Condensation dynamics of L-proline and L-hydroxyproline in solution

Abstract

We employ high-performance liquid chromatography with evaporative light scattering and mass spectrometric detection (HPLC/ELSD and LC/MS) to monitor the dynamic behavior of L-Pro, L-Hyp, and L-Pro–L-Hyp in 70% aqueous methanol. The individual amino acid solutions show evidence of oscillatory oligomerization. In the binary solution, the behavior is controlled by the dynamics of L-Pro oligomerization. A simple model involving oligomerization, formation of catalytic oligomer aggregates and cross-catalysis provides qualitative insight into the process.

---

Introduction

Spontaneous chiral inversion of nonsteroidal anti-inflammatory drugs (NSAIDs) in living organisms and in tissue culture has been studied for nearly half a century.¹In vitro evidence that this process can proceed in an oscillatory fashion in low molecular weight carboxylic acids dissolved in aqueous and non-aqueous abiotic media was first obtained in profen drugs,² a class of NSAIDs. Later, the phenomenon was shown to be quite general, occurring in chiral derivatives of acetic, propionic, and butyric acids with a chiral center located either on the α- or the β-carbon atom (e.g., ref. 3). Subsequent work found that oscillatory chiral conversion can be accompanied by oscillatory oligomerization.

The ability of amino acids to undergo spontaneous peptidization in aqueous solution was first demonstrated by means of the biuret test with L-phenylglycine.⁴ Several other amino acids were investigated with high performance liquid chromatography accompanied by diode array and mass spectrometric detection (e.g., ref. 5). A general scheme for chiral conversion of low molecular weight carboxylic acids in aqueous solution can be represented as:⁶

where X: –R (aliphatic) and Y: –NH₂, –OH, or –Ar (aromatic).

In anhydrous media and in the presence of trace amounts of water, the probable mechanism of chiral conversion is:⁷

In amino acids, the parallel processes of chiral conversion and peptidization can be illustrated by the following scheme:⁸

In this study, we present the results of experiments with L-proline (L-Pro) and L-hydroxyproline (L-Hyp), two important endogenous amino acids, both of which contain a five-membered pyrrolidine ring, as shown below.

We selected these two amino acids because of their important role as building blocks of collagen, which is omnipresent in the connective tissue of mammals and largely responsible for tissue architecture and strength. More specifically, our interest in structural transformations of these amino acids was attracted by the reluctance of L-Pro to participate in alpha helices^9,10 and to the frequent presence of L-Pro and L-Hyp in the collagen triple helix.^11,12

A detailed mechanism for the spontaneous oligomerization of Pro is depicted in the scheme below. Analogous mechanisms can be written for the homo-oligomerization of Hyp and the hetero-oligomerization of Pro–Hyp. Similar mechanisms of spontaneous amino acid oligomerization can be found in the organic chemistry literature, e.g., ref. 13 and 14.

The oligomerization process shown above is promoted by acidic media (which is the case with carboxylic acids dissolved in water and/or alcohol). The condensation can continue beyond the dipeptide stage shown to produce n-mers of various lengths. In mixtures of amino acids, a variety of heteropeptides may be generated. Structures of the simplest homo- and hetero-condensation products derived from Pro and Hyp are shown below:

We seek to trace the influence on the oscillatory condensation dynamics of the constraints produced by fixing the pyrrolidine or hydroxypyrrolidine ring to the amino acid asymmetry center (α–C*) with two (C–H and C–N) σ-bonds, and also of the steric shielding of the asymmetric center by the pyrrolidine or hydroxypyrrolidine ring. We study both the individual amino acid solutions and the binary mixture by achiral HPLC/ELSD and LC/MS. We also introduce a simple model that qualitatively accounts for the observed oscillatory behavior.

Experimental

Reagents

Analytical purity grade amino acids were purchased from Sigma-Aldrich (St Louis, MO, USA; L-Pro, cat. # P0380) and from Fluka (Buchs, Switzerland; L-Hyp, cat. # 56250). Methanol was of HPLC purity (Merck KGaA, Darmstadt, Germany), and water was de-ionized and double distilled with an Elix Advantage model Millipore system (Molsheim, France).

For the HPLC/ELSD and LC/MS experiments, we prepared 1.0 mg mL⁻¹ solutions of L-Pro, L-Hyp, and L-Pro–L-Hyp in 70% aqueous MeOH (i.e., 8.69 × 10⁻³ mol L⁻¹ for L-Pro and 7.63 × 10⁻³ mol L⁻¹ for L-Hyp). The 70% aqueous methanol, known for its strong antiseptic properties, was selected in order to protect the amino acids, which were stored for prolonged periods, from microbial action.⁵

High-performance liquid chromatography with evaporative light scattering detection (HPLC/ELSD)

The achiral HPLC mode was employed to separate the oligopeptides from the monomeric amino acids and also to separate L-Pro and L-Hyp in the binary amino acid mixture. The analyses were carried out using a Varian model 920 liquid chromatograph (Varian, Harbor City, CA, USA) equipped with a Varian 900-LC model autosampler, a gradient pump, a Varian 380-LC model ELSD detector, and Galaxie software for data acquisition and processing. The analyses were carried out on 20 μL aliquots in the isocratic mode, using a Pursuit 5C18 (5 μm particle size) column (250 mm × 4.6 mm i.d.; Varian; cat. no. A3000250C046). With single amino acids, we employed methanol–water (80 : 20, v/v) at a flow rate of 0.6 mL min⁻¹, while with the binary L-Pro–L-Hyp mixture, we used methanol–water (50 : 50, v/v) at the same flow rate. The chromatographic column was thermostatted at 30 °C in a Varian ProStar 510 model column oven. The analyses were carried out at 17 min intervals for 60 h.

High-performance liquid chromatography with mass spectrometric detection (LC/MS)

Liquid chromatographic analyses with mass spectrometric detection (LC/MS) were carried out to detect the presence of oligopeptides in our amino acid solutions. They were performed on freshly prepared samples and after 8 days (L-Pro) and 2 days (L-Hyp) of storage. An LC/MS System Varian (Varian, Palo Alto, CA, USA) was employed, equipped with a Varian ProStar pump, Varian 100-MS mass spectrometer, and Varian MS Workstation v. 6.9.1 software for data acquisition and processing.

LC analyses of the amino acid solutions were performed on 20 μL aliquots in the isocratic mode, using a Pursuit XRS3 C18 (5 μm particle size) column (50 mm × 2.0 mm i.d.; Varian, Harbor City, CA, USA; cat. no. A6001050R020) and methanol–water (90 : 10, v/v) mobile phase at a flow rate of 0.20 mL min⁻¹. Mass spectrometric detection was carried out in the ESI mode (full ESI-MS scan, positive ionization, spray chamber temperature 45 °C, drying gas temperature 250 °C, drying gas pressure 25 psi, capillary voltage 50 V, needle voltage 5 kV).

Results and discussion

HPLC/ELSD of L-proline, L-hydroxyproline, and L-proline–L-hydroxyproline

A chromatogram obtained with the ELSD detector for L-Pro dissolved in 70% aqueous MeOH remains qualitatively unchanged for 60 h of sample storage (retention time, t_R = 5.46 min). In order to visualize the time evolution of the solution, we plotted the changing peak heights against the sample storage time. The results are shown in Fig. 1. We see the non-linear signal intensity changes, which are equivalent to the respective concentration changes. From these results, we conclude that quantitative transformation of L-Pro never exceeds 10%, which is considerably less than with other amino acids investigated in a similar fashion.^4,15 Consequently, the oligomerization yields are relatively low, as no additional peak apart from that of L-Pro is distinguished by the ELSD detector.

Fig. 1 Time series of chromatographic peak heights at t_R = 5.46 min for L-Pro in 70% aqueous MeOH.

To assess whether the HPLC signal for L-Pro contains a significant periodic component, we Fourier-transformed the data shown in Fig. 1. The power spectrum calculated for the L-Pro peak is plotted in Fig. 2. It contains a large peak at zero frequency, which was removed from the plot. The predominant peak appears at 0.05 h⁻¹, implying a periodicity of ca. 20 h. However, the total length of the data is limited to 60 h by our experimental stability. A longer time series would be desirable to confirm this periodicity.

Fig. 2 Power spectrum calculated from time series of chromatographic peak heights shown in Fig. 1.

A chromatogram obtained with the ELSD detector for L-Hyp dissolved in 70% aqueous methanol remains qualitatively unchanged after 60 h of storage. The chromatographic peak height (equivalent to [L-Hyp]) is plotted as a function of time in Fig. 3. Visual inspection of this plot does not reveal an evident periodicity, although we note several pronounced minima, which suggest periods of enhanced condensation.

Fig. 3 Time series of chromatographic peak heights at t_R = 5.36 min for L-Hyp in 70% aqueous MeOH.

In Fig. 4(a), we present a typical chromatogram of L-Pro–L-Hyp dissolved in 70% aqueous methanol obtained with the ELSD detector. In monitoring the kinetic behavior of this mixed solution, we sought to balance the need for relatively short chromatographic runs in order to improve the time resolution against the desire to get a clean separation of L-Pro from L-Hyp. We compromised on a 17 min time interval, which gives a less than perfect (i.e., a non-baseline) separation, but does allow us to follow quantitative changes in the two amino acid concentrations from their changing peak heights.

Fig. 4 (a) Chromatogram of L-Pro–L-Hyp in 70% aqueous MeOH taken with the ELSD detector. Retention times (t_R) for L-Hyp and L-Pro are, respectively, 6.90 and 7.30 min; (b) time series of chromatographic peak heights for L-Hyp (t_R = 6.90 min) and L-Pro (t_R = 7.30 min) in a 70% aqueous MeOH mixture of L-Pro–L-Hyp.

The chromatographic peak heights for the peaks recorded at t_R = 6.90 min (L-Hyp) and 7.30 min (L-Pro) against the sample storage time are shown in Fig. 4(b). Visual inspection of these plots shows that the time changes of the two concentrations have a similar periodicity. Each of the traces resembles that of L-Pro alone (Fig. 1), rather than the trace of L-Hyp alone (Fig. 3). These observations suggest a process of mixed oligomerization governed by the dynamics of L-Pro.

These results suggest that neither the constraints produced by fixing the pyrrolidine or the hydroxypyrrolidine ring to the amino acid asymmetric center, nor the steric shielding of that center by the pyrrolidine or hydroxypyrrolidine ring, exerts a perceptible hampering effect on the dynamics of the nonlinear concentration changes with Pro and Hyp dissolved in 70% aqueous methanol, either individually or as a binary mixture.

LC/MS of L-proline, L-hydroxyproline, and L-proline–L-hydroxyproline

Liquid chromatography with mass spectrometric detection provides an independent approach to monitor the process of condensation. This technique is particularly useful with L-Pro, where the oligomerization yields are low and the ELSD detector fails to provide evidence of oligomer formation. We performed chromatographic separations on freshly prepared and aged (after 8 days of storage) L-Pro samples on an LC column, and recorded the mass spectra of the column effluents. The respective chromatograms are shown in Fig. 5(a and b). For each chromatogram, insets show the mass spectra corresponding to the key peaks.

Fig. 5 Chromatogram of L-Pro solution taken with the LC/MS system after (a) 0 days and (b) 8 days in 70% aqueous MeOH. Insets show mass spectra at the maxima of the respective chromatogram peaks.

It is evident that the intensity of the main chromatographic peak in the fresh sample is higher (maximum ca. 60 megacounts) than that of the main chromatographic peak in the aged sample (maximum ca. 25 megacounts). This result alone attests to the consumption of the starting material in the course of aging. With the fresh L-Pro sample, there is only one major peak in the chromatogram (Fig. 5(a)). The predominant signals in the corresponding mass spectra recorded from the peak maximum and from the two shoulders of the peak are at m/z 138, which is consistent with the structure [Pro + Na]⁺. The chromatogram of the aged L-Pro sample shows a more complex structure, with the main chromatographic peak at t_Rca. 2 min, and two additional peaks, one eluting faster than the main peak (t_R ≈ 1 min), and the other at t_R ≈ 3 min. The predominant signal in the main peak (mass spectra were collected from two points close to its maximum) is at m/z 138, which confirms the identity of Pro. However, in these two mass spectra for the main peak, many signals with m/z values higher than 138 can also be seen, pointing to the presence of oligomers. The predominant signal in this peak is at m/z 301, which can be considered as related to the Pro-derived trimer. The low-intensity diffuse peak at t_R ≈ 3 min shows an abundance of signals with m/z values much higher than that of Pro, which offers further evidence of oligopeptidization. There are three mass spectrometric signals with a distinctly higher abundance, at m/z 263, 383, and 743. The signal at m/z 263 may be attributed to [Pro₃–COOH–H]⁺, while those at m/z 383 and 743 may be fragments of a Pro tetrapeptide and octapeptide, respectively. We conclude that, although peptidization yields with L-Pro are rather low, oligomers with multiple condensed monomer units are indeed observed.

In Fig. 6(a and b), we show the progress of L-Hyp condensation with LC-MS. The intensity of the main peak in the fresh sample is higher (maximum ca. 125 megacounts) than that of the main peak in the aged sample after 2 days (maximum < 70 megacounts), which implies consumption of the starting material. Further evidence is provided by the appearance of several chromatographic peaks in the aged sample and a more complex structure of the chromatographic envelope than in the freshly prepared solution. Finally, mass spectra recorded from different parts of the chromatogram of the aged sample (Fig. 6(b)) show considerably more abundant condensation products with higher m/z values than in the mass spectrum of the fresh sample (Fig. 6(a)).

Fig. 6 Chromatogram of L-Hyp solution taken with the LC/MS system after (a) 0 days and (b) 2 days in 70% aqueous MeOH. Insets show mass spectra at the maxima of the respective chromatogram peaks.

Because the condensation rate is quite rapid, L-Hyp-derived condensation products are present both in the freshly prepared and the aged samples, although peaks attributable to the various oligomers are much higher in the aged sample than in the freshly prepared one. The m/z value suggests the presence of relatively large oligomers. For example, the signal recorded in the freshly prepared L-Hyp sample at m/z 1908 can be ascribed to [Hyp₁₇–OOH]⁺. The predominant signals recorded from the peaks in the aged L-Hyp sample (e.g., m/z 1850, 2637, and 2892) might correspond to oligomers composed of considerably more than a dozen monomer units, i.e., [Hyp₁₆ + Na]⁺, [Hyp₂₃ + OH]⁺, and [Hyp₂₅ + COOH]⁺, respectively.

Finally, we monitored the condensation of L-Pro–L-Hyp by LC-MS (Fig. 7(a and b)). As in the case of the individual amino acids, sample aging results in accumulation of condensation products. Similar to the HPLC/ELSD results, the pattern of condensation products implied by the mass spectrum also resembles that in L-Pro rather than the more complex spectrum of L-Hyp, again implying that the overall condensation dynamics in the mixed sample is influenced more strongly by L-Pro than by L-Hyp.

Fig. 7 Chromatogram of an L-Pro–L-Hyp solution taken with the LC/MS system after (a) 0 days and (b) 7 days in 70% aqueous MeOH. Insets show mass spectra registered at the maxima of the respective chromatogram peaks.

In the fresh sample (Fig. 7(a)), the two pronounced signals at m/z 139 and 154 can be attributed [Pro + Na]⁺ and [Hyp + Na]⁺, respectively. Moreover, in both the fresh and the aged samples, there is a distinct signal at m/z 647, and in the aged sample, an additional signal at m/z 953. The signal at m/z 647 can be attributed to [Pro₃–Hyp₃–H]⁺ and that at m/z 953 might correspond to [Pro₅–Hyp₄–2H]⁺. If the assignment of these latter two signals to the respective hetero-oligomer structures is correct, then the molecular proportions of Pro and Hyp in these species, 3 : 3 and 5 : 4, resemble the molar proportions of L-Pro and L-Hyp in the solution, which are almost 1 : 1.

Three important conclusions can be drawn based on the LC/MS results. First, the abundance of high m/z species in the aged samples excludes any major impact of amino acid esterification with methanol (a low molecular weight solvent) and supports the presence of homo- and hetero-amino acid condensation products in the aged samples. Second, the observed m/z values corresponding to the mixed L-Pro–L-Hyp entities suggest hetero-condensation taking place in the system. Finally, the overall yields of condensation products in the L-Pro–L-Hyp and L-Pro samples are similar to but lower than in the L-Hyp sample, which implies that the condensation dynamics in the L-Pro–L-Hyp mixture is governed by that of L-Pro.

Modeling

In an earlier study of oscillatory L-lactic acid oligomerization, we proposed a model¹⁶ that takes into account oligomerization of monomers, catalyzed and uncatalyzed aggregation of oligomers into aggregates that catalyze the formation of further aggregates (autocatalysis), and decomposition of aggregates. Here, we extend that model to allow for the presence of two types of monomers (corresponding, for example, to L-Pro and L-Hyp). We also allow for the possibility of cross-catalysis, i.e., that aggregate 1 can catalyze production of aggregate 2 and vice versa. By setting either or both of the cross-catalysis rates to zero, we obtain a family of four models: (a) no cross-catalysis – the two species behave independently; (b and c) control by one species – either 1 catalyzes 2 or 2 catalyzes 1; (d) mutual catalysis – each species catalyzes aggregation of the other.

The model equations are:

Oscillator 1:

n₁₁P₁ → E₁ rate = k₀₁P₁ (oligomerization)

n₂₁E₁ → M₁ rate = k_u1E₁ (uncatalyzed aggregation)

2M₁ + n₂₁E₁ → 3M₁ rate = k_a1M₁²E₁ (catalyzed aggregation)

M₁ → products rate = k_b1M₁ (decomposition)

Oscillator 2:

n₁₂P₂ → E₂ rate = k₀₂P₂ (oligomerization)

n₂₂E₂ → M₂ rate = k_u2E₂ (uncatalyzed aggregation)

2M₂ + n₂₂E₂ → 3M₂ rate = k_a2M₂²E₂ (catalyzed aggregation)

M₂ → products rate = k_b2M₂ (decomposition)

Cross catalysis:

2M₁ + n₂₂E₂ → 2M₁ + M₂ rate = k_c1M₁²E₂ (1 catalyzes 2)

2M₂ + n₂₁E₁ → 2M₂ + M₁ rate = k_c2M₂²E₁ (2 catalyzes 1)

here, P_i, i = 1, 2, is a monomer, e.g., L-Pro or L-Hyp, E_i is an oligomer composed of n_1i molecules of P_i, and M_i is an aggregate formed from n_2i molecules of E_i. Pairs of aggregates catalyze formation of a third aggregate.

Although in a real system, there is likely to be a distribution of oligomer lengths and aggregate sizes, in this simplified model we consider only a single representative value for each of these parameters and each species. As discussed in more detail elsewhere,¹⁶ the oligomerization and aggregation steps described in our model by single reactions actually consist of multiple steps involving different oligomer lengths and aggregate sizes. No data are available as to the distribution of these species nor the rate constants of the individual steps in which they participate. We have therefore adopted a coarse-graining approach similar to that of Coveney and Wattis,¹⁷ in which a family of steps involving species of different sizes is represented by a single step characterized by a rate constant k and a species length n (see Table 1 for values) chosen to give qualitative agreement with the observed behavior. We choose different parameter sets for the two species to give different periods of oscillation.

Table 1 Parameters used in model simulations

Parameter	Oscillator 1	Oscillator 2
P 0 (initial conc., M)	0.1	0.06
n 1	5	3
n 2	8	6
k 0 (s^−1)	1.5 × 10^−5	2.5 × 10^−5
k u (s^−1)	5.0 × 10^−5	8.0 × 10^−5
k a (M^−2 s^−1)	2.5 × 10^5	6.0 × 10^5
k b (s^−1)	5.0 × 10^−3	7.0 × 10^−3
k c (M^−2 s^−1)	2.5 × 10^4 or 0	6.0 × 10^4 or 0

---

In Fig. 8(a–d), we show four cases illustrating independent oscillation and cross-catalysis by either or both species. For the parameter sets investigated here, only mutual cross-catalysis gives synchronized behavior, with the faster oscillator 1 (L-Pro in our experiments) apparently controlling the behavior of the coupled system, since the joint period is nearly identical to that of free species 1, and the extrema of species 1 precede those of species 2. Note that in the cases in which only one species is capable of cross-catalysis, there is a significant perturbation of the behavior of the species whose aggregation is cross-catalyzed, but not to the extent of full synchronization.

Fig. 8 Simulations of oligomer concentrations in a solution initially containing two monomers that undergo oligomerization, uncatalyzed and catalyzed aggregation. Model is described in the text. Parameters are given in Table 1. (a) No cross catalysis (k_c1 = k_c2 = 0); the two species behave independently, equivalent to separate solutions. (b) 1 catalyzes 2 (k_c2 = 0); (c) 2 catalyzes 1 (k_c1 = 0); (d) mutual cross-catalysis. [E1] is upper curve, [E2] is lower curve.

Conclusions

Spontaneous peptidization of L-Pro and L-Hyp as individual amino acids in 70% methanol (monitored by means of HPLC/ELSD and LC/MS) shows that the peptidization dynamics and yields were considerably different for L-Pro and L-Hyp. In the binary L-Pro–L-Hyp solution, both amino acids also undergo non-monotonic concentration changes, which closely resemble the behavior of the L-Pro solution. These observations as well as the mass spectra of the individual and binary solutions suggest that the non-monotonic concentration changes of the two amino acids in a binary mixture are interconnected, probably through formation of hetero-oligopeptides. A simple model that allows for varying degrees of interaction between a pair of species undergoing oscillatory oligomerization gives results in qualitative agreement with a mechanism in which aggregates of L-Pro are able to catalyze the formation of catalytic aggregates of L-Hyp oligomers. The model, however, is an abstract one that involves a number of oversimplified assumptions, such as the existence of only a single oligomeric form and a single aggregate for each species. It should therefore not be construed as a mechanism for this particular system but rather as a “proof of principle” that such a mixed system with cross catalysis is capable of giving rise to behavior of the sort observed here.

The spontaneous peptidization of amino acids in purely abiotic media has not previously attracted a great deal of attention. To our best knowledge, this phenomenon has been systematically investigated only in ref. 4 and 5. The microwave-instigated peptidization of amino acids^13,14 is perhaps the closest process to the one studied here. Those results demonstrate that peptidization of amino acids can occur in the absence of a catalyst in simple aqueous and non-aqueous solvents. We note that the present investigation has been inspired by an interest in prebiotic chemistry, in which the spontaneous peptidization of amino acids may have played a significant role.

Acknowledgements

This work was supported by National Science Foundation grant CHE-1012428. We thank Prof. Bing Xu for enlightening discussions of the mechanism of amino acid condensation.

References

1. V. Wsol, L. Skalova and B. Szotakova, Curr. Drug Metab., 2004, 5, 517–533.
2. M. Sajewicz, R. Piętka, A. Pieniak and T. Kowalska, Acta Chromatogr., 2005, 15, 131–149.
3. M. Sajewicz, D. Kronenbach, M. Gontarska and T. Kowalska, J. Liq. Chromatogr. Relat. Technol., 2010, 3, 1047–1057.
4. M. Sajewicz, M. Gontarska, D. Kronenbach and T. Kowalska, Acta Chromatogr., 2009, 21, 151–160.
5. M. Sajewicz, M. Gontarska, D. Kronenbach, M. Leda, T. Kowalska and I. R. Epstein, J. Syst. Chem., 2010, 1(7), 1–16.
6. P. Belanger, J. G. Atkinson and R. S. Stuart, J. Chem. Soc. D, 1969, 1067–1068.
7. Y. Xie, H. Liu and J. Chen, Int. J. Pharm., 2000, 196, 21–26.
8. M. Sajewicz, M. Matlengiewicz, M. Leda, M. Gontarska, D. Kronenbach, T. Kowalska and I. R. Epstein, J. Phys. Org. Chem., 2010, 23, 1066–1073.
9. R. H. Yun, A. Anderson and J. Hermans, Proteins, 1991, 10, 219–228.
10. J. Y. Suh, Y. T. Lee, C. B. Park, K. H. Lee, S. C. Kim and B. S. Choi, Eur. J. Biochem., 1999, 266, 665–674.
11. B. Brodsky and A. V. Persikov, Adv. Protein Chem., 2005, 70, 301–339.
12. J. Engel and H. P. Bächinger, Top. Curr. Chem., 2005, 247, 7–33.
13. P. Lindström, J. Tierney, B. Wathey and J. Westman, Tetrahedron, 2001, 57, 9225–9283.
14. K. A. Nonappa, M. Lahtinen and E. Kolehmainen, Green Chem., 2011, 13, 1203–1209.
15. M. Sajewicz, M. Gontarska and T. Kowalska, J. Chromatogr. Sci. DOI:10.1093/chromsci/bmt033.
16. M. Sajewicz, M. Dolnik, D. Kronenbach, M. Gontarska, T. Kowalska and I. R. Epstein, J. Phys. Chem. A, 2011, 115, 14331–14339.
17. P. V. Coveney and J. A. D. Wattis, Mol. Phys., 2006, 104, 177–185.

---