Introduction

The molecules important in biology are remarkable, considering that they can display a wide variety of structures and functions. Most of the interesting functions carried out by these biomolecules are dependent on stable well-defined structures. This factor can be considered in the development of unnatural backbone modified peptides. β-amino acids [1] belong to such a group. The additional carbon atom incorporated into the backbone structure in natural α-amino acids results in unnatural oligomers containing side chains, which are significantly different from that found in natural compounds, but that can be recognized by side chain of receptors. The title molecule, phenylisoserine (Scheme 1), is one of such examples.

Scheme 1
scheme 1

 Phenylisoserine

Full size image

Among the new families of anticancer agents the natural product paclitaxel (Taxol®, Scheme 2) and its derivative docetaxel (Taxotere®) are used in the treatment of breast, ovarian cancers and lung carcinomas [2–4]. Paclitaxel is a natural product isolated from the bark of the western pacific yew tree. It contains 10-deacetylbaccatin-III (DBAC, Scheme 3) - a natural substance isolated from the leaves of the Taxus baccata species and phenylisoserine side chain [5, 6]. It has been shown that the side chain at C-13 and the diterpene moiety of taxol (DBAC) are both essential for the anticancer activity of the taxoid family. The baccatin III tetracyclic ring system has been shown to be a rigid segment. Moreover, the structure-activity studies demonstrated that the presence of the phenylisoserine side chain (as N-benzoyl-(2R,3S)-3-phenylisoserine, Scheme 2) with correct absolute stereochemistry is significant for strong cytotoxicity. As all amino acids, the side chain of paclitaxel, is flexible and can adopt different conformations. The total synthesis of paclitaxel was done in the early 70s [7], however, because of complicated multistep synthesis and small yield it is not applied in the pharmaceutical industry. A semisynthetic way of producing paclitaxel seems to be the most promising: it is based on semiproduct (10-deacetylbaccatin-III) extracted from leaves of Taxus baccata to condensate it with N-benzoyl-(2R,3S)-3-phenylisoserine [8–10]. Therefore, study on conformational landscape of phenylisoserine may provide insights and may support practical efforts to obtain the very valuable medicine.

Scheme 2
scheme 2

Taxol®

Full size image
Scheme 3
scheme 3

DBAC

Full size image

To study the conformational landscape of phenylisoserine as one of the β-amino acids is important not only because of paclitaxel and related compounds. It was found that β-amino acids can form stable α-helix, β-sheet, and hairpin motifs, similar to those in α-peptides. Indeed, while ca.15-20 residues are required to obtain a stable α-peptide α-helix, only six β-amino acids are necessary to form a stable β-peptide-based helix [11–14]. Because of its special importance in so many aspects, it is highly desirable to explore conformations of the β-amino acids thoroughly.

Likewise other amino acids, the phenylisoserine molecule has no symmetry and exhibits a significant internal rotational degree of freedom due to rotations around the three single bonds in the central part of the carbon skeleton. Therefore, it is likely to occur as numerous low energy conformers. Amino acids are known to exist as zwitterions in the solid state and in solution and in the nonionized, neutral form in the gas phase. Study of the conformational behavior of the neutral amino acids is important for understanding the dependence of the peptides and protein backbone.

The gas-phase data are, in general qualitatively consistent with the matrix-isolation results. Several amino acids have been studied experimentally by using the IR inert-matrix-isolation method [15–24]. In the low-temperature matrices three conformers were found for glycine [15–18], three for valine [19], two for proline [20], four for serine [21], six for phenylalanine [24], and more than three for tryptophane [25]. Tyrosine has been studied only by ab initio method [26]. Recently, for cysteine and β-alanine we identified in the low-temperature matrices, 3 and 6 conformers, respectively [22, 23]. Also, in few laboratories, investigations on gaseous amino acids have been carried out by microwave spectroscopy during the past two decades [26–41]. To the best of our knowledge, so far, phenylisoserine has not been studied either by matrix-isolation IR spectroscopy nor by microwave spectroscopy.

The conformational characteristic of phenylisoserine has been the subject of only a few theoretical studies. In the most detailed theoretical study, the conformation of the side chain (with the glycolate fragment) have been investigated at the HF/6-31G(d,p) and PM3 levels [42]. Nine conformers have been found and it has been suggested that only in one of them, the most stable, the C3 phenyl is accommodated in a favorable position with its π orbital far away from the carbonyl oxygen lone pairs. Onsager continuum approach (SCFR) and supermolecular models with two water molecules have been used to study the conformation in polar solvent. The conformation similar to the molecule of the crystal structure of paclitaxel is now the most favorable [42] (and refs. therein). The second part of paclitaxol – DBAC have also been studied by the calculations. The benzoate and acetate are necessary to allow binding to tubulin in the humane body. The importance of the oxetane ring was found by ZINDO method, which is consistent with experimental data [43].

The structure-activity relationship in the phenylisoserine moiety of the paclitaxel are dominated by the discussion of its different conformations named as polar and non-polar [42, 44, 45]. If the solvent is non-polar (CDCl3), then one assumes that paclitaxel binding site on microtubules is hydrophobic and the molecule adopts a staggered conformation around the C2-C3 bond. On the other hand, if the solvent is polar, the C3 phenyl group is positioned close to C2 benzoyl and C4 acetyl groups [42]. To summarize, the conformation of the phenylisoserine moiety in different solvents depends on several factors.

Continuing our previous study on β-amino acids [22, 23] we undertook a detailed investigation on the conformation behavior phenylisoserine isolated in the gas-phase and in the water solvent. Since there has been no experimental attempt to determine either microwave and matrix isolation IR spectra of gaseous phenylisoserine nor x-ray structure, the calculated data may be useful in searching of both gas-phase and condensed phases conformers with the spectroscopic methods. Therefore, the goal of this study is twofold: (i) to estimate the conformer population of phenylisoserine in the gas-phase; (ii) to predict the conformer population of phenylisoserine in the water solution considering zwitterionic structures. In this way, a platform for a better distinction of various conformers in future liquid-phase experiments is settled. To the best of our knowledge, no systematic and accurate study on the conformational behavior of phenylisoserine in gas-phase and in water solution has been reported.

Computational details

Modeling the gas phase

All the calculations were performed by using the hybrid Becke three-parameter Lee–Yang–Parr DFT B3LYP functional [46, 47] which reliability in calculations of the ground state geometries has been widely assessed [48]. The aug-cc-pVDZ Dunning’s [49, 50] (and for the most stable gas-phase conformers the aug-cc-pVTZ) basis sets were employed. These basis sets are known to be adequate to describe both organic molecules and their hydrogen-bonded systems. The conformational space of phenylisoserine was explored through a systematic variation of dihedral angles: d(O4C1O5H), d(O5C1C2C3), d(C1C2O6H), d(C1C2C3N7), d(C2C3N7H) and d(C2’C1’C3C2) (for atom labeling see Scheme 1). The stationary structures are found by ascertaining that all the harmonic frequencies are real. The relative abundances of the most stable conformations were then estimated by using the Gibbs free energy at 298.15 K, ΔG, relative to the most stable conformer. At the B3LYP/aug-cc-pVDZ level, 54 conformers were found to be stable in the gas phase.

All the calculations were performed by using the Gaussian 03 [51] and Gaussian 09 [52] programs.

Modeling the aqueous phase

A systematic search for the stable zwitterions within the supermolecule approach goes far beyond this study; therefore, we decided to study the solvent effect by the IEF-PCM model [53, 54], which includes long-range interactions only. Continuum models are widely used nowadays to simulate solvent effects on conformation [55]. Among many methods used, the IEF-PCM method introduced in the late 1990s, was shown to simulate the aqueous phase fairly well [55]. Combined supermolecular-continuum approaches would probably be more successful, if they could be applied for phenylisoserine surrounded by hundreds of water molecules. To this aim, 54 gas-phase nonionized structures were reoptimized in water solvent at the IEFPCM/B3LYP/aug-cc-pVDZ level to yield 51 nonionized and six stable zwitterion structures.

Results and discussion

Geometry and energy of the gas-phase phenylisoserine monomers

The phenylisoserine molecule has six single bonds around which free rotations may lead to as much as ca. 40 000 local minima. To generate the structures of it the relevant torsion angles were varied. Based on semiempirical AM1 search, 158 stable phenylisoserine conformers were appointed and reoptimized at the B3LYP/aug-cc-pVDZ level yielding 54 stable structures. Each of the phenylisoserine functional groups, carboxyl and alcoholic OH, C=O, and NH2, may operate as both a hydrogen bond donor or acceptor. Thus, because of internal hydrogen bonds some conformers are stabilized more than the other. Six low-energy structures of phenylisoserine (Fig. 1, Table 1) were further studied by using the B3LYP/aug-cc-pVTZ and MP2/aug-cc-pVDZ methods. They were found to be in the 3.5 kcal mol−1 energy range relative to the most stable form. Thermochemical properties at 298.15 K were computed for all the conformers and the relative abundances were estimated for six of them. Figures, energetic and selected dihedral angles of all 54 phenylisoserine conformers optimized at the B3LYP/aug-cc-pVDZ level are available in the electronic supplementary materials (Tables 1SI and 2SI, Fig. 1SI). In Table 3SI Cartesian coordinates are gathered for the six most stable structures calculated at the B3LYP/aug-cc-pVDZ level.

Fig. 1
figure 1

The B3LYP/aug-cc-pVDZ optimized the most stable six structures of phenylisoserine in the gas phase. The relative total energies (ΔE, kcal mol−1) and free Gibbs energies (ΔG, kcal mol−1) referred to the most stable conformer are given in parentheses and square brackets, respectively

Full size image
Table 1 The B3LYP/aug-cc-pVDZ, B3LYP/aug-cc-pVTZ and MP2/aug-cc-pVDZ optimized for the six most stable structures of phenylisoserine (PhIS) in gas phase, in water as neutral and zwitterion forms. The relative total energies (ΔE, kcal mol−1) and free Gibbs energies (ΔG, kcal mol−1) referred to the most stable conformer. The populations X(G) were calculated based on ΔG values (DFT) and ΔE values (MP2) separately for the 6 structures of each type
Full size table

Now let us compare the energetic results generated by using different methods and basis sets. The conformers are ordered according to the increasing total energy obtained at the B3LYP/aug-cc-pVDZ level (Table 1, Table 1SI). First and foremost, at the B3LYP/aug-cc-pVDZ level the most stable conformer 1 is more stable than the next one by as much as 2.5 kcal mol−1. Upon the thermochemical corrections, the energetic order changes but only slightly: conformer 3 is located at the 2nd position and conformer 5 at the 4th position (Table 1). However, the order of conformers 4, 5, and 6 is slightly perturbed (Table 1). Conformer 1 is found to be still the global minimum when the basis set is increased to aug-cc-pVTZ and is still separated from the next one by more than 2.3 kcal mol−1 (Table 1). The energetical picture undergoes significant change at the MP2/aug-cc-pVDZ level. First, conformer 2 becomes global minimum, separated from the next one, 1, by only 0.13 kcal mol−1. Moreover, the third stable conformer 3 is only 0.86 kcal mol−1 less stable than 2, and should be seriously taken into account if equilibrium conformer mixture in the gas phase is considered (Table 1).

The conformers may be characterized by the intramolecular hydrogen bonds (IHB). The presence of three hydrogen bond donors (NH2, OHc, OHa) and four hydrogen bond acceptors (O=C, OHa, NH2, OHc) allows for a wide range of hydrogen bond combinations and large number of stable forms. Let us now discuss the six lowest energy conformers. All the most stable six structures, exhibit the presence of the intramolecular hydrogen bonds (Fig. 1, Table 2). Two of them, 1 and 4, exhibit the OHc...NH2 hydrogen bond, in which the OHc carboxylic group, as a proton donor, is in trans position to the C=O group. This IHB forms a six-membered ring (Fig. 1). In conformers 1 and 4 this intramolecular H-bond is quite short, i.e., ca. 1.74 Å (Table 2). This type of hydrogen bond OHc...NH2 is the same as that of phenylalanine [24], tyrosine [26] and tryptophan [56] global minimum. In structures 2 and 3, the OHa alcoholic hydroxyl group is the proton donor and proton acceptor at the same time. In conformer 2 it is the proton donor to the NH2 group and proton acceptor from the OHc group forming the OHa...NH2 and OHc...OHa IHB, respectively. In conformer 3 the situation is different: the OHa group is the proton acceptor from the NH2 group and the proton donor to the C=O group, forming NH2...OHa and OHa...O=C IHB, respectively. The other very important IHB is the OHa...O=C bond (ca. 2.0 Å) stabilizing the 1, 3, and 6 conformers by forming a five-membered ring (Fig. 1). The next IHB, OHc...OHa is quite short (ca. 1.9 Å) and it is present in two conformers 2 and 5. In the case of conformers 3, 4, 5, and 6, there is also an IHB in which NH2 group is a proton donor to either OHa or O=C groups (Table 2).

Table 2 The intramolecular hydrogen bonds present in the most stable six conformers of phenylisoserine calculated at B3LYP/aug-cc-pVDZ level in the gas phase and at IEF-PCM/B3LYP/ aug-cc-pVDZ level in “water”
Full size table

The population of conformers at room temperature are shown in Table 1. In the gas phase the distribution has a peak: the most stable form 1 is at the DFT level populated in at least 93% (X(G) in terms of ΔG/B3LYP) (Table 1). On the other hand, the populations based on total energies calculated at the MP2/aug-cc-pVDZ level indicate three conformers: 1, 2, and 3 to be stable in significant amounts of ca. 37%, 46%, and 11%, respectively. Thus the population analysis is definitely dependent on energy correlation factor, however, to go far beyond the level applied here seems now to be quite a difficult task for so many conformers.

Energy of the aqueous-phase phenylisoserine conformers

Neutral forms

To estimate the influence of water on phenylisoserine molecule conformations, all 54 conformers stable in the gas phase were reoptimized by using the IEF-PCM/B3LYP/aug-cc-pVDZ method. The aqueous surrounding affects both the shape of the conformer distribution curve and the energetic differences between PhIS conformers. It is noticeable that in the first six stable conformations there are four conformers stable in the gas phase, yet, in a different order. Conformer 1w (1 stands for number of the conformer in the gas phase, w denotes water) remains the most stable one and is still separated from the next one by 2.4 kcal mol−1 (in terms of ΔG, Table 1). Furthermore, the conformers are ordered differently (Table 1); in particular, the energy differences between them are larger. Additionally, among the most stable conformers there are forms denoted as 29w and 10w, that have been much less populated in the gas phase. This confirms a significant effect of solvation on geometrical parameters and solvatation energies of solutes.

All the most stable six neutral structures exhibit the presence of the intramolecular hydrogen bonds (IHB, Fig. 2, Table 2). Out of these six neutral conformers, three exhibit the OHc...NH2 and two OHa...NH2 internal hydrogen bonds (Fig. 2, Table 2). The shortest OHc...NH2 hydrogen bond is found for 1w, which is the most stable neutral form in water. It is noticeable, that all structures with the OHc...NH2 IHB (1w, 29w, 4w) are the most stable in water. The longest OHc...NH2 hydrogen bond is detected for the 2w conformer, which is the fourth stable neutral form in water. Thus, the presence of the internal OHc...NH2 hydrogen bond is probably the main factor controlling conformer stability in water. Additionally, the distance of OHa...C1’(Phe) is shorter in 29w than in the other conformers. This suggests that the interaction with the ring π-electrons may also be an important factor. Figures, energies and selected dihedral angles of all 51 neutral PhIS conformers in water optimized at the IEF-PCM/B3LYP/aug-cc-pVDZ level are available in the electronic supplementary materials (Tables 4SI and 5SI, Fig. 2SI). Additionally, Table 3SI lists Cartesian coordinates of the six most stable neutral forms in water.

Fig. 2
figure 2

The B3LYP/aug-cc-pVDZ optimized the most stable six neutral structures of phenylisoserine in water modeled by IEF-PCM method. The relative total energies (ΔE, kcal mol−1) and free Gibbs energies (ΔG, kcal mol−1) referred to the most stable conformer are given in parentheses and square brackets, respectively

Full size image

Zwitterionic forms

The zwitterionic phenylisoserine conformers are unstable in the gas phase and thus, there are no such minima on the potential energy surface of an isolated monomer. Yet, in the water medium they become quite populated. Therefore, we carried out calculations for the zwitterion form of PhIS in the water phase. Out of 51 neutral structures stable in water only six appeared to exist as zwitterions (z) (Fig. 3, Table 1). Moreover, the zwitterionic 1z conformer came out to be the most stable and definitely dominating in water. Energies and selected dihedral angles of the six zwitterion PhIS conformers in water, optimized at the IEF-PCM/B3LYP/aug-cc-pVDZ level, are available in the electronic supplementary materials (Tables 6SI and 7SI). Additionally, in Table 3SI Cartesian coordinates of these zwitterion structures are listed.

Fig. 3
figure 3

The B3LYP/aug-cc-pVDZ optimized the most stable six structures of phenylisoserine zwitterions in water modeled by IEF-PCM method. The relative total energies (ΔE, kcal mol−1) and free Gibbs energies (ΔG, kcal mol−1) referred to the most stable conformer are given in parentheses and square brackets, respectively

Full size image

According to energetic ΔG results (Table 1), the most stable five zwitterion forms are those where the internal hydrogen bond is formed between the protonated amino group, NH +3 , and the carboxylic COO− moiety. For example, in 1z this distance is equal to 1.565 Å and is shorter by 0.056 Å than OHc...NH2 in neutral 1w form and shorter by 0.178 Å than OHc...NH2 in 1 in the gas phase. Additionally, the interaction of the NH +3 group with the OHa group stabilizes conformer 6z. Moreover, the interaction of OHa with the ring, OHa...C1’(Phe), seems to stabilize all zwitterionic conformers.

Let us now discuss conformer population in water solution. For water phase it is necessary to consider both the neutral and zwitterionic forms because for all aminoacids there are equilibria of these forms in solution. However, first let us focus on each type separately. The population of neutral conformers in water (Table 1) shows the maximum peak for conformer 1w, which is populated in at least 98% (X(G) in terms of ΔG/B3LYP). In contrast to the gas phase population, the MP2 level does not change the quantitative population picture. Indeed, according to MP2/aug-cc-pVDZ total energies conformer 1w is populated in ca. 87% (Table 1). Similarly, the population of zwitterionic conformers in water (Table 1) shows the maximum peak for conformer 1z, which is populated in ca. 99% (X(G) in terms of ΔG/B3LYP and 98% at the MP2 level (Table 1). Now, when the most stable six neutral and six zwitterionic conformers are considered all together (Table 8SI) it appears that, at the DFT level, only the most stable neutral conformer 1w and the most stable zwitterionic conformer 1z are present in amounts of ca. 86% and 12%, respectively. Qualitatively, the MP2 calculations yield a similar result: in practice only the most stable neutral conformer 1w and the most stable zwitterionic conformer 1z are present. However, the MP2 level indicates that the zwitterions can be populated much more (ca. 36%) than predicted by the DFT calculations. Thus we expect both forms to be detected in equilibrium in water.

At the end let us comment on the question of how much stabilization is provided by each of the hydrogen bonds identified in each of the important conformers, both in vacuum and in water. Notice, that the studied conformers are stabilized by a variety of different hydrogen bonds and close contacts (see Table 2). Moreover, the repulsive interactions are also important for conformer diversification.

A more detailed inspection into the influence of intramolecular hydrogen bonds can be obtained by AIM analysis [57]. It was shown that Laplacian value of the electron density as well as electron density in bond and ring critical points of the hydrogen bonds correlates with hydrogen bond strength [58–61]. Also, the hydrogen bond length correlates with its strength. Therefore, an interdependence between intramolecular hydrogen bond length and the Laplacian in its critical point is presented for all the most stable conformers in Fig. 4 (the values are listed in Table 9SI in Supplementary Information file). It is clear that the stronger the hydrogen bond the shorter it is and the more negative the Laplacian value. We found that for the neutral conformers in vacuum and in water, d(H...B) is linearly correlated with ∇2ρ (R = 0.930, Fig. 4), whereas the points corresponding to the zwitterions are spread out. Thus, the weak CH...OHa bond is placed in the top right of Fig. 4, the medium strong OHa...NH2 and OHa...O=C bonds are located in the center, whereas the strongest OHc...NH2 bonds are concentrated in the bottom left of Fig. 4. We found that for zwitterions the hydrogen bond length is not correlated with ∇2ρ value (Fig. 4), which is not surprising because of strong localization of the charges at the ends of the H-bonds.

Fig. 4
figure 4

The hydrogen bond lengths d(H...B) vs Laplacians ∇2ρ in the hydrogen bond BCPs calculated for the most stable neutral conformers in vacuum (v) and in water (w) and for zwitterionic forms in water (z) based on B3LYP/aug-cc-pVDZ optimized structures

Full size image

As we stress above, the hydrogen bond strength is not a unique factor ordering the conformers stability. Thus, despite the fact that one is able to scale the intramolecular H-bond strength in the studied conformers, the other factors are hidden yet not negligible.

Summary

Continuing our earlier studies on β-amino-acid conformations we found 54 conformers of penylisoserine to be stable in the gas phase, and 51 neutral and six zwitterionic conformers to be stable in water. The calculations were performed at the B3LYP/aug-cc-pVDZ, B3LYP/aug-cc-pVTZ and MP2/aug-cc-pVDZ levels. In the gas phase, in the most stable six conformers the following intramolecular hydrogen bonds are present: OHa...O=C, OHc...NH2, OHa...NH2, OHc...OHa, HNH...O=C, and HNH...OHa. The most stable conformer in the gas phase and in water phase is stabilized due to strong OHc...NH2 and OHa...O=C intramolecular hydrogen bonds. On the other hand, the zwitterionic forms in water are stabilized not only by interaction of NH +3 and COO− groups but also by interaction between H2NH and the phenyl ring. In the gas phase, at the DFT level, the most stable conformer is populated in at least 93%, whereas at the MP2/aug-cc-pVDZ level three conformers seems to be present in significant amounts of ca. 37%, 46%, and 11%. This shows that the population analysis is strongly dependent on electron correlation. Population analysis of the conformers in water (six neutral and six zwitterionic conformers) performed at the DFT and MP2 levels show concordantly that only the most stable neutral conformer and the most stable zwitterionic conformer are present. However, conformer ratio is again strongly dependent on electron correlation. Thus we expect both forms to be detected in equilibrium in water.