In situ monitoring of molecular aggregation using circular dichroism

The aggregation of molecules plays an important role in determining their function. Electron microscopy and other methods can only characterize the variation of microstructure, but are not capable of monitoring conformational changes. These techniques are also complicated, expensive and time-consuming. Here, we demonstrate a simple method to monitor in-situ and in real-time the conformational change of (R)-1,1′-binaphthyl-based polymers during the aggregation process using circular dichroism. Based on results from molecular dynamics simulations and experimental circular dichroism measurements, polymers with “open” binaphthyl rings are found to show stronger aggregation-annihilated circular dichroism effects, with more negative torsion angles between the two naphthalene rings. In contrast, the polymers with “locked” rings show a more restrained aggregation-annihilated circular dichroism effect, with only a slight change of torsion angle. This work provides an approach to monitor molecular aggregation in a simple, accurate, and efficient way.

T he term "chiral" is used in chemistry to describe molecules which are non-superimposable on their mirror image [1][2][3][4] . Chiral molecules are abundant in nature, with many natural chemicals such as amino acids and deoxyribonucleic acid (DNA) exhibiting chirality. There are various ways to characterize chirality, and amongst them circular dichroism (CD) is the most commonly used tool to determine the absolute configuration, both in solution and the solid state, because of its simple operation, high accuracy and efficiency [5][6][7][8] . Particularly, CD has been widely used to analyze the secondary structure of opticallyactive biomacromolecules such as proteins and DNA, which is sensitive to environmental variation. The dynamic processes of these chiral molecules are important to their biological activity. For example, the aggregation of proteins is associated with many pathological conditions 9 . How to monitor the conformational changes during protein aggregation is significant to clarify the working mechanism of biological processes. Traditionally, these processes have been studied by techniques such as electron microscopy and NMR spectroscopy. In these methods, pretreatment of samples is required and the analyses are carried out under harsh conditions [10][11][12][13][14][15][16][17] . Thanks to its capability for in-situ and real-time detection, CD spectroscopy may be a good choice to investigate the relationship between molecular aggregation and conformational change. In previous work, the effect of aggregation-induced circular dichroism has been observed in some self-assembly systems 18 . The electron microscopic results suggested that CD-inactive molecules were induced to generate strong CD signals upon helix formation in the aggregate state. However, the detailed conformational changes of a single molecule during the self-assembly process is difficult to determine using CD spectroscopy since the detected CD signal results from the whole assembly structure.
Recently, an opposite effect called aggregation-annihilated circular dichroism (AACD) was observed in 1,1′-binaphthyl derivatives. Further research revealed that the annihilation might be caused by the conformational change of 1,1′-binaphthyl (BN) 19 . BN is a well-studied chiral molecule 20,21 . Mason et al. have set up a classical model to calculate the dependence of the CD couplet intensity (Δε max ) and Davydov splitting (Δλ max ) on the torsion angle (θ) in 1,1′-binaphthyl 6,22,23 . Figure 1a shows the electronic transitions and coupling between the 1 B b and 1 B b ′ transitions in 1,1′-binaphthyl, while Fig. 1b demonstrates the simulated CD spectra of (R)-1,1′-binaphthol with θ varying from −45 to −110°. The plots of Δε max /Δε max-0 and (Δλ max − Δλ max-0 ) versus the θ value are shown in Fig. 1c,d, respectively, where Δε max-0 and Δλ max-0 are the values at θ = −45°. To investigate the rate, the first-order derivatives of d|Δε max |/d|θ| and d|Δλ max |/d|θ| are also plotted in Fig. 1c,d with blue scatter lines. As suggested by Fig. 1, the maximum |Δε max | is located at around −65°and the Δλ max gradually decreases when θ evolves from −45°to −110°. The established relationship is expected to determine the absolute configuration of chiral materials using the CD measurement.
In this work, four BN-based chiral polymers are synthesized through simple Suzuki coupling reactions. For the polymers with "open" BN units, the AACD effect is clearly observed. However, the annihilation is restrained once the BN units are "locked". Combining the experimental CD data with the simulation results shown in Fig. 1, the process of molecular aggregation can be monitored simply and accurately. Meanwhile, the relevant molecular dynamics simulation proves that this method is reliable.

Results
Synthesis and characterization. The phenomenon of aggregation-induced emission (AIE) was coined by Tang et al. in 2001. AIE luminogens (AIEgens) such as tetraphenylethylene (TPE) usually possess propeller-shaped structures and show no or weak photoluminescence (PL) when molecularly dissolved in their "good" solvent. However, their emission can be dramatically enhanced once aggregates are formed by adding 'poor' solvent. Thus, AIE exhibits an exact opposite photophysical phenomenon to the traditional aggregation-caused quenching effect discovered by Förster in 1954. Further study has revealed that the restriction of intramolecular motion (RIM) is the cause of the AIE effect. Since TPE possesses attractive PL properties in the aggregate state, it is considered to be a promising candidate to construct chiral materials with unique chiroptical properties [24][25][26] . The CD signal of binaphthyl and PL signal of TPE can be used as a dual signal to monitor molecular aggregation. In this work, four (R)-1,1′-binaphthyl and TPE-based polymers (P-1 to P-4) were synthesized through simple Suzuki coupling reactions (Fig. 2a). The detailed synthetic procedures are given in Supplementary Note 1. In P-1 and P-2, the TPE unit is linked to an "open" binaphthyl ring at the 6-and 3-position, respectively. P-3 and P-4 are structurally similar to P-1 and P-2, respectively but the binaphthyl ring is locked. The structures of all polymers were characterized by 1 H and 13 C NMR spectroscopy with satisfactory results (Supplementary Fig. 1-8). Gel permeation chromatography revealed that their weight-average molecular weight is 11900, 9200, 14000, 4800 and polydispersity is 2.53, 1.87, 2.17, 2.00, respectively (Supplementary Table 1). Thermogravimetric analysis (TGA) revealed that all polymers show 5% weight loss at high temperatures of up to 420°C (Fig. 2b), suggesting that they are thermally stable.
Circular dichroism. According to previous studies, it is anticipated that P-1 and P-2 show aggregation-annihilated circular dichroism (AACD) effects, while this effect should be restrained in P-3 and P-4 as the 1,1′-binaphthyl moiety is locked by the methylene group. The CD spectra of these polymers in solution and aggregate state were first investigated. All the polymers were molecularly dissolved in tetrahydrofuran (THF). When a poor solvent for the polymer was gradually added in their THF solutions, they started to form aggregates. Then, CD spectra of different polymers at different water fraction (f w ) were measured. All the polymers show a strong CD signal in pure THF solution. However, the CD signals of P-1 and P-2 at f w = 90% are much weaker than those of pure THF solution (Fig. 3a, b). Under the same conditions, the CD signals of P-3 and P-4 also become lower at high f w but the extent is much weaker than in P-1 and P-2 (Fig. 3c, d). As expected, both P-1 and P-2 show an AACD effect, which suggests that the torsion angle θ changes during the aggregation process. Once the binaphthyl ring is locked by the methylene group, the rigid structures of P-3 and P-4 preserve the CD couplet intensity even in the aggregate state.
The effects of water on the CD couplet molar ellipticity [Θ] and wavelength splitting (Δλ max ) have been intensively studied.
[Θ] is a parameter to describe the molar ellipticity and is usually obtained from experiments. Δε is the symbol of molar circular dichroism collected from theoretical calculation. The equation of [Θ] ≈ 3300 Δε indicates that [Θ] is proportional to the CD couplet intensity Δε. The plot of [Θ]/[Θ] 0 against f w is depicted in Fig. 4a, where [Θ] 0 is the molar ellipticity at f w = 0% and the values of CD molar ellipticity were acquired from peaks at 280, 240, 280, and 250 nm for P-1, P-2, P-3 and P-4, respectively. It is clear that the CD signals of P-1 and P-2 are tremendously annihilated. Compared to [Θ] 0 , only 30% of [Θ] remains at f w = 90%. However, for P-3 and P-4, their [Θ] stays at~80% of the initial value even at f w = 90%. These results are in good agreement with our hypothesis that the locked structures of P-3  and P-4 restrict the rotation of the naphthalene rings to suppress the AACD effect. By careful examination of the CD spectra in Fig. 3, it was found that the wavelength splitting (Δλ max ) changes continuously when water is added into the polymer solutions. To make it clear, the plots of Δλ s versus f w are presented in Fig. 4b, where Δλ s = Δλ max -Δλ max-0 and Δλ max-0 = wavelength splitting at f w = 0%. It was found that the Δλ s of P-1 keeps decreasing during the process of aggregation. In contrast, for the "locked" polymers (P-3 and P-4), their Δλ s increases slightly with increasing f w . It is worth mentioning that the positive splitting peak of P-2 is located in a fairly short-wavelength region, which is even shorter than the cut-off wavelength of THF. Therefore, the Δλ s could not be calculated. As the previous study found that the CD couplet intensity and wavelength splitting showed a direct relationship with the torsion angle θ, careful analysis on all these experimental results could provide valuable insights into the conformational change during the aggregation process.

Discussion
To further study the relationship between CD annihilation and the conformational change during the aggregation process, molecular dynamics (MD) simulation of the polymers was performed in THF and water, respectively. Polymers P-1 and P-3 were selected as representatives for the simulation and the details of the calculation are shown in the Methods section. Figure 5a, e show the single P-1 and P-3 molecules in THF and simulated aggregates with 30 P-1 and P-3 molecules in water are demonstrated in Fig. 5b, f. The MD simulation results indicate that P-1 shows a broad distribution of dihedral angle θ but P-3 exhibits a narrow distribution. For example, the full width at half maximum (FWHM) of P-1 is 35°and 45°in THF and water, respectively. However, the "locked" polymer P-3 only possesses a 10°FWHM in both THF and water, suggestive of its higher rigidity. For P-1, the most probable conformation in THF is at θ = −76°. When aggregates are formed in water, the θ of part of the conformation decreases from −76°to −102°though the most probable conformation is still located at~75°. Meanwhile, compared with the most probable conformation in THF, only a 2°increase in θ from −52°to −50°is observed in water for P-3. The maintenance of conformation in P-3 could be attributed to the restriction from the "locked" binaphthyl.
As suggested by the classical model of BN constructed by Mason (Fig. 1), the Δε max and Δλ max decrease simultaneously when θ decreases from −76°to −102°. However, when θ is slightly increased from −52°to −50°, the Δε max decreases but Δλ max becomes higher. Expectedly, the simulated change of Δε max and Δλ max from THF to water coincides with the experimental plots shown in Fig. 4. The conformational change along with molecular aggregation is schematically summarized in Fig. 6. From solution to aggregate, the θ in "open" polymers (P-1 and P-2) becomes more negative and part of the conformers relax from cisoid to transoid. On the contrary, the θ in "locked" polymers (P-3 and P-4) increases slightly and the cisoid conformation is preserved throughout the whole aggregation process.
Study of the λ abs /λ em change could be another tool to monitor molecular aggregation. TPE is a well-studied molecule with  aggregation-induced emission characteristics. Previous studies have shown that the λ abs /λ em of TPE shifts little when its molecules aggregate. So, the relationship between λ abs and β/θ was further investigated by theoretical calculation. A semi empirical CIS/ZINDO method was applied with NStates = 10 and all the calculations were performed using the Gaussian 09 program. The simulation results are shown in Fig. 8 and Supplementary Fig. 10-11. In "open" polymers such as P-1, the λ abs shows almost no change when θ decreases from −80°to −110°. However, an obvious 5 nm hypochromic shift of λ abs is observed when β increases from 26°to 46° (Fig. 8a-c). This indicates that the λ abs / λ em value could be used to monitor the change of β in P-1. However, in "locked" P-3, both β and θ contribute mainly to λ abs . As shown in Fig. 8d-f, decreasing θ from −45°to −55°or increasing β from 32°to 42°results in a 1 nm hypochromic shift of λ abs . This suggests that the change of λ abs /λ em in P-3 should be ascribed to the variation of β and θ. P-2 shows the same effect as P-1, and P-4 and P-3 also exhibit similar behaviors. Then, the PL spectra of these polymers were measured in THF/ water mixtures with different f w (Fig. 9 & Supplementary Fig. 12-13). All the polymers show no or quite weak emission in pure THF solution. However, their emission is tremendously enhanced upon water addition. For example, the PL intensity (I) of P-1 increases by almost 60 times at f w = 90%, demonstrative of an AIE effect (Fig. 8c) [27][28][29] . Similarly, the fluorescence quantum yield (Φ F ) of the polymers also increases upon aggregate formation and their maximum solid-state Φ F reaches 60% (Supple-mentary Fig. 14 and Supplementary Table 6). Except for P-2, all the polymers show an insignificant change of λ em during aggregation (Fig. 9d). For P-1 and P-2, as the decrease of θ was proved to exhibit negligible impact on λ em , so we can conclude that the β value remains almost unchanged from solution to aggregates in P-1. On the other hand, the 14 nm hypsochromic shift of λ em in P-2 suggests an increase of β in the aggregate state. In P-3 and P-4, both θ and β play important roles in the λ em change. However, the λ em stays almost unchanged before and after the aggregation. As suggested by the CD results, only a small increase of θ is observed in P-3 and P-4. We can infer that the β value also suffers small change by aggregation. Among these polymers, it seems that only P-2 shows an obvious increase of β during aggregation presumably due to its crowded structure, and only a slight change of β exists in the other three polymers. In order to verify this conclusion, the distribution of β for P-1 and P-3 obtained from the MD simulation was extracted and the corresponding results are shown in Supplementary Fig. 20-23. Unexpectedly, the dihedral angle β is not fixed both in THF and water. There is an interchange between the conformers with β ≈ 30°, −30°, 150°, and −150°, which indicate the phenyl ring rotation is active in these polymers but with several probable conformers. From the point of view of electronic conjugation, the energy gap almost shows no difference among β = 30°, −30°, 150°, and −150°. Then all the angles of β were transformed and centralized in the range of 0-90°. In this case, the angle with β = 30°, −30°, 150°and −150°could be classified into β = 30°. The processed data are shown in Supplementary Figure 24. From THF to water, the distribution of processed β does not show an obvious change both in P-1 and P-3. Meanwhile, their most probable angle is also similar. For example, P-1 and P-3 shows the highest probability of β at 31°and 32°, respectively, both in THF and water, which is consistent with the PL results. Based on the above analysis, we further conclude that the PL method cannot accurately monitor the conformational change during the process of aggregation, due to the insufficient information provided by the PL spectra. On the contrary, in CD spectra, both molar ellipticity and Davydov splitting could work synergetically to monitor the conformational change, which makes CD spectroscopy a promising tool to monitor molecular aggregation.
In summary, four 1,1′-binaphthyl and TPE-based chiral AIE polymers were designed and synthesized. Polymers P-1 and P-2 with "open" binaphthyl exhibited a typical AACD effect and their CD couplet intensity was largely annihilated in the aggregates.  Fig. 6 Schematic representation of the conformational change of (R)−1,1′binaphthyl moieties during the aggregation process However, the AACD effect was suppressed in P-3 and P-4 as the binaphthyl moieties were locked by methylene. These results indicate that the AACD effect is mainly caused by the twisting of naphthalene rings. The combination of MD simulation and analysis on the change of CD couplet intensity and wavelength splitting during the aggregation process is thus an appealing method for in-situ and real-time monitoring of the conformational change. From the solution to aggregate state, the θ of "open" polymers is largely decreased and part of the conformers relax from cisoid to transoid. On the contrary, the "locked" polymers show a slightly larger θ and their cisoid conformation is preserved during the aggregation process. Thus, CD spectroscopy may be used for real-time and in-situ monitoring of molecular aggregation, especially for biomacromolecular aggregations in physiological conditions. Methods Molecular dynamics simulations. Molecular dynamics simulations were performed using the GROMACS software package (version 5.1.5) 30 . The atom types and parameters of P-1 and P-3 were set from the general amber force field 31 . The electrostatic potential of P-1 and P-3 were calculated by Gaussian 09 package 32 and the corresponding partial charges reproducing the electrostatic potential were obtained by the restrained electrostatic potential fit method 33,34 . To mimic the dilute THF solution of P-1 and P-3, the single P-1 and P-3 molecules were placed in a cubic box of pre-equilibrated THF molecules [35][36][37] , respectively. For each system, the box length was 6.5 nm containing one P-1 or P-3 molecule and 2000 THF molecules (see Fig. 5a, e). To simulate the aggregate state of P-1 and P-3, we first placed 30 P-1 or P-3 molecules in a small cubic box with length 4.5 nm. Then we centered the small cubic box in a large cubic box with length 8.0 nm and solvated by the pre-equilibrated TIP3P water molecules (see Fig. 5b, f) 38 . For all the systems, including single molecule THF solution and aggregates of P-1 and P-3, we first carried out energy minimization, followed by 5 ns NPT (T = 300 K and P = 1 bar) simulations to relax the system. After equilibration, we further ran four independent 50 ns production MD simulations for single P-1 and P-3 molecules in THF solution and carried out 6 independent 50 ns production MD simulations for P-1 and P-3 aggregates in water solution, respectively (Supplementary Fig. 15-18). The temperature and pressure were controlled by the velocity rescaling thermostat 39 ). c, f Plots of λ abs -(λ abs ) 0 versus θ and β. λ abs : simulated absorption maximum, (λ abs ) 0 : simulated absorption maximum wavelength of P-1 at θ = −90°and β = 36°; and P-3 at θ = −50°and β = 37°. θ = torsion angle between two naphthalene rings, β = torsion angle between the naphthalene and phenyl ring of the TPE moiety. A semi empirical CIS/ZINDO method was applied with NStates = 10 and all calculations were performed using Gaussian 09. For the purposes of a clear demonstration and simplified calculation, the hexyl groups are replaced with methyl groups in P-1 reciprocal space summation was evaluated by the particle mesh Ewald method 41,42 . The direct space summation was computed at a cutoff distance of 1.2 nm. The cutoff distance of VdW interactions was 1.2 nm. All bond lengths here were constrained via the LINCS algorithm 43 . Periodic boundary conditions were applied in all three directions to minimize the edge effects in a finite system. The time step here was 2 fs. The configurations were stored at a time interval of 40 ps for statistical analysis of dihedral angles and the corresponding free energy calculation. In total, for single molecule THF solution, we collected 25,000 configurations, while for aggregates, we collected 37,500 conformations. The obtained distributions of the dihedral angle and the Gibbs free energy were demonstrated in Fig. 5 and Supplementary Fig. 19, respectively.
Quantum mechanics calculations. All the quantum mechanism calculations here were performed by the Gaussian 09 program. The optimized conformations of these four polymers were calculated in the gas phase by B3LYP/6-31 G(d). To simplify the calculation, only two repeating units were chosen to do the optimization. Calculation on the CD spectra of BINOL was carried out at CAM-B3LYP/6-31 G* level with Nstates = 30. Calculation on the UV spectra of the polymers was performed with semiempirical CIS/ZINDO method with NStates = 10. The CD and UV spectra were processed by Multiwfn 3.6 Program 44 .