Hierarchical formation of helical supramolecular polymers via stacking of hydrogen-bonded pairs in water

Abstract

Bifunctional ureido-s-triazines provided with penta(ethylene oxide) side chains are able to self assemble in water, leading to helical columns via cooperative stacking of the hydrogen-bonded pairs (DADA array). Monofunctional ureido-s-triazines do not form such helical architectures. The presence of a linker, covalently connecting the two ureido-s-triazine units, is essential as it generates a high local concentration of aromatic units, favorable for stacking interactions. This hydrophobic stacking of the aromatic units occurs at concentrations as low as 5⋅10⁻⁶ M and can be visualized by using fluorescence spectroscopy. The stacking generates a hydrophobic microenvironment that allows intermolecular hydrogen bonding to occur at higher concentrations because the hydrogen bonds are shielded from competitive hydrogen bonding with water. This hierarchical process results in the formation of a helical self-assembled polymer in water at concentrations above 10⁻⁴ M. Chiral side chains attached to the ureido-s-triazine units bias the helicity of these columns as concluded from CD spectroscopy and “Sergeants and Soldiers” experiments.

In DNA, the well known double helical motif originates from self assembly and is directed by lateral hydrogen bonds between bases and stabilized by solvophobic interactions between the covalently linked bases perpendicular to the hydrogen bonds (1). Control over the intrinsic helicity of the structure is governed by the peripheral chirality in the sugar–phosphate backbone. In synthetic systems, noncovalent interactions have been used to obtain well defined self-assembled architectures in organic solvents (2–7). Peripheral chiral centers in assemblies (8–11) and chiral side chains attached to a polymer backbone (12–20) have been shown to bias chirality at the supramolecular level. Highly ordered multimolecular supramolecular structures stable in water, held together by hydrophobic interactions or with additional hydrogen bonding, are also known (21–25). However, it remains difficult to exploit directional noncovalent interactions in a cooperative way for the formation of discrete multimolecular assemblies stable in water. In this solvent, water molecules strongly compete with directional polar interactions such as hydrogen bonds, resulting in low association constants.

The formation of helical self-assembled polymers in solution by both stacking and hydrogen bonding has recently been demonstrated by us (26, 27). Self-complementary apolar molecules 1 and 2 dimerize via strong cooperative 4-fold hydrogen bonding (ADAD) in chloroform (28), giving rise to the formation of a dimer by 1 and a random coil polymer by 2, the latter featuring viscous solutions at higher concentrations. Additional to the hydrogen bonding, solvophobic interactions between the aromatic surfaces arise for 1 and 2 in alkanes. Association via hydrogen bonding of the ureido-s-triazine functional groups leads to the formation of a large and planar aromatic core surrounded by six flexible chains, a structure that is conducive to the formation of aggregates with a columnar architecture in environments that are unable to solvate the core (29). The concomitant occurrence of hydrogen bonding and solvophobically induced stacking of the aromatic cores results in the formation of columnar polymeric architectures for both 1 and 2. The columnar integrity of the assemblies was verified by using small-angle neutron scattering experiments, revealing a column radius of 16 Å, matching with a disk formed by the dimerization of two ureido-s-triazines. In the columns built up by 1, the discs are rotating freely, but in the columns formed by 2, this rotation is restricted by the hexamethylene spacer, linking two discs to each other. The rotational restriction and short length of the spacer of 2 account for the induction of a twist between two consecutive discs, resulting overall in the formation of helical columns. The homochiral side chains of 2b bias the chirality of the helix and generate homochiral columns. Cooperativity within the helix was shown by mixing chiral 1b with achiral 2a, giving rise to a Cotton effect, whereas the two compounds individually are CD inactive. Two molecules 1b act as chiral end groups of the helix formed by 2a, and express their side-chain chirality in the backbone of the helix.

Studies have shown that selective hydrogen-bond formation by organic molecules in bulk water almost never occurs because of the interference of the competitive solvent molecules (30). Natural molecules use a hydrophobic microenvironment that shields the hydrogen bonding from the water such as in the double helix of DNA (1) and the interior of proteins (31, 32). Synthetic systems using this principle have been demonstrated (21–25, 33–41). It was anticipated that such a microenvironment, which is required for the creation of multimolecular helical columnar architectures in water analogous to those observed for 2 in dodecane, would be created by the arene–arene stacking of the hydrophobic aromatic surfaces of the molecules. To ensure water solubility molecules, 1 and 2 were modified by providing them with penta(ethylene oxide) monomethyl ether side chains. Compounds 3 and 4 were accordingly designed and synthesized in an enantiomerically pure form and subsequently studied in aqueous solutions (26). In this paper, we present optical studies on these molecules in solution, revealing a hierarchical assembly process in water (Scheme S1).

Scheme 1.

Results and Discussion

Compounds 3 and 4 were synthesized by using previously described precursor molecules (11) (see supporting information on the PNAS web site, www.pnas.org). Monofunctional 3 was obtained as a colorless oil, whereas bifunctional molecules 4a,b were obtained as waxy white solids that showed birefringence in polarized optical microscopy. Heating of the samples resulted in loss of birefringence at 115 and 73°C for 4a and 4b, respectively. The formation of flower-like patterns on cooling indicated the formation of a columnar liquid crystalline mesophase.

¹H-NMR Spectroscopy.

Self assembly of compounds 3 and 4 in solution was investigated by ¹H-NMR spectroscopy. In DMSO, all compounds exist as single molecules resulting in ¹H-NMR spectra featuring the signals of the monomeric species. When dissolved in deuterated chloroform, monofunctional 3 gave ¹H-NMR spectra featuring three NH protons between δ = 9 and 10.5 ppm and one at δ = 5.8 ppm (Fig. 1 Upper). These positions are typical for molecules dimerized via self-complementary hydrogen bonding; the three hydrogen-bonded protons are visible at low field, whereas the nonhydrogen-bonded amine proton can be found at high field (28). Monofunctional 3 also dissolves readily in water, and the spectra between 11 and 5.3 ppm at different temperatures are given in Fig. 1 for a 10 mM solution. At 80°C, only the signals of the aromatic protons and the two urea protons, which coincide, are visible. The amine protons exchange rapidly with the water hydrogens at this temperature, and therefore their signal is not visible. Lowering of the temperature gives rise to the decoalescence of the NH signals, resulting in the emergence of two unequal signals for the urea protons and the appearance of two broad signals for the amine protons. At low temperatures (<10°C), broadening of the urea and aromatic signals occurs, suggesting aggregation of the molecules, most probably via arene–arene stacking. The spectra thus indicate that at high temperatures in water, the monofunctional molecules are molecularly dissolved and on lowering of the temperature, aggregation takes place via stacking of the hydrophobic aromatic units, however without significant occurrence of intermolecular hydrogen bonding.

Figure 1.

¹H-NMR (500 MHz) spectra between 11.0 and 5.3 ppm of 3 in CDCl₃ and in H₂O/D₂O 9/1 (≈10 mg/ml; 10 mM) at different temperatures.

Similar to monofunctional 3, bifunctional 4 smoothly dissolves in chloroform. ¹H-NMR spectroscopy revealed that 4 dimerizes in chloroform via self-complementary 4-fold hydrogen bonding (28) (Fig. 2 Lower), which affords a supramolecular polymer (4). In both methanol at room temperature and water at higher temperatures, the signals of the amine protons of 4 are invisible, indicating that they are in rapid exchange with the solvent (Fig. 2). Furthermore, no downfield NH signals are observed, indicating that the molecules dissolve in methanol and in water at high temperatures without formation of hydrogen bonds. However, in contrast to monofunctional 3, bifunctional compounds 4 did not dissolve readily in water at room temperature; 4b especially required heating and sonication for approximately 15 min before it dissolved. Lowering the temperature of the aqueous sample resulted in changes in the proton NMR spectrum suggestive for aggregation; a broadening of the signals occurred pointing to hydrophobic arene–arene stacking and, more importantly, signals at positions typical for hydrogen-bonded protons appeared between 9 and 10.5 ppm (Fig. 3) (28). Initially, only broadening of the signals was observed, similar as for 3 in water at low temperatures. However, at temperatures below 50°C, additional signals, assigned to hydrogen-bonded protons, appear. Because of the concomitant occurrence of both stacking and hydrogen bonding, the signals become rather broad.

Figure 2.

¹H-NMR (500 MHz) spectra between 11.0 and 5.0 ppm of 4b in CDCl₃ and CH₃OH/CD₃OD 9/1 at room temperature and in H₂O/D₂O 9/1 at 90°C (≈10 mg/ml; 5 mM).

Figure 3.

¹H-NMR (500 MHz) spectra of 4b in CDCl₃ at room temperature and in H₂O/D₂O 9/1 (≈10 mg/ml; 5 mM) at different temperatures in the regime of 5.3–11.0 ppm. The baseline of the water samples is curved near 5 ppm because of the suppression of the water signal with the JUMPRET measurement technique.

UV-Vis and CD Spectroscopy.

To further investigate the self assembly of 3 and 4 in solution, UV-Vis and CD spectroscopy measurements were performed in various solvents. In the UV-Vis spectra, an absorption band around 290 nm is present in dilute (10⁻⁴ M) solutions. More specifically, the absorption maximum in methanol, acetone, and acetonitrile (hydrogen-bond breaking solvents) lies at 288 nm, whereas in chloroform, a small but significant red shift to 292 nm occurs for both 3 and 4. The absorption maximum in chloroform is at the same position as the maximum observed in the solid state (hydrogen-bond supporting conditions) and that of apolar analogs 1 and 2 in chloroform and dodecane. From this, we conclude that this red-shifted maximum at 292 nm refers to the hydrogen-bonded form of the molecules, and the maximum at 288 nm corresponds to molecularly dispersed, not hydrogen-bonded, molecules. The UV-Vis spectrum of 4b in water at 10⁻⁴ M reveals a maximum at 292 nm, indicating that the molecules are dimerized via hydrogen bonding.

When the same solutions are monitored with CD spectroscopy, no Cotton effect is observed for the methanol, acetone, acetonitrile, and chloroform solutions, but intriguingly a Cotton effect is observed in water, at the position of the UV maximum. In the polar organic solvents methanol, acetone, and acetonitrile, the molecules are molecularly dispersed and are thus not present in a preferred supramolecular chiral conformation. In chloroform, the ureido-s-triazine groups dimerize via quadruple hydrogen bonding, but the aromatic units do not stack, resulting in a disordered supramolecular random coil polymer in analogy with its ureidopyrimidinone counterpart (4). In water, the molecules both dimerize via quadruple hydrogen bonding and stack via solvophobic interactions in a supramolecular chiral conformation. The occurrence of both of these interactions accounts for the formation of a superstructure in which the chiral side chains can transfer their chirality to the helical arranged aromatic core, reflected in the presence of a Cotton effect. The results obtained for 4b in water are highly analogous to those of 2b in dodecane. It is proposed that the superstructure formed by 4 in water is a helical column of stacked hydrogen-bonded pairs, in analogy to 2 (Fig. 4) (26). The liquid crystalline nature of these molecules does not allow for the generation of high-resolution scattering data, and as such the exact helical nature of the columns cannot be established. The comparison might be made here to the difficulty of obtaining information concerning the structural arrangement of molecules in micelles beyond the globular or columnar shape.

Figure 4.

Proposed mode of association of bifunctional 4b in water into helical columns.

The self assembly of 4b in water was further investigated with temperature-dependent UV-Vis and CD spectroscopy. An increase of temperature results in a gradual blue shift of the absorption maximum from 292 to 288 nm (Fig. 5 Upper). Simultaneously, the Cotton effect gradually decreases on increase of the temperature until it has disappeared around 90°C (Fig. 5 Lower).† The changes observed on increase of the temperature indicate the coincidental loss of positional order of the molecules within the columns (CD) and loss of intermolecular hydrogen bonding (UV-Vis and ¹H-NMR), thus showing that increased temperatures denaturate the hydrogen-bonded pairs (analogous to the observations made with ¹H-NMR).

Figure 5.

UV-Vis (Upper) and CD spectra (Lower) of 4b in water (2.4⋅10⁻⁴ M) at different temperatures. The Cotton effects are stable in time, reversible, and independent of cooling or heating rate.

A 2⋅10⁻⁴ M aqueous solution of 4b becomes turbid above 80°C. This is ascribed to the phase separation of the oligo(ethylene oxide) side chains with the water, known as the lower critical solution temperature behavior (42). However, in contrast to previously studied discotics (43), the temperature-induced oligo(ethylene oxide) side-chain aggregation does not result in the reappearance of a Cotton effect, affirming the loss of structuring hydrogen-bonding interactions at high temperatures.

To investigate the importance of a covalent linker between two ureido-s-triazine units, i.e., bifunctional molecules, UV-Vis and CD spectra of chiral monofunctional 3 were recorded at different concentrations in water. At 10⁻⁴ M, the UV-Vis spectrum displays an absorption maximum at 288 nm, and no Cotton effect is observed with CD spectroscopy (Fig. 6). In line with the ¹H-NMR experiments performed for 3 (Fig. 1), the UV-Vis spectra thus indicate that intermolecular hydrogen bonding is not occurring in water. As a result, monofunctional 3 does not form a chiral supramolecular assembly in water at the applied low concentrations. A red-shifted absorption maximum in the UV-Vis spectra becomes visible only at very high concentration (≈10⁻² M) and concomitantly, a small Cotton effect appears. Apparently, at low concentrations the monofunctional molecules are molecularly dispersed, and only a strong increase of the concentration accounts for enhanced stacking probability and hence some intermolecular hydrogen bonding. Within the aggregates thus formed (not necessarily of the columnar type), the side-chain chirality is only weakly transferred to the aromatic system. The covalent linker of 4 is thus responsible for the generation of helical architectures at low concentrations and positional control of the molecules within the column.

Figure 6.

Normalized UV-Vis (Upper) and CD spectra (Lower) of 3 in water at different concentrations. The spectra were recorded in 1-, 0.1-, and 0.01-mm cuvettes, respectively.

Fluorescence Spectroscopy.

To further explore the differences in aggregation behavior between 4 and 3 and to elucidate the hierarchy in the occurrence of the intermolecular interactions, fluorescence measurements in water and chloroform were undertaken. In water, fluorescence of compounds 4a and 4b is highly quenched when compared with 3 in water or 4a or 4b in chloroform, acetone, methanol, or acetonitrile. The fluorescence of 4b in chloroform is almost temperature independent, even though increase of the temperature initially results in a small increase of the fluorescence intensity. Further increase of the temperature (>30°C) results in the usual small decrease of the fluorescence intensity, supporting the absence of stacking of the molecules. A similar behavior was found for monofunctional 3 in water; an increase of temperature from 0 to 100°C resulted in a gradual decrease of the fluorescence intensity (Fig. 7 Upper). The fluorescence maximum is somewhat red-shifted in water with respect to chloroform, which is most probably because of the difference in solvent polarity.

Figure 7.

Fluorescence spectra of monofunctional 3 in water (Upper) and bifunctional chiral 4b in water (Lower), recorded at ≈5⋅10⁻⁶ M with λ_exc at 290 nm. It should be noted that, because of normalization, the approximately 20-fold stronger fluorescence of 3 in water at room temperature with respect to 4b is not reflected in the depicted intensities.

Molecularly dispersed π-conjugated oligomers feature a strong fluorescence intensity, whereas aggregation/stacking of these oligomers results in a strong quenching of their fluorescence intensity (44, 45). Aggregation of these oligomers can generally be induced by lowering of the temperature or addition of a nonsolvent. The fluorescence of 4 in water shows a behavior analogous to that of π-conjugated oligomers. The fluorescence of aqueous solutions of 4a or 4b is highly quenched, but an increase of the temperature results in a strong increase of the fluorescence intensity (Fig. 7 Lower). This behavior of 4 suggests that at low temperatures, the molecules are not molecularly dispersed. Bifunctional molecules 4 are capable of hydrophobic stacking interactions between the two apolar ureido-s-triazine units. This stacking presumably causes the strong quenching of the fluorescence. UV-Vis and CD measurements on aqueous solution of 4 have shown that an increase of the temperature accounts for increased motion of the molecules and concomitant decrease of stacking interactions. Accordingly, the increase of the fluorescence of 4 at higher temperatures results from a reduction of the stacking interactions, a feature typically observed for the π-conjugated oligomers as well. Monofunctional 3 does not stack in water and thus features a strong fluorescence that decreases at increased temperatures.

Compound 3 is molecularly dissolved in water, but the presence of the hexamethylene linker of 4 enables the stacking of two ureido-s-triazine units in water. The fluorescence experiments confirm that bifunctional molecules 4 have a much stronger tendency for hydrophobic stacking of the units. This is most probably because of a high local concentration of functional groups, rationalizing why self-assembled architectures will form at lower concentrations than 3. The analogy might be made here to nucleic acid oligomers; whereas single mononucleotides do not form a hydrogen-bonded supramolecular structure in water, covalently connected oligonucleotides do (1). That stacking is observed already at concentrations as low as 5⋅10⁻⁶ M indicates that it precedes hydrogen bonding and concomitant formation of helical columns, which occurs around 10⁻⁴ M. The use of a hydrophobic interaction as the driving force for self assembly, generating a hydrophobic microenvironment to allow expression of hydrogen bonding, has also recently nicely been demonstrated for the hierarchical formation of membranes (24).

Amplification of Chirality.

In water, bifunctional molecules 4a and 4b aggregate at low concentrations (10⁻⁴ M) in helical columns. In principle, intrinsically helical columns are suitable for the transfer of chiral information from one molecule to the other. The stacking of a molecule with achiral side chains on top of a molecule with chiral side chains would possibly result in the amplification of chirality from the chiral to the achiral molecule. To investigate this cooperativity in the stacking of the molecules, “Sergeants and Soldiers” (13) experiments were performed on mixtures of achiral bifunctional 4a (“soldiers”) and chiral 3 or chiral 4b (“sergeants”). In analogy with previously studied discotics (43), the Cotton effect of the mixtures was studied as a function of their composition.

When solutions of varying amounts chiral 3 and achiral 4a—keeping the total concentration of chromophores constant—were prepared at 10⁻⁴ M in water, no Cotton effect was observed for any of the mixtures. The maximum of the absorbance spectrum shifts linearly from 288 to 292 nm on going from 100% 3 to 100% 4a. Even though at this concentration the achiral bifunctional molecules 4a are aggregating, their helicity is not biased by the presence of molecules of 3, because the monofunctional chiral molecules 3 are not aggregating at this low concentration. When mixing experiments were performed at high concentration (5⋅10⁻³ M) and low temperature (5°C), conditions under which monofunctional 3 aggregates to some extent, the maximum UV-Vis absorption was observed at 292 nm, independent of composition. When these mixtures were examined by CD spectroscopy, Cotton effects were indeed observed. A nonlinear increase of the g_abs was observed on addition of small amounts of 3 to 4a (Fig. 8). The deviation from linearity shows that chirality is being transferred from chiral monofunctional 3 to achiral bifunctional 4a. In Fig. 8, the proposed mode of the action of chirality transfer is visualized.

Figure 8.

(Upper) Dependence of the overall chirality on the mol fraction of chiral 3 in mixtures of 3 and achiral 4a in water at 5°C, expressed in terms of the g value and measured at the maximum of the Cotton effect at 287–293 nm. Measurements were recorded at a constant total concentration of functional groups of 5 × 10⁻³ M in a 0.1-mm cell. (Lower) Proposed mode of amplification of chirality in mixtures of 3 and 4a. Achiral 4a forms nonbiased helical columns, mixing with monofunctional chiral 3 results in the bias of the columns of 4a.

Inducing a bias in the helicity of columns of 4a has also been investigated by addition of chiral bifunctional 4b. UV-Vis spectra of different mixtures of 4a and 4b in water are identical, indicating that similar helical structures are being formed by all mixtures. Surprisingly, however, the intensity of the Cotton effect was found to depend linearly on the amount of chiral 4b, showing that amplification of chirality does not occur for these mixtures. Two explanations can be thought of: there is no cooperative order in the helix, or the molecules do not form mixed aggregates. The first explanation seems unlikely, because amplification of chirality was observed in mixtures of 4a and 3, as well as for their apolar analogs 1 and 2 (26). A strong difference in solubility between 4a and 4b in water exists. Whereas achiral 4a dissolved readily in water at room temperature after several minutes, chiral 4b could be dissolved only on sonication at elevated temperatures for 15 min. This difference in solubility most probably accounts for the formation of aggregates consisting of one type of molecule only. Apparently the exchange of molecules between these aggregates is kinetically hindered.‡

Conclusion

The creation of columnar architectures in water by 3 and 4 using hydrogen-bonded pairs requires both hydrophobic and hydrogen-bonding interactions. The hydrophobic microenvironment is the driving force for self assembly and is subsequently used to shield the hydrogen bonding from the solvent. Monofunctional 3 does not stack via hydrogen-bonded pairs in water, but bifunctional 4a and 4b do. This stacking of 4 finds its origin in a high local concentration of hydrophobic units because of the hexamethylene linker. Hydrophobic stacking, occurring at concentrations below 10⁻⁵ M, precedes the hydrogen bonding at 10⁻⁴ M. Within the helical columns, a certain degree of cooperativity exists, as chiral monofunctional 3 is capable of biasing the helicity of columns formed by achiral bifunctional 4a at high concentrations.

The findings presented in this paper provide design rules for the formation of highly ordered aggregates in water by using a combination of noncovalent interactions. Furthermore, we have shown that the self-assembly process on the basis of these interactions is a hierarchical process with the solvophobic interactions occurring at lower concentrations than the solvent-sensitive hydrogen bonds. The implementation of these design criteria will allow the generation of a variety of ordered supramolecular architectures in water.