Stereochemistry of furofuran lignans from optical rotatory dispersion: potential and pitfalls

The absolute configuration assignment of chiral secondary metabolites is one of the main bottlenecks in natural product chemistry. Many methods are available to solve this structural puzzle, such as X-ray crystallography, nuclear magnetic resonance, and chiroptical spectroscopy. Among chiroptical spectroscopic methods, optical rotation (OR) and electronic circular dichroism have been by far the most frequently used by natural product chemists. Such a use, however, has been predominantly based on empirical comparisons of data obtained for structurally related molecules. This is a risky approach especially considering optical rotation, which is commonly obtained at a single wavelength (sodium D line). Besides the wavelength range, another important aspect in chiroptical spectroscopy is the effect of solvents on the properties of interest. Recently, we have demonstrated the impact of solvents of different polarities and H-bond capabilities on the vibrational circular dichroism of furofuran lignans. In order to further investigate the solvation effects on the chiroptical properties of this class of secondary metabolites, as well as to explore more reliable protocols to assign absolute configuration based on OR, herein, we present a systematic experimental and theoretical investigation of the optical rotatory dispersion (ORD) of a series of furofuran lignans in CH 3 CN, CHCl 3, and CH 3 OH. Specific rotations were measured at 589, 578, 546, 436, 405, and 365 nm and compared to DFT simulations at the B3LYP/PCM/6–311G++(d,p) level. Comparisons between experimental and calculated data revealed that the absolute configurations of the target lignans could be unambiguously assigned in all cases, but that of sesamin (2) in CH 3 OH solution. The latter case was then investigated using molecular dynamics (MD) and quantum mechanics/molecular mechanics (QM/MM) calculations. Our results indicate that CHCl 3 is the preferable solvent for ORD, while CH 3 OH should be used with caution as it unpredictably affects the simulations of electronic properties of this class of natural products.

Nature has a remarkable ability to produce secondary metabolites with intricate molecular scaffolds by means of specialized biosynthetic machinery. These molecules commonly present multiple stereogenic elements, which gives rise to various possibilities for stereoisomers, including enantiomers (Novak and Trauner, 2020). Additionally, natural products can be produced with variable enantiomeric purities/excesses by the same or different species, which complicates their structural characterization even further (Batista et al., 2018). The stereochemistry of secondary metabolites affects not only their physicochemical properties but also their biological activities and functions in the environment. In fact, stereochemical recognition may be considered the ultimate level of chemical communication in nature (Leal et al., 1995). Therefore, reliable methods to assess the stereochemistry of natural product molecules are constantly in demand.

The absolute configuration of a given molecule can be assessed using different methods. The golden standard for stereochemical studies is single crystal X-ray crystallography. However, although it is considered the most reliable methodology and presents unquestionable importance, it has limitations that can often either delay the analysis of some molecules or, in some cases, preclude it entirely (Harada, 2008; Flack and Bernardelli, 2008). Its application to absolute configuration studies is restricted to substances in the solid state with single crystals of high quality. Another important technique is nuclear magnetic resonance (NMR) spectroscopy. In recent decades, NMR methods have grown in both sensitivity and signal dispersion by the introduction of spectrometers with stronger magnetic fields, leading to relevant instrumental progress. Nevertheless, it is intrinsically achiral in isotropic media resulting in no sensitivity to the differentiation of enantiomers. To be used for this purpose, a chiral auxiliary must be introduced during the experiment such as chiral derivatizing agents, chiral solvation agents, chiral hosting compounds, metal complexes, among others (Wenzel and Chisholm, 2011). As an alternative, chiroptical spectroscopic methods have faced a renaissance in the past decades, following important developments in both instrument and quantum chemical software to predict spectra (Polavarapu, 2007). They are non-destructive spectroscopic methods that investigate different aspects of optical activity directly in solution and without requiring either crystallization or derivatizations. The main chiroptical methods include optical rotation (OR), its wavelength dependent version, optical rotatory dispersion (ORD), electronic circular dichroism (ECD), vibrational circular dichroism (VCD) and Raman optical activity (ROA). All chiroptical methods arise from the differential interaction of a chiral molecule in a non-racemic enantiomeric mixture with left- and right-circularly polarized light beams (Mándi and Kurtán, 2019).

In the natural product structural elucidation pipeline, once the relative configuration of natural compounds is determined, mainly by NMR spectroscopy, optical rotation is the most common chiroptical property used to assess chirality. The most used practice is the comparison of OR signs and values with those of structurally related molecules. Such empirical comparisons are also performed for ECD spectra, when available. Despite the widespread use of OR, it represents simply a number that does not carry any structural information (Batista et al., 2024). Additionally, OR is very sensitive to experimental parameters, such as solvent, concentration, pathlength, and temperature (Polavarapu, 2002). Regarding comparison of structurally related molecules, it has been demonstrated that slight structural changes can result in complete inversion of the optical rotation sign for very similar structures (Nakahashi et al., 2011; Zhou et al., 2016). The only acceptable protocol for empirical comparisons of OR involves the case of the enantiomers of a given molecule. Even in that case, the assignment of the absolute configuration is warranted only if one of the enantiomers has had its absolute stereochemistry secured by independent methods and the comparison is performed under the same experimental conditions.

Among different classes of secondary metabolites, furofuran lignans represent one of the most important subclasses of lignans, being widespread in the plant kingdom. These compounds are formed by the central-carbon fusion of dimers of phenyl-propanoids (C6-C3) and contain the 2,6-diaryl-3,7-dioxabicyclo [3.3.0] octane skeleton (Dewick, 2009). There are numerous furofuran lignans that have been reported in the literature, exhibiting a range of biological activities, including antioxidant, anti-inflammatory, cytotoxic, and anti-microbial effects (Xu et al., 2019). Regarding stereochemistry, although their relative configurations can be easily determined by NMR methods (Shao et al., 2018), the assignments of absolute configurations are more challenging (Li et al., 2025) and may be considered ambiguous in the literature. The reason for that lies on the wide use of empirical correlations of OR and ECD data obtained for structurally related molecules (Biavatti et al., 2001). Recently, VCD has been proposed as a more reliable method to determine both the absolute configuration and main conformations in solution for a series of furofuran lignans (Mestriner et al., 2024; Lee et al., 2025; Lee et al., 2026). The enhanced structural sensitivity of VCD, however, led to additional complications for the quantum chemical simulations of vibrational spectra. Some VCD spectral features in furofuran lignans were demonstrated to require the inclusion of explicit solvent molecules, even in the absence of direct H-bond interaction with the solute as in acetonitrile. In the case of the lignan sesamin, explicit solvation was unprecedently required for chloroform solutions as well. A computational protocol combining molecular dynamics (MD) simulations followed by quantum mechanics/molecular mechanics (QM/MM) calculations proved particularly effective for both acetonitrile and chloroform solutions. The presence of explicit solvation shells stabilized conformations that led to effective orientation and, consequently, coupling of the aromatic ring vibrational modes, which was essential for the reproduction of the couplet-like signal at around 1,510 cm -1 observed for all lignans. The effects of solvent, however, were found to vary depending on structural features such as flexibility and symmetry elements. It was demonstrated that the inclusion of explicit solvation becomes necessary for VCD simulations as both molecular symmetry and flexibility increase.

Despite a significant increase in the number of natural product chemists using VCD over the last decade (Batista et al., 2024), it is still unknown to a significant part of the community. Additionally, the above-mentioned solvent effects seemed to be less pronounced in the electronic optical activity method ECD. Therefore, in order to further investigate the solvation effects on the chiroptical properties of furofuran lignans, as well as to explore more reliable protocols to assign absolute configuration based on OR, herein, we present a systematic experimental and theoretical investigation of the ORD of a series of furofuran lignans (Figure 1) in CH 3 CN, CHCl 3, and CH 3 OH. The advantages and limitations of this specific rotation-based approach will be investigated with the aid of quantum chemical calculations and discussed in terms of solvent effects, conformational preferences and symmetry. Extensive experimental and theoretical investigations of the ORD of chiral molecules of various classes have already been reported in the literature (Grimme, 2001; Stephens et al., 2001; Stephens et al., 2002; Stephens et al., 2003; Norman et al., 2004; Crawford et al., 2005; Stephens et al., 2005; Polavarapu et al., 2005; Crawford et al., 2007; Rudolph and Autschbach, 2011; Srebro et al., 2011; Lambert et al., 2012; Haghdani et al., 2016; Carrano et al., 2021) and, although the effects of different solvents, as well as methods and levels of theory for the calculations have been investigated (Mennucci et al., 2002; Giorgio et al., 2005; Mukhopadhyay et al., 2006; Mennucci et al., 2011; Lahiri et al., 2013; Alenaizan et al., 2017; Howard and Crawford, 2018; D'Cunha and Crawford, 2021), no such a study has been performed for furofuran lignans. Given the widespread use of empirical comparisons for stereochemical assignments of lignans, the results presented herein will contribute to safer applications of OR to this class of secondary metabolites.

Plant material and lignan isolation

Compounds 1 and 3 were isolated from Aristolochia gigantea (Aristolochiaceae), compound 2 from A. pubescens, compounds 4 and 6 from A. warmingii, and compound 5 from A. galeata. The procedures for their isolation and structural identification have been previously reported in the literature (Lopes and Bolzani, 1988; Nascimento and Lopes, 1999; Holzbach and Lopes, 2010; Cunha et al., 2023). The enantiomeirc purities of compounds 1–6 were determined by chiral HPLC and reported previously (Lee et al., 2025; Lee et al., 2026).

Experimental

ORD was measured for compounds 1–6 in chloroform, acetonitrile, and methanol, at 589, 578, 546, 436, 405, and 365 nm, using a Jasco P2000 polarimeter equipped with sodium and mercury lamps, along with appropriate filters. Sample concentration was kept at 0.25 mg mL -1, and the optical rotation was recorded using a cylindrical glass sample cell of 1 dm pathlength and 10 mm ID. Specific rotation values were reported as a mean of 3–5 measurement blocks of 5 repetitions each.

Calculations

Calculations were performed for arbitrarily chosen configurations as described in Figure 1. The chiroptical properties of their enantiomers were obtained by multiplying the theoretical spectra by (−1). All conformational searches were initially carried out at the molecular mechanics level of theory employing a MMFF force field incorporated in the Spartan 08 software package. The DFT calculations were carried out at 298 K in either chloroform, acetonitrile or methanol solutions using the polarizable continuum model (PCM) in its integral equation formalism version (IEFPCM), incorporated in Gaussian 09 software (Revision A.02) (Frisch et al., 2009). Conformers with relative energy (rel E.) within 10 kcal mol -1 of the lowest energy conformer were selected and further geometry was optimized at the B3PW91/PCM/6–311G(d,p) level. ORD properties were calculated at 589, 578, 546, 436, 405, and 365 nm at the B3LYP/6–311G++(d,p) (Polavarapu, 2002) level in chloroform, acetonitrile, and methanol solutions using PCM. Discrete specific rotation values at each wavelength of individual conformers (rel E. ≤ 2.0 kcal mol -1) were Boltzmann averaged using B3PW91/PCM/6–311G(d,p) Gibbs free energies. Vibrational analysis at the level used for geometry optimizations resulted in no imaginary frequencies for all conformers, confirming them as real minima. Molecular dynamics (MD) simulations were performed using GROMACS software (Berendsen et al., 1995) (version 2024.1) with periodic conditions with explicit CH 3 OH for 2. A cubic box of 50 × 50 × 50 Å dimensions was used and filled with a single lignan molecule and 1,600 molecules of CH 3 OH. Simulations were performed at 300 K and 1 bar using the V-rescale thermostat and the Berendsen barostat for 20 ns with an integration time step of 2 fs. OPLS-AA parameters (Jorgensen et al., 1996) were used with a cut-off radius of 1.2 nm for both Coulomb and Lennard-Jones interactions with PME (particle-mesh Ewald) (Essmann et al., 1995) correction for long-rang Coulomb interactions. Snapshots were taken at every 50 ps and subjected to geometry optimization using the two-layer ONIOM method. Only solvent molecules within a 7–8 Å distance from any atom of the lignan were kept (70 solvent molecules in average). The high layer/QM region was treated at the B3PW91/6–311G(d,p) level of theory for geometry optimizations and harmonic frequency calculations. The low layer/MM CH 3 OH region was treated with the UFF force field without electronic embedding. After geometry optimization using the QM/MM ONIOM method, the low layer was deleted and only the lignans (high layer) were subjected to ORD calculations at the B3LYP/PCM(MeOH)/6–311G++(d,p) and CAM-B3LYP/SMD(MeOH)/aug-cc-pVDZ (Carrano et al., 2021) levels. Final chiroptical data were obtained by simple averages of individual snapshots taken over the last 5 ns. The inclusion of individual solvent molecules (micro-solvation) was performed using the GaussView software. Solvent-solute complexes were subjected to geometry optimization, frequency and ORD calculations as described above.

ORD overcomes some of the limitations of the well-established and simplified specific rotation method, widespread within the natural product community, by recording optical rotations in different wavelengths. This leads to a dispersion curve in the non-resonant region as opposed to the single signed value commonly reported. The comparison of such a curve with DFT-based simulations allows for a more reliable assignment of the absolute configuration of the target compounds (Giorgio et al., 2005). Specific rotation (and consequently ORD), however, is sensitive to a series of experimental parameters, such as concentration, path length, solvent and temperature. Although the effect of changing the concentration, path length and temperature on the optical rotation is somehow predictable, the effect of solvent is more complicated to anticipate and control as it may affect conformations, inter- and intramolecular interactions, as well as electronic and vibrational transitions. Therefore, in this work ORD data for all furofuran lignans were recorded in three different solvents, CH 3 CN, CHCl 3, and CH 3 OH. Acetonitrile has already been described as the solvent of choice for concerted analyses of OR, ECD, and VCD of furofuran lignans as it allows for larger spectral windows (Lee et al., 2025). Nevertheless, its solubility properties and the requirement of explicit solvation to reproduce the chiroptical data recorded in this solvent may limit its applicability. In natural product chemistry, methanol and chloroform are far more common for spectroscopic analysis of furofuran lignans. The comparison between experimental and calculated ORD for all target lignans in three different solvents are presented in Figure 2.

From the data presented in Figure 2, it is possible to assign the absolute configuration of each compound reliably. For the dextrorotatory compounds 1–3 and 5, the absolute configuration arbitrarily chosen for the calculations were opposite to that of the natural compounds. This way, compounds 1–3 could be assigned as (7 S,8 R,7′ S,8′ R), while 5 was assigned as (7 S,8 R,7′ R,8′ R). Regarding the levorotatory compounds 4 and 6, the absolute configurations used for the calculations were identical to those of the isolated compounds leading to their assignments as (7 R,8 S,7′ S,8′ S). For each compound, the tendency of dispersion and magnitude of the specific rotation values were comparable between experiments and calculations. The errors of experimental and calculated OR data at each wavelength and solvent are presented in Supplementary Tables S1-S6. Overall, experimental data across different solvents resulted in similar magnitudes, which compared better to calculations in CHCl 3 by a small margin for compounds 2–4. On the other hand, calculated dispersion data in CH 3 CN and CH 3 OH presented better correspondence between each other. These observations are in accordance with literature results (Mennucci et al., 2002). Regarding experimental data in different wavelengths, data recorded in CHCl 3 presented better dispersion properties and less variation for all lignans due to lower saturation of the photomultiplier tube detector. This behavior may also be linked to smaller conformational perturbation induced by the less polar solvent. Thus, CHCl 3 represents the solvent of choice for experiments and calculations of specific rotation of furofuran lignans. Even though the ORD in the non-resonant region is expected to result in smooth dispersion curves with no anomalous behavior, some of the experimental specific rotations in CH 3 CN and CH 3 OH for compounds 1, 4 and 6 presented important deviations. This seemed to be more critical at lower wavelengths as both solvents led to higher counts at the photomultiplier tube detector even at the low concentrations used in the experiments. Regarding molecular structure, compounds 1, 4 and 6 differ from 2, 3 and 5 mainly on the flexibility of the aromatic ring substituents. It is noteworthy that compound 6 is the only capable of acting as a H-bond donor. The combination of the electronic properties (refractive index and/or polarizability) of the solvents (Alenaizan et al., 2017) and higher conformational flexibility of the lignans may help explain the experimental deviations observed.

Despite the overall good agreements between experimental and calculated ORD data for the lignans in different solvents, it is noteworthy that for sesamin (2) the simulated data in CH 3 OH solution using PCM led to the assignment of the dextrorotatory compound as (7 R,8 S,7′ R,8′ S). This result is in contrast with the data obtained in both CH 3 CN and CHCl 3, as well as with literature results (Lee et al., 2026). The theoretical dispersion curve for sesamin in CH 3 OH solution shown in Figure 2 was created by Boltzmann averaging the lowest-energy conformers identified. Upon assigning equal weights for the identified conformers (simple average), however, the dispersion curve adopted a negative trend, in line with the results obtained for the other solvents. This behavior may indicate problems with the selection of the conformational ensemble either due to inaccurate geometries or incorrect energy estimations. These latter properties are known to be directly affected by the effect of solvents. Experimental specific rotation sign inversions upon changing solvents are relatively common (Mukhopadhyay et al., 2006; Batista Jr et al., 2015; Das et al., 2024), and these variations are frequently well reproduced by quantum chemical calculations by means of PCM, despite challenging cases (Howard and Crawford, 2018). Interestingly, in the case of selected rigid and flexible small chiral cyclic molecules, solvents such as CH 3 OH (and CH 3 CN) have been demonstrated to provide the best correlations between experimental and calculated ORD data (Mennucci et al., 2002; Lahiri et al., 2013). In the case of (+)-sesamin, however, experimental dispersion curves remained consistently positive across the three different solvents, and only the calculations in CH 3 OH failed to reproduce the correct trend for the (7 R,8 S,7′ R,8′ S) configuration. A similar situation was reported for (S)-(+)-carvone (Lambert et al., 2012) for which optical rotations calculated in CH 3 OH at different wavelengths resulted in opposite signs for the same configuration. The authors considered that further vibrational corrections would need to be considered to bring solvent phase calculations into agreement with experimental results. Nevertheless, even though zero-point vibrational corrections to optical rotations are sizable, including vibrational effects alone is in general not sufficient to improve agreement with experimental solution data (Mort and Autschbach, 2005; Srebro et al., 2011). A combination of vibrational and mainly solvent effects is expected to result in better agreements between theory and experiment.

Accordingly, we proceeded with a different protocol to try to better account for the effects of solvents on the ORD simulations. To that end, a combination of MD simulations and QM/MM calculations was employed. Due to convergence problems, however, ORD data was simulated using the geometries optimized with the QM/MM ONIOM approach (Lee et al., 2025; Lee et al., 2026), but without the inclusion of explicit solvent molecules; PCM was used instead. Although a more realistic approach was employed, which included explicit solvent molecules during geometry optimization steps, the ORD data remained largely positive for the (7 R,8 S,7′ R,8′ S) configuration of 2 in CH 3 OH solution (Figure 3). ORD calculations were performed for 100 snapshots, corresponding to the last 5 ns of the 20 ns long MD run. The convergence of the MD trajectory was verified by the analysis of the Root Mean Square Deviation (RMSD) and Solvent Accessible Surface Area (SASA) presented in Supplementary Figures S1, S2. Specific rotations were initially simulated at the B3LYP/PCM(MeOH)/6–311G++(d,p) level. Additional calculations were also performed at the CAM-B3LYP/SMD(MeOH)/aug-cc-pVDZ level, which includes a long-range corrected functional, a widely used Dunning basis set, and the SMD variant of PCM that gave the smallest deviations compared to experiment for aziridines and oxirane derivatives (Carrano et al., 2021). This latter method resulted in largely similar dispersion curves to those presented in Figure 3 for the (7 R,8 S,7′ R,8′ S) configuration of 2 in CH 3 OH solution (Supplementary Figure S3). A similar approach combining MD and ORD calculations was reported previously in water, methanol and carbon tetrachloride (D'Cunha and Crawford, 2021). In their work, however, D’Cunha and Crawford included up to 12 solvent molecules in the solvent shell of the snapshots subjected to TDDFT calculations. According to the authors, despite the extensive computational effort invested in the above protocol and the careful convergence of the methodology with respect to each degree of freedom, the comparison with experimental solution-phase data remained inadequate for the relatively smaller compound (S)-methyloxirane. These findings reinforce the challenges of correctly reproducing the OR of chiral molecules in a solvent environment.

In another attempt to accurately simulate the experimental data, we tested the inclusion of solvent molecules in the quantum mechanical Hamiltonian by treating the solvation shell in the same level of theory as the solute. Full individual MD snapshots, however, would be untreatable considering the computational power at our disposal. Even the reduced solvent shell employed by D’Cunha and Crawford would be prohibitive for us. As an alternative, the method described as micro-solvation (Merten, 2023) was used. Although micro-solvation still results in a higher computational cost as compared to hybrid approaches, it may be feasible within the computational resources available depending on the number of solvent molecules to be included. For this step, the lowest-energy conformer of 2 obtained using the classic approach was selected. This conformer with (7 R,8 S,7′ R,8′ S) configuration presented positive optical rotation in CH 3 OH solution for all simulated wavelengths (Figure 4). Inclusions of 1 up to 4 individuals explicit CH 3 OH molecules were tested. Solvent molecules were positioned as to favor H-bond interactions with the oxygen atoms of the methylenedioxy and/or furofuran groups (Figure 4). After inclusion of solvent molecules, the structures were re-optimized at the B3PW91/6–311G(d,p) level. Subsequently, the solute-solvent complexes were subjected to ORD calculations with and without PCM. The results of ORD simulations demonstrated that the inclusion of explicit solvent molecules had an almost null effect on the specific rotation of the target molecule, even though important structural changes were observed. Changes from twisted to flat furofuran moieties were observed upon including explicit solvent molecules and, while largely affecting VCD properties (Lee et al., 2025; Lee et al., 2026), the ORD values remained largely positive (Figure 4), with slightly smaller rotations for PCM simulations. The dominant structural factor for specific rotation sign inversion seemed to be the relative orientation of the dipole moments of the identical aromatic rings. Such orientations and the resulting energy changes seemed to be overestimated when geometry optimizations and frequency calculations were performed using implicit solvation. Since this behavior was observed only for CH 3 OH solution, its use is discouraged for optical rotation measurements of furofuran lignans whenever possible, especially if absolute configuration assignments are intended. The difficulties in simulating solvent effects in this class of secondary metabolites also render empirical comparisons of optical rotation values for structurally correlated molecules a very risky approach.

Solvation plays a crucial role on the chiroptical properties of chiral molecules in general (Merten, 2023) and furofuran lignans in particular (Lee et al., 2025; Lee et al., 2026). Herein, we demonstrated that the electronic optical activity of furofuran lignans is affected by the polarity and H-bond propensities of different solvents. For the first time, the effect of CH 3 CN, CHCl 3, and CH 3 OH was investigated on the ORD properties of a series of furofuran lignans (1–6). Comparisons between experimental and calculated data revealed that the absolute configurations of the target lignans could be unambiguously assigned in all cases, except for sesamin (2) in CH 3 OH solution. Calculations for the (7 R,8 S,7′ R,8′ S) configuration resulted in a positive dispersion curve leading to the assignment of the dextrorotatory (+)-2 as (7 R,8 S,7′ R,8′ S), in contrast to the results in CH 3 CN and CHCl 3, as well as literature results. Even after MD simulations and the inclusion of explicit solvation in the geometry optimization steps, the theoretical ORD values remained largely positive for (7 R,8 S,7′ R,8′ S)-2. Therefore, although frequently applied in the isolation and structural characterization of furofuran lignans, CH 3 OH should be used with caution for specific rotation. Given that empirical comparisons of specific rotation for structurally related molecules is a common practice in natural product chemistry for stereochemical assignments, these results are far more reaching. The solvent CHCl 3 is the preferable choice for ORD as it leads to higher intensities, lower intermolecular interactions, and improved reproducibility. If solubility is an issue, then the use of CH 3 CN is recommended. Finally, obtaining OR in different solvents is recommended for more reliable assignments.