curves of HM and BD nanocubes lie below those of BM and PM nanocubes, T(ΔG=0) of HM and BD nanocubes are similar to those of PM and BM nanocubes (Fig. 7a).

Fig. 7
figure 7

Plots of ΔGd of nanocubes with temperature and stability curves. Stability curves were obtained so that Eq. (1) would fit the experimentally obtained ΔGd values using T(ΔG=0) and ΔH(T(ΔG=0)) as variable parameters under the assumption that ΔCP is constant with temperature. a Stability curves of free nanocubes (BM, PM, HM, and BD). b Stability curves of G2@nanocubes (G2@BM, G2@PM, G2@HM, and G2@BD). c–f Effect of encapsulation of guests on the stability of the nanocubes. c BM and G2@BM. d PM and G2@PM. e HM and G2@HM. f BD and G2@BD. G indicates 1,3,5-tribromomesitylene. Black crosshairs indicate the data obtained by variable temperature 1H NMR spectroscopy

Full size image

If we suppose that the GSAs in the nanocube tightly mesh with each other, it is logical to think that they are densely packed. This characteristic can be assessed by partial-specific volume (PSV). A good correlation was found between the stability of the nanocubes and their PSV (Table 1), indicating that a more densely packed nanocube is more stable. Although the structure of BM and PM GSAs is very similar, the PSV of BM nanocube is slightly smaller than that of PM nanocube, suggesting that the stronger cation–π interactions in BM nanocube slightly tighten the molecular meshing. Compared with proteins, whose PSV45 ranges from 0.58 to 0.76 cm3 g–1, the PSV of the nanocubes is relatively high, which indicates that the density of the nanocubes is lower than that of proteins. This is quite reasonable because the nanocubes contain about a 1-nm-sized inner void space. Considering that thermally stable proteins have small void spaces46, the high thermal stability of PM and BM nanocubes is surprising, demonstrating that the tight molecular meshing between the GSAs is strong enough to overcome the destabilization coming from having a large void space.

Improvement of thermal stability of nanocubes

The stability of the nanocubes was improved by filling their inner void space with appropriate molecule(s) (Fig. 1e). When 1,3,5-tribromomesitylene (G) was added into the nanocube solutions, a couple of Gs were encapsulated in the nanocubes, which was confirmed by 1H NMR spectroscopy (Supplementary Fig. 16). All the signals for G were shifted downfield due to the shielding caused by the aromatic rings of the GSAs. Filling the void space with Gs decreased the PSV of the nanocubes and T1/2 increased of 20–70 K (Table 1). The T1/2 of G2@BM is higher than 423 K, at which 20% of the nanocube was disassembled into the monomer (exact T1/2 could not be determined because of the temperature limit of our NMR instrument (Supplementary Fig. 23)). G2@BM is thermally more stable than the most stable thermophilic protein ever reported, PhCutA1 (Tm = 421.5 K). The less stable HM and BD nanocubes were additionally stabilized by the encapsulation of the guest molecules (Table 1 and Supplementary Figs. 27 and 29), as a well-ordered nanocube structure was induced from a more fluctuated structure. Upon the encapsulation of the guests, the stability curves for all four nanocubes became similar to each other (Fig. 7b), suggesting that the guest induced the nanocubes to assume a similar shape. All the stability curves of the G2@nanocubes shifted to the right (Fig. 7c–f), suggesting that filling the void space corresponds to the third strategy to improve the thermal stability (Fig. 6c). The top part of the parabola of BM and PM nanocubes slightly decreased (Fig. 7c, d), suggesting the decrease in the enthalpic contribution. This is mainly because the volume of two Gs is slightly larger than the inner space of the nanocube, which causes the molecular meshing between the GSAs to decrease. The nanocubes are thus slightly destabilized enthalpically in the meshing, but additional vdW interactions between the nanocube and G enthalpically contributes to the stability. On the other hand, the stability curves for BD and HM nanocubes rose upon the encapsulation (Fig. 7e, f), which is due to the increase in the molecular meshing between the GSAs caused by the induced-fit to the guests.

Discussion

In conclusion, a self-assembled structure with extremely high thermal stability, BM nanocube, was created from gear-shaped amphiphilic molecules based on Hozo with an indented hydrophobic molecular surface utilizing only weak molecular interactions and the hydrophobic effect. As all the nanocubes have large desolvation surface areas (ΔSAS) comparable to those of small proteins, the hydrophobic effect is a driving force of the formation of the nanocube and also much contributes to the stability of this. However, the low stability of HM and BD nanocubes with large ΔSAS (3431 and 3642 Å2) indicates that other molecular interactions such as vdW and cation–π interactions also play a significant role for the thermal stability. Some thermophilic proteins contain many charged residues (Asp, Glu, Lys, and Arg)26, 47, 48, suggesting the importance of electrostatic interaction. Likewise, BM nanocube, which has slightly stronger cation–π interactions (between the benzene and the N-methylpyridinium rings) than PM nanocube (which has a pyridine instead of a benzene), has greater thermal stability, which indicates that even slight differences in the electrostatic potential surfaces between benzene and pyridine clearly affect the stability. Molecular meshing, which mainly contributes to vdW interactions between GSAs, is another important factor to increase the thermal stability. Therefore, the extremely high thermal stability of the nanocube is attained only if all the contributions work together. In other words, for the creation of a highly stable self-assembly or folded structure in water, (1) a desolvation surface area (ΔSAS) of over several thousands Å3 (the hydrophobic effect), (2) complementary hydrophobic surfaces (vdW interactions), and (3) electrostatic interactions such as cation–π interactions should properly be included in designed molecules. Such new design principles would enable to pave the way toward discrete molecular self-assemblies with high stability in water that can be utilized for the purpose of the interplay with biological molecules under physiological conditions and for industrial purposes such as the transformations of molecules under harsh conditions.

Methods

Synthesis of GSAs

Synthesis of GSAs is detailed in the Supplementary Methods. NMR characterization of GSAs is provided in Supplementary Figures 1–12. 1H and 13C NMR spectra of all the new compounds are provided in Supplementary Figures 51–86.

Mass spectrometry

ESI-MS spectra of nanocubes are provided in Supplementary Figures 13–15 and the observed species are summarized in Supplementary Table 1.

Preparation of nanocube with guests

Preparation of nanocube with guests is detailed in the Supplementary Methods. 1H and 1H DOSY NMR spectra are provided in Supplementary Figures 16–19.

H/D exchange

H/D exchange of nanocubes in D2O at high temperature is described in the Supplementary Methods. 1H NMR spectra are provided in Supplementary Figures 20 and 21.

Determination of T 1/2 of nanocubes

Determination of T1/2 of nanocubes is detailed in the Supplementary Methods. 1H NMR spectra are provided in Supplementary Figures 22–29 and the data are provided in Supplementary Table 2.

Dilution ITC experiments

Dilution ITC experiments are detailed in the Supplementary Methods. Titration curves are provided in Supplementary Figures 30–40 and the data are provided in Supplementary Tables 3 and 4.

DSC measurement

DSC measurement is detailed in the Supplementary Methods.

Partial-specific volume

Determination of partial-specific volume is detailed in the Supplementary Methods. The results are provided in Supplementary Figures 41–48 and Supplementary Tables 5 and 6.

Desolvation surface area of nanocubes

Computational methods to determine ΔSAS of nanocubes are detailed in the Supplementary Methods. Geometrically optimized structures of nanocubes are provided in Supplementary Figure 49. Plots of the T1/2 with respect to the ΔSAS of nanocubes are shown in Supplementary Figure 50. The detailed data of ΔSAS are provided in Supplementary Table 7.

Determination of stability curves for nanocubes

Determination of stability curves for nanocubes is detailed in the Supplementary Methods. The data for the analysis of the nanocubes and the stability curves with the experimentally obtained ΔGd(T) are shown in Supplementary Tables 8–15.

Data availability

The authors declare that all the other data supporting the findings of this study are available within the article and its supplementary information files and from the corresponding author upon request.