Karolina Di Remigio
Eikås
*a,
Monika
Krupová
a,
Tone
Kristoffersen
b,
Maarten T. P.
Beerepoot
b and
Kenneth
Ruud
*ac
aHylleraas Centre for Quantum Molecular Sciences, Department of Chemistry, UiT The Arctic University of Norway, 9037 Tromsø, Norway. E-mail: karolina.s.eikas@uit.no; kenneth.ruud@uit.no
bDepartment of Chemistry, UiT The Arctic University of Norway, 9037 Tromsø, Norway
cNorwegian Defence Research Establishment, P.O. Box 25, 2027 Kjeller, Norway
First published on 16th May 2023
Cyclic peptides show a wide range of biological activities, among others as antibacterial agents. These peptides are often large and flexible with multiple chiral centers. The determination of the stereochemistry of molecules with multiple chiral centers is a challenging and important task in drug development. Chiroptical spectroscopies such as vibrational circular dichroism (VCD) can distinguish between different stereoisomers. The absolute configuration (AC) of a stereoisomer can be determined by comparing its experimental spectra to computed spectra of stereoisomers with known AC. In this way, the AC of rigid molecules with up to seven chiral centers has been assigned (Bogaerts et al., Phys. Chem. Chem. Phys., 2020, 22, 18014). The question arises whether this is possible with more conformationally flexible molecules such as cyclic peptides. We here investigate to what extent the AC of cyclic peptides can be determined with VCD. More specifically, we investigate the maximum number of chiral centers a cyclic peptide can have in order to be able to unambiguously assign the AC with VCD. We present experimental and computed IR and VCD spectra for a series of eight tetrapeptides and hexapeptides with two, three and four chiral centers. We use our recently developed computational protocol with a conformational search based on sampling with meta-dynamics. We use visual inspection to compare the computed spectra of different stereoisomers with an experimental spectrum of the corresponding cyclic peptide with known AC. We find that the AC of the investigated cyclic peptides with two chiral centers can be unambiguously assigned with VCD. This is however not possible for all of the cyclic peptides with three chiral centers and for none of those with four chiral centers. At best, one can limit the number of possible stereoisomers in those cases. Our work shows that other techniques are needed to assign the AC of cyclic peptides with three or more chiral centers. Our study also constitutes a warning that the spectra of all stereoisomers should be computed before attempting to match to an experimental spectrum, to avoid an accidental erroneous match.
Various methods of chiroptical spectroscopy can be used to determine the chirality of organic compounds.7–9 Vibrational circular dichroism (VCD)10 has recently been extensively used in the pharmaceutical industry to determine the AC of chiral drugs.8,11,12 While IR measures the absorption of infrared light, VCD measures the difference in absorption of left- and right-circularly polarized light in the infrared energy range. Since this differential absorption varies for enantiomers, VCD, in contrast to IR, can distinguish between enantiomers and help to determine the AC of chiral compounds, since two enantiomers have equal VCD spectra with opposite sign. The situation gets more complicated, however, for molecules with more than one chiral center where one distinguishes between enantiomers and diastereomers. Enantiomers have different stereochemistry in all chiral centers, are mirror images of each other and have equal VCD spectra with opposite signs. Diastereomers have different stereochemistry in at least one but not all chiral centers and may have VCD spectra that are more different than only a sign change.
Unambiguous assignment of the stereochemistry for a range of both synthetic and natural chiral compounds has been done using VCD.8,13–15 In most cases where VCD alone has been used to determine the stereochemistry, the compounds were limited to a maximum of three chiral centers.16 However, a recent work by Bogaerts et al. demonstrated that VCD could be used to unambiguously assign the AC of artemisinin and artesunate with six and seven chiral centers, respectively.17 Another success was achieved by Koenis et al., who could determine AC and diastereomeric contamination levels as low as 5 percent for a molecule with six chiral centers, dydrogesterone.15 This was facilitated by the fairly rigid structure of these molecules. The situation becomes more complex and challenging for conformationally flexible systems such as cyclic peptides,18 where VCD alone may not be successful in an unambiguous AC determination. For cyclic peptides, an additional complication may arise from the fact that the dominating VCD bands of these molecules are amide modes in which the chiral centers are not significantly involved, the bands instead reflect the secondary structure of the peptides. Moreover, the VCD spectral signatures may be influenced by the nature of the solvent,19,20 which places further demands on the calculations needed to interpret the experimental spectra. Exploring whether the VCD spectra reflect the nature of the chiral centers or the secondary structure for cyclic peptides of increasing size is therefore of importance in order to assess the predictive power of VCD for this class of compounds.
To assign the unknown AC of a chiral compound, a measured VCD spectrum is compared with computed spectra for different stereoisomers of the compound. A computational protocol to simulate the spectra typically consists of a conformational search and quantum-chemical calculations of the VCD spectra.8,18 Since the spectra are measured in solution, the solvent environment is often included in calculations through the use of a continuum model.21,22 The conformational search is a key step in the computational protocol for flexible molecules, since different conformers with the same AC may provide VCD intensities with different signs in the same frequency range. When relevant conformers are missing, this can easily lead to an erroneous assignment of the AC. In a previous work,18 we established a reliable computational protocol for the calculation of conformationally flexible molecules with known AC, using several cyclic peptides as test systems. This computational protocol was based on a meta-dynamics conformational search as implemented in CREST,23–25 combined with energies, geometries and VCD spectra obtained using DFT. We showed that the protocol allows for the identification of relevant conformations, and that the Boltzmann-averaged spectrum agrees well with the corresponding experimental spectrum.
In this work, we investigate to what extent the AC of cyclic peptides can be determined with VCD using this computational protocol. More specifically, we investigate the maximum number of chiral centers a cyclic peptide can have in order to be able to unambiguously assign the AC with VCD.
With the exception of β-alanine in 1 and glycine in 2–6 and 8, all amino acids have one chiral center. The AC of a chiral peptide is indicated by naming the stereochemistry of each chiral amino acid, following the order of the amino acids. For example, SR-cyclo(Ala-β-Ala-Pro-β-Ala) (1) indicates that alanine has the S configuration and proline has R wheres as RSR-cyclo(Val-Gly-Ser-Gly-Ala-Gly) (5) indicates that the chiral amino acids are R-valine, S-serine and R-alanine.
The experimental spectra for tetrapeptides 1,26419 and 719 as well as for the hexapeptide 827 were taken from the literature. Cyclic hexapeptides 2, 3, 5 and 6 were synthesized and their VCD spectra measured as part of this study.
Final VCD spectra were generated as Boltzmann averages based on free energies for the unique conformers by the DrawSpectrum program.39 The calculated spectral intensities are scaled such that the peak with the highest intensity in the computed spectrum matches the peak with the highest intensity in the experimental spectrum. For the molecules measured in this work (2, 3, 5 and 6), a Lorentzian bandshape with a full width at half maximum (FWHM) of 10 cm−1 was applied. For the molecules measured elsewhere, the broadening was chosen to be the same as used for the calculated spectra in those works: 10 cm−1 for 1 measured by Vass et al.,26 16 cm−1 for 4 and 7 measured by Merten et al.19 and 10 cm−1 for 8 measured by Bouř et al.27
Since the stereoisomers occur in pairs of enantiomers that have equal IR spectra and equal but opposite-signed VCD spectra, we have reduced the computational workload by calculating the spectra for only one enantiomer in each pair. Thus, we applied the computational protocol including the conformational search for two diastereomers of the cyclic peptides with two chiral centers; for four diastereomers of the cyclic peptides with three chiral centers; and for eight diastereomers of the cyclic peptides with four chiral centers.
In general, frequencies calculated with DFT do not match the experimentally observed frequencies15,40–42 and they are usually scaled to obtain a better agreement with experiment. To our knowledge, there is no established method to determine the frequency scaling factors for spectra recorded in solvent and several ad hoc schemes are in use. One way is to choose the scaling factor that gives the best visual agreement with the experiment. In this work, however, we compute the overlap between the experimental spectrum and spectra computed with different frequency scaling factors and choose the frequency scaling factor giving the largest combined VCD and IR overlap between experimental and computed spectra. Overlap integrals Sspec are calculated with DrawSpectrum39 as43–46
(1) |
As discussed in our previous work,18 the experimental gap between the amide I and II spectral regions is poorly reproduced in computed spectra of cyclic peptides. Therefore, the same frequency scaling factor cannot be used for both spectral regions to calculate the overlap integral of both regions combined. Hence, we have chosen to compute overlap integrals for the amide I region (1800–1600 cm−1) and amide II region (1600–1400 cm−1) separately and use two different scaling factors for the two regions. For 8, the experimental spectra is measured from 1800 cm−1 to 1500 cm−1 and the amide II region for this molecule is 1600–1500 cm−1.
For each region, the scaling factor that gives the highest value for combined VCD and IR overlap given as
(2) |
To mimic a situation in which an entirely unknown AC is assigned, scaling factors were determined individually for all stereoisomers of each chiral peptide. This makes chemical sense due to their different specific solute–solvent interactions, but only for different diastereomers. However, without prior information on the AC, the same procedure has to be followed for the two stereoisomers in an enantiomeric pair, to avoid optimizing the frequency scaling factor of the correct stereoisomer to the VCD spectrum of its enantiomer. Hence, frequency scaling factors were also determined individually for enantiomers, resulting typically in positive overlap integrals for both enantiomers and VCD spectra that are not mirror images of each other. For each stereoisomer, scaling factors that give a negative SVCD are discarded. In cases where all investigated scaling factors lead to a negative SVCD, the spectra are not scaled. The resulting frequency scaling factors are collected in Tables S1–S3 (ESI†).
One could in theory use overlap integrals to determine which stereoisomer matches the experimental VCD spectrum best. Indeed, Bogaerts et al. used overlap integrals between calculated and experimental VCD and Raman optical activity (ROA) spectra to determine the AC of two rigid compounds with six and seven chiral centers.17 Koenis et al. used spectral overlap between calculated and experimental VCD to show that the AC of a rigid molecule with six chiral centers can be determined with VCD only.15 For the molecules investigated in this work, we found that the overlap integrals are not reliable enough to be used for this purpose. The overlap integrals calculated using eqn (1) are extremely sensitive to small changes in frequencies. Indeed, a small change in the frequency scaling factor may change the AC assignment based on the overlap integral. In addition, the largest overlap integrals may correspond to a stereoisomer that shows an entirely different pattern in the VCD spectrum and which would therefore immediately be discarded by the trained eye. In Fig. 2, the experimental VCD spectrum (SS) and the calculated spectra of all stereoisomers of 3 are shown together with the respective overlap integrals. By solely looking at the overlap integrals, the RS stereoisomer is clearly in best agreement with the experiment. However, by looking at the pattern of the amide I region of the spectra, the RS stereoisomers has a −/+/− pattern which does not correspond to the experimental +/− pattern.
Fig. 2 Experimental and computed (B3LYP/6-31+G*/CPCM) VCD spectra (left) and overlap integrals (right) of 3 in DMSO-d6. Different scaling factors are used for the amide I and amide II regions, see Table S1 (ESI†). The experimental spectrum is overlaid with the computed spectra as a dotted black line. |
Thus, we included all possible stereoisomers in our analysis and use visual inspection to determine which computed spectrum/spectra match the experimental spectrum best.
We primarily use the experimental pattern in the intense amide I region to select the matching computed spectrum. If this is not possible or when several computed spectra match the pattern in the amide I region, we also include the amide II region in this analysis. Another source of information is the relative intensity of the amide I and amide II regions. We do not take into account the separation between the amide I and II regions, as it is likely that explicit solvent modelling is needed to correctly reproduce this separation.18,27
Fig. 3 Experimental and computed (B3LYP/6-31+G*/CPCM) IR (top) and VCD spectra (bottom) of 1 in ACN-d326 and 2 and 3 in DMSO-d6. Different scaling factors are used for the amide I and amide II regions, see Table S1 (ESI†). The experimental spectrum is overlaid with the computed spectra as a dotted black line. |
For tetrapeptide 1, the experimental VCD spectrum (SS) is in fair agreement with calculated spectra for both the SS and SR stereoisomers in the amide I region. However, the experimental positive band slightly below 1450 cm−1 is only reproduced for the SS stereoisomer. In this particular case, the IR spectrum can help in assigning the AC. Indeed, the experimental absorption band around 1625 cm−1 is reproduced as a third peak in the amide I region only for the calculated spectra of the SS and RR stereoisomers. For hexapeptide 2, the experimental (SS) +/+/− pattern in the amide I region is reproduced only in the computed VCD spectrum of the SS stereoisomer.
For hexapeptide 3, the experimental (SS) +/− pattern in the amide I region is similarly reproduced only in the computed VCD spectrum of the SS stereoisomer.
In summary, the AC of the three investigated cyclic peptides with two chiral centers can be determined by VCD. For tetrapeptide 1, the calculated SS spectrum reproduces the experimental pattern in the entire investigated frequency range (1800–1400 cm−1), whereas for hexapeptides 2 and 3, only the amide I region in the experimental spectra is reliably reproduced by the calculations. This may be caused by the longer and more flexible peptide backbone in 2 and 3 compared to 1.
Fig. 4 Experimental and computed (B3LYP/6-31+G*/CPCM) IR (top) and VCD spectra (bottom) of 4 in ACN-d319 and 5 and 6 in DMSO-d6. Different scaling factors are used for the amide I and amide II regions, see Table S2 (ESI†). The experimental spectrum is overlaid with the computed spectra as a dotted black line. |
For tetrapeptide 4, the experimental (RSR) −/+ couplet in the amide I region as well as the negative band slightly above 1500 cm−1 are reproduced in the computed VCD spectra of RSR, SRR, and RRR stereoisomers. However, the intensity of the amide I region in the computed spectrum of the SRR stereoisomer is too low when compared to the experiment. The relative intensities of bands in the amide I and II regions are reasonable for the RRR stereoisomer, but the negative band in experimental VCD spectra around 1450 cm−1 has a −/+ couplet shape in the computed spectra. Thus, the correct RSR stereoisomer matches best with the experimental VCD spectrum. For hexapeptide 5, the experimental (SSS) +/− couplet in both the amide I and II regions is reproduced only in the computed VCD spectrum for the SSS stereoisomer. For hexapeptide 6, the experimental (SSS) +/− couplet in the amide I region is reproduced in the computed VCD spectrum of both the SSR and SSS stereoisomers. The experimental +/− couplet in the amide II region is reproduced better for the SSR stereoisomer than for the SSS stereoisomer, which shows a +/−/+ spectral pattern. Based on the amide II region, one could erroneously assign SSR as the AC, or conclude that the AC cannot be reliably assigned with VCD only.
In summary, the AC of two out of three cyclic peptides with three chiral centers can be correctly assigned with VCD.
Fig. 5 Experimental and computed (B3LYP/6-31+G*/CPCM) IR (top) and VCD spectra (bottom) of 7a in ACN-d3,197b in ACN-d319 and 8 in TFE.27 Different scaling factors are used for the amide I and amide II regions, see Table S3 (ESI†). The experimental spectrum is overlaid with the computed spectra as a dotted black line. |
For tetrapeptide 7a, the experimental (RSSR) −/+ couplet in the amide I region is reproduced in the computed VCD spectra of six stereoisomers. Two of these can be eliminated due to a negative band between 1800–1700 cm−1 that is not observed in the experimental spectrum (RRSS and SSSS). For the four remaining stereoisomers, only calculated spectra for the RSSR and RSSS stereoisomers reproduce the +/− couplet in the amide II region. At best, the determination of the AC can be limited to two options. For tetrapeptide 7b, the experimental (RSRR) −/+ couplet in the amide I region is reproduced in the computed VCD spectra of six stereoisomers. The correct RSRR stereoisomer is not among those six, showing a positive band at 1700 cm−1 that is not observed in the experiment. The RSRR stereoisomer is, however, the only one reproducing the experimental −/+ couplet in the amide II region. For hexapeptide 8, the experimental (SRSS) −/−/+ spectral pattern in the amide I region is reproduced in the computed VCD spectra of three stereoisomers: SRSS, RRSS and RRSR. The experimental spectrum does not show an intense signal in the amide II region, which is in line with computed spectra of these three stereoisomers. At best, the determination of the AC can be limited to three options.
In summary, the AC cannot be correctly assigned with VCD for any of the three investigated cyclic peptides with four chiral centers.
Although somewhat disappointing, it is not surprising that VCD alone cannot always determine the AC of molecules with three or more chiral centers. Other works have been successful in determining the AC of such molecules and have achieved this by combining VCD with other types of spectroscopies. Hopmann et al. combined VCD, ROA and NMR to determine the AC of a highly flexible natural product with two chiral centers and an asymmetrically substituted double bond.48 Dinku et al. combined VCD and NMR to determine the AC of a tricyclic triterpene acid with three chiral centers and three asymmetrically substituted double bonds.49 Bogaerts et al. combined VCD and ROA to determine the AC of two rigid compounds with six and seven chiral centers.17 Thus, combining VCD with other techniques such as ROA and NMR is a preferred strategy to correctly and reliably determine the AC of large and flexible molecules with multiple chiral centers.8,14 We hypothesize that this strategy also works for cyclic peptides 6–8, for which the AC could not be assigned with VCD only. Although we have shown that VCD independently can determine the correct AC for cyclic peptides 1–5, a more unambiguous assignment can be made by combining several techniques.
Our study also shows that in the absence of additional experimental data, it is important to calculate spectra for all possible stereoisomers of a flexible molecule with multiple chiral centers, to avoid an erroneous assignment based on an accidental agreement. Such an erroneous assignment becomes more probable with a higher number of chiral centers in a molecule.
Footnote |
† Electronic supplementary information (ESI) available: Additional IR and VCD spectra of 7a, 7b and 8. Detailed description of peptide synthesis and purification for 2, 3, 5 and 6. NMR spectra of the peptides synthesised in this work: 2, 3, 5 and 6. See DOI: https://doi.org/10.1039/d2cp04942b |
This journal is © the Owner Societies 2023 |