Anal. Chem. 2005, 77, 3113-3122
Adsorption Equilibria of Benzodiazepines on a Hybrid Polymeric Chiral Stationary Phase Alberto Cavazzini,*,† Francesco Dondi,† Stefania Marmai,† Erik Minghini,† Alessandro Massi,† Claudio Villani,‡ Romina Rompietti,‡ and Francesco Gasparrini‡
Department of Chemistry, University of Ferrara, via L. Borsari 46, 44100 Ferrara, Italy, and Dipartimento di Studi di Chimica e Tecnologia delle Sostanze Biologicamente Attive, Universita` degli Studi di Roma, “La Sapienza”, P.le A. Moro 5, I-00185 Roma, Italy
The chromatographic behavior of a series of racemic benzodiazepines was evaluated under linear and nonlinear conditions on a new hybrid polymeric (DACH-ACR) chiral stationary phase (CSP). Differently substituted benzodiazepines were employed as probes to make hypotheses concerning possible molecular interaction mechanisms originating between target compounds and active sites on the CSP. Hydrogen bonds were found to be pivotal for chromatographic retention and chiral selectivity. The competitive effect from a mobile-phase (MP) modifier able to interact with the CSP through H-bonds was investigated. The performance of the polymeric DACH-ACR CSP for preparative purposes was also evaluated. The competitive adsorption isotherms of two benzodiazepines, lorazepam and temazepam, were measured at different MP compositions through the so-called inverse method. The adsorption data were fitted with a competitive biLangmuir adsorption isotherm. Enantiomeric separations under nonlinear conditions were modeled by using the equilibrium dispersive (ED) model of chromatography. Theoretical overloaded band profiles (obtained by solving the system of partial differential equations described by the ED model) matched, in a significantly accurate way, the profiles experimentally measured. Large-scale preparative chromatography represents one of the most attractive and useful means for purification of racemates.1-4 With respect to the other techniques in use for racemic resolution, chromatographic separations offer some advantages. Stereoselective syntheses, for instance, usually take longer to develop, involve numerous steps, and rarely give the required enantiomeric purity.5 Enzymatic reactions, on the other hand, lead to the destruction of one of the enantiomers. Finally, crystallization * To whom the correspondence should be addressed. E-mail: cvz@ unife.it. Fax: +39 0532 240709. † University of Ferrara. ‡ Universita ` degli Studi di Roma. (1) Fornstedt, T.; Guiochon, G. Anal. Chem. 2001, 73, 608A-617A. (2) Ahuja, S. Chiral separations by chromatography; Oxford University Press: Washington DC, 2000. (3) Cavazzini, A.; Kaczmarski, K.; Szabełski, P.; Zhou, D. M.; Liu, X. D.; Guiochon, G. Anal. Chem. 2001, 73, 5704-5715. (4) Ariens, E. J. Med. Res. Rev. 1986, 6, 451-466. (5) Beesley, T. E.; Scott, R. P. W. Chiral chromatography; John Wiley & Sons: Chichester, 1998. 10.1021/ac048101t CCC: $30.25 Published on Web 04/19/2005
© 2005 American Chemical Society
techniques are often used in addition to chromatographic separations.6 The extremely large number of chiral stationary phases (CSPs) available makes direct separations the most versatile choice when HPLC separation of enantiomers is planned. Enantiorecognition processes on CSPs originate due to a series of specific interactions of significantly different physicochemical nature: hydrogen bonding, van der Waals interactions, ionic forces, dipole stacking, π-π aromatic stacking, enthalpic-entropic compensation effect, chelate effect, and hydrophobic and steric effects. The presence of basketlike structures on the adsorptive material or of a rigid conformation on the enantiomers has been noted as one of the key factors in the chiral recognition process. Contrarily, when molecules without rigid structures interact with CSPs through chemical moieties positioned far from the stereogenic center, resolution can be difficult to attain.7-13 The progressive understanding of the molecular-level mechanisms, which lead to chiral recognition, has significantly contributed to the development of new adsorptive materials whose structures are designed based on this information. Modern techniques of silica gel derivatization make adsorptive materials able to support extremely different mobile-phase (MP) compositions, with beneficial effects for the separation. Different MP compositions can be used, for instance, to improve the selectivity of chiral separations; moreover, by varying the MP polarity, target compound solubility can be increased (an extremely important issue in preparative chromatography).14 When employed as chromatographic supports, new adsorptive materials must also possess enhanced mass-transfer properties to ensure high efficiency. CSPs based on chiral polymers covalently linked to a silica matrix have been observed to exhibit (6) Lorenz, H.; Sheehan, P.; Seidel-Morgenstern, A. J. Chromatogr., A 2001, 908, 201-214. (7) Armstrong, D. W.; Tang, Y. B.; Chen, S. S.; Zhou, Y. W.; Bagwill, C.; Chen, J. R. Anal. Chem. 1994, 66, 1473-1484. (8) Berthod, A.; Liu, Y. B.; Bagwill, C.; Armstrong, D. W. J. Chromatogr., A 1996, 731, 123-137. (9) Berthod, A.; Chen, X. H.; Kullman, J. P.; Armstrong, D. W.; Gasparrini, F.; D’Acquarica, I.; Villani, C.; Carotti, A. Anal. Chem. 2000, 72, 1767-1780. (10) Popieniek, P. H.; Pratt, R. F. J. Am. Chem. Soc. 1991, 113, 2264-2270. (11) Cavazzini, A.; Nadalini, G.; Dondi, F.; Gasparrini, F.; Ciogli, A.; Villani, C. J. Chromatogr., A 2004, 1031, 143-158. (12) Armstrong, D. W.; Rundlett, K. L.; Chen, J. R. Chirality 1994, 6, 496-509. (13) Jandera, P.; Buncˇekova´, S.; Mihlbachler, K.; Guiochon, G.; Bacˇkovska´, V.; Planeta, J. J. Chromatogr., A 2001, 925, 19-29. (14) Cavazzini, A.; Massi, A.; Bergamaschi, G.; Braga, S.; Dondi, F.; Dondoni, F. Biotechnol. Prog. 2004, 20, 603-612.
Analytical Chemistry, Vol. 77, No. 10, May 15, 2005 3113
poor chromatographic performaces if the polymerization degree, the grafting density, and the polymer architecture are not adequate. The advantages of these types of CSPs are the wide chemical and structural variability exploitable in the preparation of the chiral selectors and the chemical and thermal inertness of the packing material that derives from the covalent attachment of the chiral polymer to the solid support. The new hybrid silica/chiral polymeric material employed in this work was prepared through a so-called “g-from” polymerization approach, in which the initiators have been immobilized on the silica surface. This approach permits the achievement of a thin, ordered layer of polymeric material (“polymer brushes”) directly on the surface of mesoporous silica particles.15,16 The scope of this work is to study the adsorption equilibria of some racemic benzodiazepines on this new CSP. Different MP compositions have been investigated and the performance of the chromatographic system evaluated at the different conditions. Hypotheses regarding molecular phenomena possibly involved in the chiral recognition process have been made by considering (1) the chromatographic behavior, under linear conditions, of the differently substituted benzodiazepines lorazepam (LO), temazepam (TE), and oxazepam (OX) and (2) by measuring the competitive adsorption isotherms of LO and TE racemates under different MPs. Overloaded separations of racemic LO and TE have been modeled, and the polymeric DACH-ACR CSP was characterized with respect to its properties as preparative medium. EXPERIMENTAL SECTION Chemicals and MPs. TE, LO, and OX were prepared in our laboratory according to ref 17. 1,3,5-Tri-tert-butylbenzene (TTB) was purchased from Aldrich (Milwaukee, WI). The MPs consisted of solutions of dichloromethane and methanol. These solvents were HPLC grade compounds from Fluka-Riedel-De Hae¨n (Buchs, Switzerland). Under linear conditions, MPs were prepared by varying MeOH amounts between 0 and 5 v/v. Competitive adsorption isotherms were, instead, measured under the following MP compositions: 3 and 5% v/v MeOH in CH2Cl2 for LO; pure CH2Cl2 and 1% v/v MeOH in CH2Cl2 for TE. CSP Preparation. The CSP was prepared according to refs 15 and 16. Chromatographic System and Column. An Agilent-1100 series HPLC, equipped with a two-pump delivery system, a DAD, a manual injector (Rheodyne 7725i, Rheodyne, Cotati, CA), and a data acquisition system, was employed (Chemstation 9.01, Agilent Technologies). The column used for all the measurements was a 50 × 4.6 mm stainless steel column packed with hybrid polymeric DACH-ACR.16 The particle diameter was 5 µm and the pore diameter 300 Å. The holdup volume of the system, measured through TTB, was 0.52 ( 0.03 mL with the total porosity t ) 0.63. No significant changes in the holdup volumes were observed by changing the MP composition. All the measurements were performed at 1 mL/min flow rate and at 25 °C. Injection loops used were 2 and 160 µL for analytical and overloaded injections, respectively. The wavelength used in linear conditions for all the (15) Gasparrini, F.; Misiti, D.; Villani, C. PCT Int. Appl. CODEN PIXXD2 WO 2003079002, 2003. (16) Gasparrini, F.; Misiti, D.; Rompietti, R.; Villani, C. J. Chromatogr., A 2005, 1064, 25-38.
3114
Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
benzodiazepines was 242 nm. Calibration of the detector (overloaded conditions) was done at 342 nm. The number of theoretical plates (apparent efficiency of the column) estimated under linear conditions was ∼1200. This value (calculated on the second eluted compound) was used for all the simulations (see below). First moments of peaks were obtained by Agilent Chemstation (Version Release 9.01). Perturbation on Plateau Measurements and Inverse Method Calculations. Perturbation peaks were performed through 2-µL injections of pure CH2Cl2 pulses (vacancies) and detected at 206 nm. Five runs were repeated for each of the five MP compositions exploited (MeOH amount varied between 1 and 5% v/v). The average perturbation volumes were used for calculations. Inverse method calculations were performed by the simultaneous use18 of four experimental chromatograms (maximum total injected concentration =6 g/L). A simplex algorithm was employed for parameter optmization. THEORY Linear and Nonlinear Chromatographic Measurements. The expressions “infinite dilution” and “finite concentration” are used to refer to solute concentrations in any phase, be it a bulk phase or an adsorbed (interfacial) phase.19 At infinite dilution (or under linear conditions), all solute molecules are assumed to behave independently and properties under investigation vary linearly with concentration. For equilibrium properties, linearity implies a linear distribution isotherm.20,21 Chromatographic measurements are typically performed under these conditions. However, when chromatography is used to characterize adsorption surfaces or to investigate mechanisms involved in the recognition processes between eluates and stationary phases, measurements under nonlinear conditions are necessary.19,22 The adsorption data obtained are modeled through an isotherm model that intrinsically accounts for a specific adsorption mechanism.3,23-25 This information is also pivotal for separation modeling under overloaded conditions and for the achievement of proper experimental conditions under which to perform the separation. In practice, empirical limitations (most often solubility limitations) may not allow attainment of the pertinent information in a sufficiently wide concentration range to permit a discrimination between different models of adsorption. In such situations, the empirical adsorption isotherms are considered mere “working curves” to be used for optimization study.14 Gathering nonlinear adsorption information becomes especially important in chiral separations on CSPs since, in these cases, a heterogeneous adsorption mechanism has to be hypothesized to explain the separation. The chiral discrimination (17) Bell, S. C.; Childress, S. J. J. Org. Chem. 1962, 27, 1691-1695. (18) Marchetti, N.; Dondi, F.; Felinger, A.; Guerrini, R.; Salvadori, S.; Cavazzini, A. J. Chromatogr., A In press (19) Guiochon, G.; Shirazi, S. G.; Katti, A. Fundamentals of preparative and nonlinear chromatography; Academic Press: Boston, MA, 1994. (20) Conder, J. R.; Young, C. L. Physicochemical Measurement by Gas Chromatography; John Wiley & Sons: New York, 1979. (21) Giddings, J. C. Unified Separation Science; Wiley-Interscience: New York, 1991. (22) Huang, J.-X.; Horva´th, C. J. Chromatogr., A 1987, 406, 275-284. (23) Go ¨tmar, G.; Stanley, B. J.; Fornstedt, T.; Guiochon, G. Langmuir 2003, 19, 6950-6956. (24) Stanley, B. J.; Szabełski, P.; Chen, Y. B.; Sellergren, B.; Guiochon, G. Langmuir 2003, 19, 772-778. (25) Felinger, A.; Cavazzini, A.; Guiochon, G. J. Chromatogr., A 2003, 986, 207225.
process requires the existence of at least two transient diastereomeric analyte-CSP complexes (one for each enantiomer), whose thermodynamic stability is different due to specific molecular interactions. Linear measurements (namely, capacity factor values) cannot be used to differentiate between true and apparent contributions to resolution.11,26 It is the knowledge of the adsorption behavior over concentration ranges where nonlinearity effects become relevant that represents the key to understanding the adsorption fundamentals. Clearly no direct knowledge concerning molecular interactions involved in the recognition process will be obtained through mere chromatographic measurements, unless chromatographic data are coupled with information coming from more specific techniques, such as NMR investigations. A less sophisticated but still effective approach to use chromatography as a molecular-investigation tool27 relies on the employment of proper target compounds, systematically modified via organic synthesis to substitute/block chemical moieties that are responsible for specific interactions between these molecules and the adsorptive material.11 This is the typical approach of so-called inverse chromatography, which establishes the properties of an “unknown” stationary phase from its interaction with a “known” injected substance.27,28 When characterizing the conditions under which a chromatographic separation is carried out, it is common practice in preparative chromatography to use the adimensional parameter loading factor (LF), defined as
LF ) CinjVinj/qsVads
(1)
where Vinj and Cinj are the injected volume and concentration, respectively, qs is the saturation capacity, and Vads is the volume of the adsorptive material contained into the column. Isotherm Determination through Numerical Approaches. When the amounts of the target compounds whose adsorption isotherms have to be determined are reduced, their cost elevated, or both, traditional techniques for isotherm determination, such as frontal analysis (FA), perturbation on a plateau (PP), etc., may become inadequate. In fact, even when microtechnologies (micropumps, -columns, -detectors) are used to reduce the amount of material, these classical approaches appear globally expensive or simply too difficult to utilize.29,30 Additionally, if adsorption isotherms must be measured for the same compound under different experimental conditions, such as when gradient elution under overloaded conditions is planned, the problem may become insurmountable.18 In all these cases, numerical methods represent a solution to the problem of isotherm determination. Inverse methods (IMs) attempt to derive the isotherm parameters from recorded band profiles,25,31 in opposition to the traditional problem of chromatography that consists of calculating the band profiles of the (26) Go ¨tmar, G.; Fornstedt, T.; Guiochon, G. Chirality 2000, 12, 558-564. (27) Haidacher, D.; Vailaya, A.; Horvath, C. Proc. Nat. Acad. Sci U.S.A. 1996, 93, 2290-2295. (28) Vilcu, R.; Leca, M. Polymer Thermodynamics by Gas Chromatography; Studies in Polymer Science 4; Elsevier: New York, 1990. (29) Cavazzini, A.; Felinger, A.; Kaczmarski, K.; Szabelski, P.; Guiochon, G. J. Chromatogr., A 2002, 953, 55-66. (30) Zhou, D.; Kaczmarski, K.; Cavazzini, A.; Liu, X.; Guiochon, G. J. Chromatogr., A 2003, 1020, 199-217. (31) Dose, E. V.; Jacobson, S.; Guiochon, G. Anal. Chem. 1991, 63, 833-839.
component involved in the separation knowing their adsorption isotherms.19 In synthesis, the IM consists of the optimization of the isotherm parameters by minimizing the square of the sum of the differences between experimental band profiles and theoretical peaks obtained by solving a proper model of nonlinear chromatography. To do this, an adsorption isotherm model must be chosen a priori.25,31 The choice of a convex upward/downward isotherm is suggested by the shapes of the experimental peaks recorded under nonlinear conditions. This is a drawback of IM with respect to FA, which gives the very points of the isotherm. However, since only a few overloaded band profiles have to be measured (in addition to a calibration curve), the amount of material needed to obtain the thermodynamic adsorption data is drastically reduced, as is the time necessary to perform the measurements. Previous works25,32,33 have demonstrated that the quality of the information obtained through IM is elevated: the agreement between data collected with FA (or PP) and IM coincides substantially in the concentration range included between zero and the maximum concentration value recorded at the outlet of the column. (Incidentally, this is the range over which the numerical optimization of the parameters is effectively performed.)25 IM accuracy, instead, systematically degrades in the interval enclosed within the maximum concentration eluted and the injected concentration. This is the reason why short columns should be employed if IM is used for isotherm determination (see below). Equilibrium Dispersive (ED) Model of Chromatography. The simplest macroscopic model of nonlinear chromatography consists of a differential mass balance equation for any of the retained components and their equilibrium adsorption isotherms, measured under the actual experimental conditions, on the material with which the column is packed.19,34 The ED model of chromatography assumes that the adsorption equilibrium between mobile and stationary phase is reached instantaneously. In those situations in which mass-transfer resistances are significantly reducedsgenerally if the mass transfer is controlled by diffusion in mobile phase while adsorption-desorption kinetics are fasts the ED model represents a realistic approximation. This is often the case with molecules of small dimensions. Additionally, the constructive characteristics of the new-generation adsorptive materials, as well as the improved packing technology, very often make modern HPLC columns exhibit elevated efficiency values (as a rule of thumb, a few hundred theoretical plates represents an operational limit for expecting the ED model to work).19 According to the ED model, the system of differential mass balance equations for a mixture composed of two components adsorbing on the stationary phase, as in the case of the enantiomeric separations, is written as19
∂Ci ∂qi ∂Ci ∂2Ci + F + u DL 2 ∂t ∂t ∂z ∂z
i ) 1, 2
(2)
where t and z are the time elapsed from the injection and the (32) Cavazzini, A.; Felinger, A.; Guiochon, G. J. Chromatogr., A 2003, 1012, 139-149. (33) Felinger, A.; Gritti, F.; Guiochon, G. J. Chromatogr., A 2004, 1024, 21-38. (34) Guiochon, G.; Lin, B. Modeling for Preparative Chromatography; Academic Press: San Diego, CA, 2003.
Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
3115
distance traveled by the molecules inside the column, respectively, u is the linear mobile-phase velocity, qi ) fi(C1, C2) is the adsorption isotherm for the ith species, F is the phase ratio F ) (1 - )/ ( being the total porosity), and DL the so-called apparent dispersion that accounts for all the phenomena leading to band broadening in linear conditions (namely, eddy and axial diffusion and nonequilibrium effects). DL is evaluated by
DL ) HL/2t0 ) Hu/2
(3)
where H represents the height equivalent to a theoretical plate (HETP), L the column length, and t0 the holdup time of the column (t0 ) L/u). In practice, it is assumed that all compounds have the same HETP and that DL is independent of the concentration (at least in the range of concentrations investigated). To numerically solve the system of eq 2, initial and boundary conditions have to be defined. In this work, the following equations were used:3,35
initial condition Ci(0,z) ) 0
(4)
boundary condition at the column inlet (t > 0 and z ) 0) Ci(t,0) ) C′f,i
(5)
with C ′f,i defined according to
[
C if 0 < t < t C ′f,i ) f,i if t < t p 0 p
being tp the injection time and the subscript f indicating an inletvalue;
boundary condition at the column outlet (t > 0 and z ) L) (6)
The exacter Danckwerts condition, in which the effect of axial diffusion during the very injection is accounted for, was not used. It has been shown that when the column efficiency exceeds a few hundred plates (as in the present study), the differences in the theoretical profiles calculated by solving eq 2 with the Danckwerts condition or with the simpler eq 5 are negligible.19 The ED model was solved using a program based on a finite difference algorithm.19 Isotherm Determination through PP. A short overview of the PP technique will be given in this section. In the case of a one-component system and by neglecting all kinetic effects except convection (i.e., under the hypotesis of ideal chromatography), eq 2 becomes
∂C ∂q ∂C +F +u )0 ∂t ∂t ∂z
(7)
This equation (equation of continuity) represents the basis for the calculation of the isotherm parameters through PP. Its physical meaning19 is that the retention volume of a given concentration 3116
Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
(
| )
∂q VR(C′) ) V0 1 + F C)C′ ∂C
(8)
where ∂q/∂C|C)C′ is the slope of the isotherm at C ) C′ and V0 is the holdup volume of the system. The retention time of a small injection of a solute in a column equilibrated with the pure mobile phase (typical condition of analytical chromatography) therefore gives the retention time under linear conditions (i.e., at infinite dilution), through the expression
tR ) t0(1 + aF)
(9)
where a is the initial slope of the isotherm. Similarly, the set of retention times of perturbation peaks measured in a column equilibrated with streams of the mobile phase of increasing concentrations allows the calculation of the slope of the isotherm at these different concentrations. This set of retention times constitutes the fundamental data set needed for determining the adsorption isotherm. Adsorption Isotherm Models. To conclude, a concise survey of fundamental competitive adsorption models most often used in chiral HPLC includes.3,19,36 1. Langmuir model (adsorption on a homogeneous surface):
qi )
]
∂Ci/∂z ) 0
C′ is a function of the slope of the isotherm at C ) C′:
aiCi qskiCi ) 1 + k1C1 + k2C2 1 + k1C1 + k2C2
i ) 1, 2
(10)
where ai ) qski is the so-called Henry’s adsorption constant (slope of the isotherm at infinite dilution) and ki the adsorption equilibrium constant for the ith compound. 2. Bi-Langmuir model (adsorptive material made of two different kinds of adsorption sites):
qi )
ai,ICi ai,IICi ) 1 + k1,IC1 + k2,IC2 1 + k1,IIC1 + k2,IIC2
i ) 1, 2 (11)
where ai,I ) ki,Iqs, I and ai,II ) ki,IIqs,II. Equation 11 is obtained by simply considering the sum of two Langmuirian terms (denoted by I and II). In eqs 10 and 11, the same saturation capacity value has been assumed for both enantiomers on each adsorption site. This is a necessary condition to ensure the thermodynamic consistency of the adsorption competitive model.3,19,29 RESULTS AND DISCUSSION The physicochemical characterization (elemental analysis, differential scanning calorimetry, diffuse reflectance infrared spectroscopy, scanning electron microscopy, inverse size exclusion chromatography) of the hybrid silica/chiral CSP used in the present study has been discussed in detail in a related paper.16 For the sake of clarity, only the chemical structure of the monomer (35) Kaczmarski, K.; Antos, D.; Sajonz, H. G.; Sajonz, P.; Guiochon, G. J. Chromatogr., A 2001, 925, 1-17. (36) Jaroniec, M.; Madey, R. Physical Adsorption on Heterogeneous solids; Elsevier: Amsterdam, 1988.
Table 1. Retention Factors (k′) and Selectivities (r) of LO, TE, and OX at the Different MP Compositions (MeOH:CH2Cl2, % v/v) Exploited in the Worka LO
(DACH-ACR) constituting the base unit of this polymeric CSP is represented in Figure 1 (top). The numerous amide groups present on the polymeric support make it a weak H-bond donor (via amide NH group), a strong H-bond acceptor (through amide carbonyl group), or both.37 Additionally, since the surface density of chiral monomer was found to be ∼392 µmol/g of matrix,16 and each monomer possessing two chiral centers, the number of stereogenic centers available for diastereomeric associations is extremely large. Figure 1 also reports the chemical structures of the three benzodiazepines considered in the present study: LO, TE, and OX. The most significant difference between them is represented by the fact that the amidic nitrogen (N-1) is methylated in the case of TE. Additionally, LO brings an extra Cl on the nonfused phenyl ring. Because of these structural differences, the chemicophysical properties of the corresponding compounds change. This is particularly so regarding the hydrogen-bonding ability since an H-bond donor, the NH group (N-1), is absent on the secondary amide TE. The chromatographic behavior of these compounds has been evaluated under different chromatographic situations. LO and TE have been studied under both linear and nonlinear conditions and at different MP compositions; OX has, instead, been injected only at infinite dilution and for a specific MP composition (see below). Chromatographic Behavior under Linear Conditions. Linear chromatographic injections of LO and TE were performed under different MP compositions, obtained by varying the amount of MeOH between 0 (MP made of pure CH2Cl2) and 5% v/v. The effect of the amount of the strong MP modifier on the performance of the separation of LO and TE is summarized in Table 1, where the capacity factors (k′s) for the first and the second eluted (37) Dill, K. A. Biochemistry 1990, 29, 7133-7155.
OX
k′1
k′2
R
k′1
k′2
R
k′1
k′2
R
5 4 3 2 1 0
3.3 4.4 6.5 11 23 /
6.6 8.9 13.2 23 50 /
2.0 2.0 2.0 2.1 2.2 /
/ 0.12 0.15 0.21 0.34 1.3
/ 0.18 0.23 0.34 0.57 2.4
/ 1.5 1.5 1.6 1.7 1.8
3.3 4.4 6.5 11 23 /
5.4 7.6 11.0 18 40 /
1.6 1.7 1.7 1.6 1.7 /
a
Figure 1. DACH-ACR (S,S) monomeric unit and chemical structures of OX, TE, LO.
TE
MeOH %
See text for details.
compounds and the selectivities (Rs) obtained in all the cases are listed. It can be observed that both retention and R progressively decrease by increasing the amount of MeOH. This indicates a normal-phase mode chromatography. In normal-phase mode, the strong solvent acts by competing with the analytes, occupying a relatively large fraction of the adsorbent surface. Because the strong solvent is adsorbed as much as or even more than the analyte, the retention of analytes is reduced.19 (1) MP Modifier Adsorption Isotherm. Evidently, MeOH can interact with both the CSP and the analytes through hydrogen bonding. By using the PP technique,19,32 the adsorption isotherm of MeOH on the CSP was determined. Perturbation peaks were obtained through injections of vacancies38 after equilibrating the column at different MP compositions (MeOH amount varied between 1 and 5% v/v). In the limits of the experimental errors, the retention volumes of the perturbation peaks showed a slight decrease by increasing the amount of the strong modifier. Therefore, perturbation data were fitted via the simple singlecomponent Langmuir model:
q ) qskC/(1 + kC)
(12)
The plot reporting dq/dC versus C, is represented in Figure 2. Admittedly, only five experimental data points were considered due to the experimental difficulties encountered in evaluating perturbation data at MeOH concentrations smaller than 1% v/v (∼8 g/L MeOH in CH2Cl2). This was represented in Figure 2: the use of a dotted line in the concentration range below 8 g/L emphasizes that the value of dq/dC at the origin cannot be precise. Under these experimental limitations, the best estimates of the isotherm parameterssqs = 23 g/L, k ) 0.33 L/gswould seem to indicate that an MeOH concentration corresponding roughly to 3-4% v/v (i.e., 23-30 g/L MeOH in CH2Cl2) was enough to saturate the chiral support. (2) Role of the Amide N-1 in the Chiral Recognition Process. The data reported in Table 1 show that there were significantly different features in the behavior of LO and TE when the chromatographic conditions are changed. In the case of TE, in fact, both the resolution and the retention appeared almost completely lost even for an MeOH concentration roughly corresponding to 3-4% v/v. At 5% MeOH, then, completely unresolved (38) Blumel, C.; Hugo, P.; Seidel-Morgenstern, A. J. Chromatogr., A 1999, 865, 51-71.
Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
3117
Figure 2. Perturbation on plateau experimental data (points) and nonlinear fitting to the Langmuir model (continuous and dotted line). Best isotherm parameters through fitting: k ) 0.33 L/g, qs = 23 g/L. See text for details.
Figure 3. Chromatograms for TE recorded under linear conditions. Effect of the change of the strong MP modifier amount (MeOH concentration varied between 5%, top, and 0%, bottom). Vertical arrows, holdup time; Y, detector response (generic absorbance units).
(and almost unretained) peaks were eluted (in Table 1 the symbol “/” was used to indicate this fact). This trend is further evidenced in Figure 3, where the sequence of experimental chromatograms that show the progressive loss of resolution (and retention) in the separation of TE racemate due to the increase in the MeOH amount is represented. Conversely, LO enantiomers were resolved even when the amount of MeOH reached 5% v/v (R ) 1.99; see Table 1). LO 3118
Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
capacity factors were found to be several orders of magnitude larger than those of TE and impressively increased by decreasing the amount of methanol: when MeOH concentration was only 1% v/v, k′ values of about 24 and 50 for the first and the second eluted enantiomers were obtained. In pure CH2Cl2, these species were practically “stuck” in the column (the symbol “/” was again used in Table 1 to underline this fact). Various physicochemical phenomena can be invoked to provide an interpretation of these data. The strongly different properties in terms of hydrogen-bonding ability of LO and TE may be considered as the most important factor. Amide CdO groups of the CSP may accept H-bonds from the NH group of LO (but evidently not from TE). Conversely, amide NH groups on the CSP can interact with a strong H-bond donor, such as the CdO group of both TE and LO. Additionally, the competitive effect by MP modifier molecules has to be taken into account. MeOH molecules may affect the adsorption equilibria in two ways: OH groups (Hbond donors) of MeOH can interact with CSP amide CdO groups or they may donate hydrogen bond to amide carbonyl groups of benzodiazepines. Through these retention data, some observations were possible. First, the interactions between NH groups (N1) of benzodiazepines and CSP carbonyl moieties were of primary importance both for chiral recognition and for chromatographic retention. On the other hand, interactions between (H-bond donor) NH groups on the chiral support and CdO groups of benzodiazepines did not seem to have the same importance. This would explain why LO enantiomers were separated, despite the amount of MeOH in the MP, while TE enantiomers were only slightly retained and poorly separated (see Table 1). Without the amidic NH group N-1, TE was unable to behave as an H-bond donor and endured a “strong” competition for the adsorption on the active sites by MeOH molecules. This effect became dramatically evident by increasing the amount of the strong MP modifier (see Table 1 and Figure 3). MeOH competition was, instead, significantly weaker in the case of LO. LO has a strong H-bond donor character with respect to MeOH: the MP modifier amount affected the retention without determining significant changes in selectivities (see also below). To further bolster the hypothesis about the role of the amidic nitrogen N-1 in the chiral recognition process, another benzodiazepine, OX, was employed as target compound. OX differs from TE only because nitrogen N-1 is not methylated (see Figure 1). With respect to LO, OX does not bring any chlorine on the nonfused ring. Figure 4 summarizes the results of this investigation. Herein, the chromatograms corresponding to linear injections of TE, LO, and OX are overlapped (incidentally, the TE chromatogram corresponds to that reported on the top of Figure 3) in the case of a MP made of MeOH:CH2Cl2 ) 5:95% v/v. It is interesting to observe that OX enantiomers were separated under these experimental conditions (R = 1.6; k′1 ) 3.3, k′2 ) 5.4), which represented further evidence of the importance of a H-bond donor N-1. Figure 4 suggested also another interesting observation. k′s for the first eluted enantiomers of LO and OX were found to be equal (3.3), while retentions of the later-eluted enatiomers were noticeably different (k′2,OX = 5.4; k′2,LO = 6.6). Since, for both LO and OX, R and S enantiomers were chromatographated with a same elution order (as demonstrated through CD measurements,
Figure 4. Chromatographic separation of TE, LO, and OX. MP, 95:5 ) CH2Cl2:MeOH, % v/v. Vertical arrow, holdup time; Y, detector response (absorbance at 242 nm).
chromatograms not reported) this observation suggested that the way in which R and S enantiomers interacted with the CSP was substantially dissimilar. Since OX and LO differ only in the nonfused aromatic ring, an explanation of the observed chromatographic behaviors was that the nonfused ring did not take part in the adsorption processes of the formerly eluted compounds (same k′s), while it played an active role in those of the latter chromatographated enantiomers. According to this interpretation, the differences in retention and R between LO and OX second eluted enantiomers corresponded to the contribution due to chlorine. OX retention data, measured under the different MP compositions investigated in this work, confirmed the above hypothesis. These data are reported in the third column of Table 1. Chromatographic Behavior under Overloaded Conditions. The chromatographic behavior of LO and TE was also evaluated under overloaded conditions with the purpose of modeling their competitive separation. Since MeOH was demonstrated to adsorb on the CSP, a three-component system should be considered. To do this, knowledge of the adsorption isotherms of MeOH and those of pure TE and LO (i.e., their adsorption isotherms from solutions of pure CH2Cl2) is required.19 The data reported in Table 1, however, showed the impossibility of gathering this information for LO, as it stayed on the chromatography column at MeOH 0% v/v. Consequently, from a modelistic point of view, the system will be regarded as a two-component case and MeOH competitive effects somehow “lumped” in the measured adsorption parameters. Because of the significantly reduced amount of available compounds (∼100 mg for both of benzodiazepines), the IM was used to determine their adsorption isotherms. Additionally, a relatively short (50 × 4.6 mm) column was employed. On one hand, this allowed for a [further] reduction of the material needed to overload the column. On the other hand, using short columns contributes to decreased dispersion effects causing band broadening (axial dispersion, external and intrapore mass-transfer resistances, adsorption-desorption effects). This is important since it has been demonstrated that, the “sharper” the peak recorded at the column outlet, the more accurate the information provided by IM.25 As previously discussed, IM requires the a priori choice of a theoretic adsorption model, which is suggested from the peak
Figure 5. Comparison between empirical peaks (points) and simulation results (continuous and dotted lines) for different racemic solutions of LO. Cinj, injected concentration; MP, MeOH:CH2Cl2 ) 5:95% v/v. Peak simulations obtained by solving system of eqs 2 with isotherm parameters reported in Table 2.
shapes of the empirical profiles. Figures 5-8 report (with points) several overloaded band profiles recorded for TE and LO under various experimental conditions (see further for details). The shapes of the chromatographic profiles clearly indicated that the adsorption isotherms of these benzodiazepines on this chiral support are characterized by a Langmuirian (convex upward) form. (1) Choice of an Adsorption Isotherm Model and Separation Modeling. In the case of chiral separations, the bi-Langmuir isotherm (eq 11) is one of the models most often used to account for the adsorption data. Under the hypothesis that the two kinds of adsorption sites are enantioselective and nonselective, eq 11 can simply be rewritten as (5-parameter bi-Langmuir model):
qi )
ai,ICi aIICi + 1 + k1,IC1 + k2,IC2 1 + kII(C1 + C2)
i ) 1,2 (13)
where aII ) qs,IIkII. The actual recognition process between enantiomers and CSP happens on selective sites (I), while nonselective sites (II) merely contribute to retention. On them, both enantiomers behave identically (same equilibrium adsorption constant, kII). Despite its simplicity, the five-parameter bi-Langmuir model has been demonstrated to be able to satisfactorily account for different cases of chiral separations via HPLC.3,19,26,39 This is due to the fact that enantiomeric systems fit most of the assumption under which the bi-Langmuir model holds.39 Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
3119
Figure 6. Comparison between empirical peaks (points) and simulation results (continuous and dotted lines) for different racemic solutions of LO. MP, MeOH:CH2Cl2 ) 3:97 v/v.
In preparative works, an important validation of an adsorption isotherm model adequacy is represented by the agreement between empirical profiles and theoretical peaks. These last are obtained by solving a model of nonlinear chromatography (such as eqs 2) when the measured adsorption data are employed. In this work, eq 13 was used. The five parameters, optimized through IM calculations at the different experimental conditions investigated, are reported in Table 2 and the corresponding isotherms in Figure 9. To numerically perform this optimization, chromatograms corresponding to racemic solutions of roughly 7 g/L total concentration were employed (figures not reported). The biLangmuir model was then used to solve the system of eqs 2, by employing a program based on a finite difference algorithm (Rouchon’s scheme).19 Figures 5-8 summarize the results of these calculations by comparing experimental profiles and simulated peaks under different experimental conditions (see figure captions for details). The remarkable agreement observed between simulations and real data provided an important indication of the adequacy of the isotherm model considered. Admittedly, in some cases there are minor differences between experimental and simulated profiles, especially in the rear parts of the second eluted compounds. These differences could suggest the existence of more complex adsorption models accounting for a continuous adsorption energy distribution (such as, for instance, the To´th model19). (39) Jacobson, S.; Golshan-Shirazi, S.; Guiochon, G. J. Am. Chem. Soc. 1990, 112, 6492-6498.
3120 Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
Figure 7. Comparison between empirical peaks (points) and simulation results (continuous and dotted lines) for different racemic solutions of TE. MP, MeOH:CH2Cl2 ) 1:99 v/v. Since in this case peaks overlap, the sum of the first and the second eluted peaks is also represented (first + second). This would correspond to a hypothetical concentration signal recordable at the column outlet.
(2) CSP Performances. Two MP compositions, corresponding to 3 and 5% v/v MeOH in CH2Cl2, were considered for LO. Smaller amounts of MP modifier leaded to excessively large k′s, which are quite unpractical, especially for preparative purposes. In the case of TE, the choice of the investigated MPs (MeOH: 0 and 1% v/v; see Table 2) was due to the fact that, even when the amount of MeOH was 2% v/v, the compounds appeared almost unretained. A first relevant conclusion reachable from the data reported in Table 2 and from Figures 5-8 was that the polymeric DACHACR chiral support exhibited excellent properties as preparative separation medium for the important benzodiazepine class of compounds. This was especially the case for those bringing an amidic group N-1 (such as LO and OX). Even though a detailed investigation of the effects caused by the variables commonly employed in preparative chromatography, such as production rate, flow rate velocity, etc.,40-45 was not purpose of this work, there were consistent indications of the excellent perfomance of the new adsorptive material for preparative applications. (40) Golshan-Shirazi, S.; Guiochon, G. Am. Biotechnol. Lab. 1990, 8, 26-34. (41) Golshan-Shirazi, S.; Guiochon, G. J. Chromatogr. 1990, 511, 402-403. (42) Felinger, A. Data analysis and signal processing in chromatography; Elsevier: Amsterdam, 1998. (43) Felinger, A.; Guiochon, G. J. Chromatogr. 1992, 591, 31-45. (44) Felinger, A.; Guiochon, G. AICHE J. 1994, 40, 594-605. (45) Jacobson, S. C.; Felinger, A.; Guiochon, G. Biotechnol. Prog. 1992, 8, 533539.
Figure 8. Comparison between empirical peaks (points) and simulation results (continuous and dotted lines) for different racemic solutions of TE. MP, pure CH2Cl2. Table 2. Isotherm Parameters (Bi-Langmuir Model, Eq 13) Obtained through IMa compd LO TE
MeOH %
a1,I
5 3 1 0
2.8 5.9 0.81 0.010
a2,I
k1,I (L/g)
k2,I (L/g)
aII
kII (L/g)
9.3 18 1.2 1.8
0.15 0.24 0.0075 0.0028
0.49 0.73 0.011 0.51
3.2 4.9 0.0065 2.4
0.036 0.053 0.00052 0.12
a a , adsorption Henry constant of jth enantiomer on ith site (I, j,i selective site; II, nonselective site). kj,i, adsorption equilibrium constant of jth enantiomer on ith site.
For instance, the LO chromatogram represented at the top of Figure 5 (Cinj ) 5 g/L) was obtained with a LF roughly equal to 3% (calculated according to eq 1). At this relatively elevated LF, the achievement of a baseline separation was an undoubtable sign of the elevated separation efficiency of this material. The chromatographic system could be further overloaded without dramatically losing “separation quality”. Since saturation capacities estimated in this work were based on data measured at relatively small Lfs, their physical meaning has to be carefully evaluated. Having said this, through data reported in Table 2, it is possible to hypothesize that the relative amount of chiral sites on the chromatographic support changed by modifying the amount of MeOH. This is demonstrated by the values of the ratios between selective-site saturation capacities (qs,I ) ai,I/ki,I; see Table 2) and total saturation capacities (sum between qs,I and nonselective qs,II saturation capacity) at the different MPs
Figure 9. Adsorption isotherms (bi-Langmuir model) for the different cases exploited. First, second, first and second eluted compounds; MeOH %, strong MP modifier concentration, % v/v. (Best isotherm parameters in Table 2.)
considered. (At 3% MeOH v/v, ∼20% of the CSP would result paved with chiral sites, while at 5% MeOH v/v, this fraction would drop down to 0.17). On the other hand, MeOH did not have a significant effect on the stability of the transient complexes between former and latter eluted enantiomers and CSP. This was reflected by the values of k1,I/k2,I ratios (at the two MP compositions), which practically did not change when MeOH amount increased from 3 to 5% v/v (0.32 vs 0.31, respectively). A conclusion from these data was that solvents useful for competing with hydrogen bonds from benzodiazepines such as LO and OX are those that are stronger H-bond donors than N-H. As observed under linear conditions, TE chromatographic behaviors were substantially different from those of LOs due to the negative effect of MeOH on retention and R. The data measured under nonlinear conditions obviously also reflected these trends. Since at MeOH:CH2Cl2 ) 1:99% v/v TE enantiomers were almost unretained, their adsorption isotherms were expected to be practically linear. This is clearly shown in Figure 9 (bottom) where TE competitive isotherms are represented. Linearity of adsorption isotherms had some important practical consequences. First, it explained the experimental peak shapes observed in Figure 7: these forms are typical of systems in which only volume, but not concentration, overloading can be achieved.14 Second, it makes the physical meaning of the calculated isotherm parameters under these conditions (third line of Table 2) highly questionably, since the use of a five-parameter model to fit almost linear data is not justified. This is especially true for saturation capacity values, Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
3121
which are extrapolations made at infinite concentration. However, the fact that predicted peaks and experimental profiles, matched in a significantly consistent way (as shown by Figure 7), authorizes the use of these isotherms as empirical “working curves”, to be employed for the prediction of the overloaded band profiles and the definition of proper experimental conditions for preparative goals. Figure 8 refers to the overloading separation of TE enantiomers in pure CH2Cl2. The isotherm parameters corresponding to this case are reported in the fifth line of Table 2 and the corresponding isotherms plotted in Figure 9 (bottom). The top of Figure 8 (Cinj ) 5.2 g/L) corresponds to a LF of ∼16%. Although in pure CH2Cl2 the simulation results showed some discrepancies with respect to the empirical profiles (especially for the more retained enatiomer), the overall agreement was quite positive (the positions of the first and second eluted compound shocks were always accurately predicted). The conclusion is that, although with a slightly poorer performance than in the case of LO, HPLC TE preparative purication can be satisfactorily achieved on this CSP. Finally, a closer analysis of the isotherm parameters indicated that, in the case of TEs, the two enantiomers behaved in very different manners on selective sites: the second being ∼180 times more strongly retained than the first (as shown by the ratio between k1,I and k2,I). Based on previous observations (see Discussion about data measured under Linear Conditions), we concluded that in the chiral recognition process the less retained enantiomer did not interact with the CSP through the nonfused aromatic ring. These data seemed to give a further indication of the pivotal role of this ring in stabilizing the diastereomeric complex between second eluted enantiomer and active sites on the CSP. CONCLUSIONS In this work, the adsorption equilibria of the differently substituted benzodiazepines, temazepam, lorazepam, and ox-
3122
Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
azepam, were evaluated on a new hybrid polymeric CSP in normalphase mode. Although the adsorption process on this kind of CSP can involve several different physicochemical phenomena, the role of hydrogen bonding was found to be pivotal in the chiral recognition process. In particular, major molecular interactions were supposed to involve H-bond-donor amidic groups of benzodiazepines and H-bond-acceptor carbonyl groups present on the CSP. The stronger retention of the second eluted enantiomer seemed related to an extra-stabilization of the complex between CSP and benzodiazepines due to the nonfused rings of benzodiazepines. The importance of the MP modifier upon the performance of racemic benzodiazepine separations was demonstrated: the competition for the adsorption played by the MP modifier is effectively connected to its relative H-bond donor properties with respect to those of the benzodiazepines to be separated. The feasibility of overloaded HPLC separations of racemic benzodiazepines on this new chromatographic support was demonstrated by modeling the competitive separation of racemic benzodiazepines under nonlinear conditions. DACH-ACR CSP demonstrated optimal characteristics as chromatographic preparative medium: high efficiency, large selectivity, and elevated loading. ACKNOWLEDGMENT This work has been supported by the Italian University and Scientific Research Ministry (Grant 200-3039-537) and by the University of Ferrara (ex 60%).
Received for review December 23, 2004. Accepted March 6, 2005. AC048101T