Quantitative In Silico Analysis of Retention of Phenylthiohydantoin-Amino Acids in Reversed-Phase Ion-Pair Liquid Chromatography

Abstract

The retention mechanisms of phenylthiohydantoin (PTH)-amino acids in reversed-phase ion-pair liquid chromatography were quantitatively analyzed in silico. The most significant interaction for the retention was the Lewis acid–base interaction between an aromatic ring of a PTH-amino acid and a hydroxyl-group hydrogen of tetra-alkyl ammonium hydroxide. Solvent effects, addition of molecular interaction (MI) energy values between an analyte and solvent molecules, significantly improved the relationship between the MI energy values, calculated using a molecular mechanics program, and logk values, measured via chromatography. The correlation coefficient between the calculated MI energy values and the logk values was 0.98 (n = 19).

Introduction

A clear, quantitative explanation of the concept of chromatographic retention is fundamental requirement in separation technology. First, the retention in reversed-phase liquid chromatography was quantitatively analyzed using the octanol–water partition coefficient (logP) and the dissociation constant (pKa) (1). However, the combination of logP and pKa values can be used in only very limited conditions in reversed-phase liquid chromatography. Therefore, a new approach has been studied.

A molecular dynamics program has been used for a simulation analysis of proteins and designing artificial polymers; however, it did not quantitatively provide the molecular interaction (MI) energy values. The reproducibility of complex structures with solvent molecules optimized via a molecular dynamics program was not satisfactory, and the selective MI of a hydrophobic analyte with an alkyl-bonded model phase in solvents (water and acetonitrile) was not observed by the molecular dynamics simulation (data not provided). A molecular mechanics (MM) program with a personal computer is a practical tool for quantitative analysis of MIs.

Chromatography is an excellent analytical method to study MIs. Chromatographic retention has been quantitatively analyzed via computational chemistry calculations, mainly using different versions of MM (MM2) programs with several model phases. However, it is difficult to build a real model of the stationary phase in silico; therefore, a simplification of this model phase is required to study MIs in chromatography (2). The retention times measured in a variety of chromatographic techniques, from gas chromatography to liquid chromatography, including affinity liquid chromatography, were quantitatively analyzed in silico; the calculated MI energy values (kcal/mol) between the analytes and a model phase were related to the measured logarithmic capacity ratio (logk) values. The MI energy value is calculated by subtracting the energy value of the complex from the sum of the individual energy values and the model phase energy value. The energy value change after complex formation is considered as the MI energy value and comprises the final (optimized) structure (MIfs), hydrogen bonding (MIhb), and electrostatic (MIes) and van der Waals (MIvw) forces. The MIhb and MIvw indicate the contribution of the hydrogen bonding and molecular size effects, respectively.

Furthermore, the retention times could be predicted if standard compounds for a column calibration, such as alkanes in gas chromatography and alkyl benzenes in reversed-phase liquid chromatography for phenolic compounds, were available. Various chromatographic data were quantitatively analyzed in silico, and the reports were summarized in a book (2). However, the quantitative in silico analysis of the retention mechanisms in reversed-phase ion-pair liquid chromatography has remained incomplete, because the structure of the stationary phase is complicated and the hydrophobic phase is coated with components of the mobile phase. On the other hand, fast chromatographic analysis of amino acids is also of fundamental importance. First, free amino acid separation was performed using ion-exchange liquid chromatography and then using reversed-phase ion-pair liquid chromatography (3). Furthermore, the analysis of phenylthiohydantoin (PTH)-amino acids is a very important process in protein chemistry. The aforementioned analysis was performed using reversed-phase ion-pair liquid chromatography. The chromatographic data were used to study the retention mechanisms in reversed-phase ion-pair liquid chromatography via MM calculations with model phases.

Experimental

Molecular modeling and MI energy values were calculated via MM of CAChe™ program from Fujitsu, Japan. Chromatographic data were measured using a 100 × 2.1 mm ID column packed with home-made phenylhexyl-bonded silica gel with eluent containing 10 mM sodium phosphate solution (pH 7.35), 0.001 M tetrabutyl-ammonium (TBA) hydroxide and 25% acetonitrile or with home-made octadecyl-bonded silica gel with eluent containing 10 mM sodium acetate solution (pH 7.3), 0.01 M tetramethyl-ammonium (TMA) hydroxide and 30% acetonitrile. Chromatograph used was a Shimadzu model LC9. The flow rate was 0.2 mL/min. The column temperature was 40°C.

Results

Generally, the retention of hydrophobic compounds, including acidic and basic drugs, is quantitatively related to the MI energy values, which are calculated using an MM2 program. Particularly, van der Waals energy values have indicated hydrophobic interactions. First, a model pentyl-bonded silica gel phase was used to study the retention mechanisms of PTH-amino acids in reversed-phase ion-pair liquid chromatography; however, the model pentyl groups were deformed after optimization using the MM2 calculations. Therefore, a model octyl-bonded silica gel phase (2) was used to obtain the MI energy values (data not provided here).

The hydrophobic MI energy values obtained using the model octyl-bonded silica gel phase were related to the logk values of 19 PTH-amino acids and the following equation was obtained:

$$\text{MIvw}=3.805\log k+24.826,r=0.495,n=19,$$ (1)

where MIvw denotes the MI values of van der Waals energy calculated between the model octyl-bonded silica gel phase and the PTH-amino acids. The unit of MI energy values is kcal/mol.

The correlation coefficient was poor when compared with the results previously obtained for aromatic acids, phenolic compounds, acidic drugs and basic drugs (2). Further studies were performed to understand the solvent effects for these phenolic compounds. The model acetonitrile phase was the same as that used for hydrophobic compounds (4). The solvent effects were obtained as MI energy values between the model acetonitrile phase and the PTH-amino acids. The calculated MI energy values are summarized in Table I. The MI of hydrogen bonding and electrostatic energy values were used as the solvent effect, because acetonitrile is polar solvent. The calculated MI(ahb + aes) values were combined with the above results.

$$y_2=-3.651\log k+10.305,r=0.786,n=\hspace{0.17em}19,$$ (2)

where y₂ is a combination of MIahb and MIaes as calculated as the solvent effect using the acetonitrile phase.

$$\text{MIvw}+2.8\times \text{MI}(\text{ahb}+\text{aes})=14.035\log k-4.075,r=0.915,n=\hspace{0.17em}19,$$ (3)

where MI(ahb + aes) represented the combined MI values of the hydrogen-bonding and electrostatic energies calculated between the model acetonitrile phase and the PTH-amino acids. An example of the complex conformation of PTH-alanine in the acetonitrile phase is shown in Figure 1, where PTH-alanine is sandwiched by acetonitrile molecules.

Table I.

Molecular Property of PTH-Amino Acids

No.	Chemicals	logk1	fs	hb	es	vw	TBAfs1	TBAhb1	TBAes1	TBAvw1
1	Alanine	0.671	−3.169	−0.282	−17.459	10.507	3.256	−8.028	−19.358	11.947
2	Arginine	0.851	−20.085	−0.293	−39.357	13.099	−16.567	−8.109	−40.458	10.589
3	Asparagine	0.053	−23.194	−3.589	−38.574	11.002	−16.805	−11.398	−40.557	12.422
4	Aspartic acid	−0.412	−5.849	−0.282	−21.665	10.837	0.579	−8.010	−23.966	12.660
5	Glutamine	0.091	−20.640	−3.289	−35.529	11.158	−16.848	−11.045	−37.833	10.493
6	Glutamic acid	−0.079	−2.284	−0.281	−17.325	11.089	1.571	−8.038	−19.854	10.706
7	Glycine	0.612	−4.639	−0.283	−17.459	10.441	2.766	−8.062	−19.811	13.383
8	Histidine	0.544	7.152	−1.061	−19.803	9.364	11.936	−9.066	−21.341	8.551
9	Isoleucine	1.359	0.994	−0.279	−17.519	12.259	4.887	−8.019	−19.391	11.191
10	Leucine	1.359	−0.500	−0.281	−17.476	11.469	3.672	−8.017	−19.343	10.602
11	Lysine	1.339	−0.668	−0.284	−17.420	11.354	3.064	−8.093	−19.423	10.029
12	Methionine	0.996	−2.286	−0.292	−17.754	10.876	1.534	−8.101	−18.748	9.694
13	Phenylalanine	1.281	−11.725	−0.994	−18.011	12.807	−6.591	−8.779	−19.843	13.051
14	Proline	1.077	15.035	0.000	−17.229	14.448	20.727	−7.767	−19.129	15.341
15	Serine	0.145	−2.361	−0.500	−17.620	10.553	3.972	−8.216	−19.497	12.638
16	Threonine	0.300	−0.035	−0.562	−17.422	11.434	6.592	−8.296	−19.451	13.266
17	Tryptophan	1.094	−6.646	−1.442	−18.698	10.980	−3.207	−9.597	−21.098	10.486
18	Tyrosine	0.763	−13.894	−2.460	−17.998	12.742	−8.792	−10.343	−10.821	12.935
19	Valine	1.059	−1.327	−0.281	−17.476	10.771	6.736	−7.358	−19.214	12.012
TBA-OH		–	22.590	0.000	−2.881	11.247	–	–	–	–
No.	Chemicals	logk2	TMAfs2	TMAhb2	TMAes2	TMAvw2	ACNfs3	ACNhb3	ACNes3	ACNvw3
1	Alanine	0.558	−22.823	−13.100	−24.248	8.009	28.296	−0.282	24.091	−6.401
2	Arginine	0.811	−41.574	−12.949	−45.861	8.510	5.929	−0.597	−1.733	−5.454
3	Asparagine	0.074	−14.224	−16.175	−46.010	7.620	3.446	−3.627	−1.503	−7.104
4	Aspartic acid	−0.632	−27.035	−13.345	−29.363	7.642	17.839	−0.282	12.922	−5.651
5	Glutamine	0.162	−42.218	−16.530	−43.142	7.743	3.505	−3.034	−0.114	−6.792
6	Glutamic acid	−0.215	−24.299	−13.282	−25.390	7.420	22.698	−0.281	17.352	−6.377
7	Glycine	0.350	−24.651	−13.301	−24.531	7.975	27.818	−0.283	24.079	−5.427
8	Histidine	0.234	−14.426	−13.517	−27.620	5.615	33.302	−1.044	16.470	−7.709
9	Isoleucine	1.311	−21.521	−13.154	−26.010	8.582	30.512	−0.279	23.653	−6.084
10	Leucine	1.355	−21.354	−14.023	−23.175	7.418	30.031	−0.281	23.912	−6.269
11	Lysine	1.260	−21.577	−13.016	−24.392	7.721	30.010	−0.284	24.104	−6.139
12	Methionine	0.988	−24.067	−13.372	−25.045	6.848	25.445	−0.292	21.230	−7.810
13	Phenylalanine	1.206	−32.706	−13.741	−25.171	9.270	17.046	−0.896	23.417	−6.793
14	Proline	0.993	−3.307	−12.848	−24.351	11.494	46.097	0.000	24.374	−2.940
15	Serine	0.162	−23.119	−13.439	−24.418	7.774	28.410	−0.570	21.912	−6.183
16	Threonine	0.162	−19.996	−13.617	−34.192	8.856	27.582	−0.690	22.088	−6.487
17	Tryptophan	1.079	−29.307	−14.776	−26.452	6.942	20.110	−1.767	21.514	−9.261
18	Tyrosine	0.671	−34.294	−15.355	−24.929	9.949	13.685	−2.309	22.107	−6.858
19	Valine	0.993	−21.379	−13.527	−23.582	7.563	30.100	−0.281	24.252	−6.594
TMA-OH		–	−3.663	−2.943	−6.239	3.006	–	–	–	–
ACN phase		–	–	–	–	–	44.570	0.000	47.055	−9.100

Logk1 and logk2: measured in TBA-OH and TMA-OH systems; fs, hb, es and vw: final structure, hydrogen bonding, electrostatic and van der Waals energy values (kcal/mol); 1, 2, 3: TBA-OH, TMA-OH and acetonitrile phase with PTH-amino acid complexes (kcal/mol).

Figure 1.

PTH-alanine in acetonitrile phase.

The above results indicated that further improvement of the correlation coefficient was necessary. Therefore, a model TBA phase was constructed and the MI energy values of the PTH-amino acids were calculated. TBA is an ion-pair reagent of an eluent and plays an important role in the reversed-phase ion-pair liquid chromatography of PTH-amino acids. PTH-amino acids are generally not ionized; however, TBA may contribute to charge-transfer-like interactions. The hydrophobic alkyl group of TBA coats the surfaces of the hydrophobic-bonded phase. Therefore, a flat sheet of TBA was constructed using six units of TBA ions, and the MI energy values between the TBA phase and the PTH-amino acids were calculated (data not provided). The calculated MI energy values were correlated with the logk values and the following correlation was obtained:

$$y=1.828\log k+15.621,r=\hspace{0.17em}0.451,n=\hspace{0.17em}19,$$ (4)

where y is the MI van der Waals value calculated between the model TBA phase and the PTH-amino acids. Further study was carried out by incorporating the solvent effects used in previous equation.

The addition of the above MI(ahb + aes) value to the MI van der Waals values of above equation improved the correlation coefficient to 0.924 (y = 6.927 logk + 1.198, r = 0.924, n = 19), but further study was required to obtain an even better correlation coefficient. The MI energy values calculated using a single TBA molecule and PTH-amino acids were insignificant, and the correlation coefficient was 0.118 (n = 19). Further study was carried out to elucidate the charge-transfer-type interactions between TBA hydroxide (OH) and PTH-amino acids. Initially, the best complex structure of a TBA-OH and PTH-alanine complex was studied. The initial position of PTH-alanine and TBA-OH is shown in Figure 2A. After several trial optimizations, the most stable, least energy-valued complex structure was obtained, as shown in Figure 2B. Then, other PTH-amino acids were located at similar positions about 8 Å from the OH group, and their complex structures were optimized. The calculated energy values of the complexes are summarized in Table I with the molecular properties of the PTH-amino acids. These values indicated that the interactions between the OH groups and the aromatic rings of PTH-amino acids are the basic MIs in this chromatographic technique. Such interactions include charge-transfer or Lewis acid–base interactions. MI energy values between a model TBA-OH and the PTH-amino acids were correlated. However, the correlation coefficient was still poor:

$$y_1=1.203\log k+10.164,r=0.450,n=\hspace{0.17em}19,$$ (5)

where y₁ is the MI of the van der Waals energy (MIvw).

Figure 2.

Complex conformation of PTH-alanine with TBA-OH.

Then, the above hydrophobic interaction energy values (MIvw) and solvent effect (MIahb + MIaes) were combined. The combined MI energy values improved the correlation coefficient:

$$Y=4.107\log k+1.930,r=0.975,n=\hspace{0.17em}19,$$ (6)

where Y = y₁ + 0.82 × y₂.

Generally, the hydrophobic interaction energy values obtained as MIvw using the hydrophobic model phases mainly contributed to the quantitative analysis of the logk values. However, the solvent effect contribution was significant in reversed-phase ion-pair liquid chromatography for the analysis of the PTH-amino acid retention, as shown in Figure 3.

Figure 3.

Relation between MI energy values calculated (kcal/mol) and logk values measured. Numbers beside symbols: see Table I. MIvw, van der Waals energy of MI energy; MI(ahb + aes), the combined MI values of the hydrogen-bonding and electrostatic energies calculated between the model acetonitrile phase and the PTH-amino acids.

Further study was carried out using TMA hydroxide as the ion-pair reagent. The total retention time was shorter due to the lower hydrophobicity of TMA when compared with TBA. However, the molecular size of TMA was too small to allow for complete interaction between the molecules of TMA and a PTH-amino acid. Therefore, two TMA-OH molecules were used for complete interaction with one PTH-amino acid. The optimized structure of two TMA-OHs, determined using the MM2 program, is shown in Figure 4A. Then, the pair of TMA-OH molecules formed an ion-pair with a PTH-alanine. The calculated values are summarized in Table I. The structure is shown in Figure 4B.

Figure 4.

Complex conformation of PTH-alanine with twin TMA-OHs.

The calculated MIvw values were related to the logk values of the PTH-amino acids. The correlation coefficient was 0.379 (n = 19). The correlation was improved by the addition of the previously obtained solvent effects MI(ahb + aes); the value was 0.979. A similar relationship was obtained as that shown in Figure 3.

Further study was carried out on the reversed-phase ion-pair liquid chromatography of free amino acids, in which the free amino acids formed ion-pairs with dodecyl sulfate ions. The chromatograph used was a home-made octadecyl-bonded silica gel column, with an eluent containing 50 mM sodium phosphate solution (pH 2.7), 25.6 mM sodium dodecyl sulfate and 16% methanol at 45°C. Free amino acids are polar compounds; therefore, they did not demonstrate a good correlation between the logk values and the MIvw calculated with a model hydrophobic phase (2). Even the paired ions of the free amino acids and the dodecyl sulfate ions did not demonstrate a meaningful correlation coefficient. The correlation coefficient was close to zero, and the slope was flat. The correlation was improved to 0.935 by the addition of solvent effects (MImhb + MImes), where mhb and mes indicate hydrogen-bonding and electrostatic energy values, calculated in a model methanol phase, similar to the model acetonitrile phase (data not provided).

Discussion

The hydrophobic interaction between the PTH-amino acids and a model octyl-bonded silica gel phase, TBA-OH or TMA-OH, did not significantly correlate with the retention of the PTH-amino acids in reversed-phase ion-pair liquid chromatography. However, the MIs between the acetonitrile phase and the PTH-amino acids played an important role in the quantitative explanation of the retention in the reversed-phase ion-pair liquid chromatography of the PTH-amino acids. The contribution of solvent effects was more significant than that in the reversed-phase liquid chromatography of phenolic compounds.

This chromatography was carried out using an eluent containing an ion-pair reagent. The PTH-amino acids were not ionized except for arginine, aspartic acid, glutamic acid and histidine depending on the eluent's pH. However, the contribution of ion-pair reagent was significant, resulting in fast and isocratic elution. The ion-pair reagent should be coated on the surface of the hydrophobic phase. Therefore, the Lewis acid–base interaction between an OH group of the ion-pair reagent and an aromatic ring of a PTH-amino acid was considered as the main interaction. The solvent effect was also important in reversed-phase ion-pair liquid chromatography of free amino acids using dodecyl sulfate ion and methanol.

Although a molecular dynamics program has been mainly used for a simulation analysis of proteins, it did not quantitatively provide the MI energy values, and reproducible complex structures with solvent molecules. A MM program with a personal computer is a practical tool for quantitative analysis of MIs.

Conclusion

The chromatographic retention behaviors of both free and derivatized amino acids in reversed-phase ion-pair liquid chromatography were the same. Amino acids formed paired ions with hydrophobic counter ions and retained the hydrophobic phase. The hydrophobic ions played a crucial rule in the retention of analytes, and the analytes were not ionized in the system. The contribution of the hydrophobic compounds demonstrated the contribution of MIvw, but the main contribution was that of the solvent effects, as calculated from both the model acetonitrile and methanol phases. A combination of the MIs between the hydrophobic ions and the PTH-amino acids, and the solvent effects, correlated well with the logk values, measured using the reversed-phase ion-pair liquid chromatography system. If we can predict solubility, we will be able to develop a method for predicting retention time.