696 ENANTIOMER SEPARATIONS 0 5 10 15 min (R) (S) O NH OH H 3 C CH 3 Figure 14.16 Separation of propranolol enantiomers on a protein column (cellobiohydro- lase I; Chiral CBH I, ChromTech). Mobile phase: 0.01M acetate buffer, pH 5. Reprinted with permission from [83]. extreme conditions, can easily lead to chemical or biochemical changes of the protein selectors; this can adversely impact column longevity or enantioselectivity. Another disturbing limitation is a low to moderate column efficiency, as seen in the example separation of Figure 14.16 (N = 300–700). Even with large enantioselectivity values (as in Fig. 14.16), low plate numbers, combined with peak tailing, can make the analysis of 0.1% enantiomeric impurities impossible in some cases. This is especially important when the impurity is the second peak, as noted in Section 14.3; unfortu- nately, a simple switch to the CSP with opposite configuration and reversed elution order (to minimize the effects of peak tailing) is not possible for protein columns. For some samples, AGP and OVM display a (unpredictable) reversal of elution order. The complexity of the protein selectors limits insights into chiral recognition phenomena; yet several studies have been undertaken in this direction [96–103]. The value of protein phases is also considerably reduced by their very limited sample loadability [61]; the number of specific (enantioselective) interaction sites on the surface of the (large) protein molecule is usually small. The immobilization of smaller protein fragments that include the enantioselective sites might be expected to lead to higher sample loadability, but such experiments have so far proved unsuccessful [99, 100]. Consequently protein CSPs are of little value for preparative separations. To summarize, the role of protein phases for enantiomer separations is today slowly declining; these columns are more and more being replaced by other CSPs (e.g., polysaccharides and macrocyclic antibiotics). While the use of protein columns for quality control is discouraged, they may still have some relevance in bioana- lytical research, not least because of the compatibility of their mobile phases with electrospray interfaces (as opposed to typical normal-phase packings) [104]. 14.6 CHIRAL STATIONARY PHASES AND THEIR CHARACTERISTICS 697 14.6.4 Cyclodextrin-Based CSPs Cyclodextrin (CD) bonded CSPs, introduced by Armstrong [105], are based on α-, β-, or γ -cyclodextrins: macrocyclic structures that are assembled from 6, 7, and 8 glucose units, respectively (Fig. 14.17). The glucose (glucopyranose) units are connected via α-1,4-linkages. These macrocyclic molecules adopt the shape of a truncated cone, and the number of glucose units determines the size of the cavity (0.57, 0.78, and 0.95 nm in diameter for α, β,andγ -CD, respectively) into which (preferentially hydrophobic) solutes or substituents of solutes can be inserted. The internal surface of the CD-cavity is hydrophobic, due to the carbon backbone of the sugar moieties. The upper and lower rim surfaces are hydrophilic, due to the presence of the primary hydroxyls (lower narrower rim) and secondary hydroxyls (upper wider rim). These CDs are usually bonded to silica gel at the narrower ring hydroxyls, either via ether linkage (Cyclobond columns, from ASTEC) or carbamate linkage (ChiraDex and ChiraDex Gamma, from Merck; Ultron ES-CD from Shinwa). Aside from CSPs with native CDs as selector, a few CD derivatives are also available (e.g., Cyclobond I SP, RSP, RN, SN or Ultron ES-PhCD)(see Fig. 14.17). Derivatization may enhance chiral recognition by providing additional interaction sites or by varying the size and access of the chiral inclusion cavity. Hence CSPs based on derivatized CDs can exhibit significantly different enantioselectivity, compared to their underivatized counterparts (native CDs). Commercial columns: Cyclobond I (native β-CD), II (native γ-CD), and III (native α-CD) (from ASTEC) Cyclobond I SP or RSP [(S)- or (RS)-2-hydroxypropylether-β-CD] (ASTEC) Cyclobond I RN or SN [(R)-or (S)-1-(1-naphthyl)ethylcarbamate-β-CD] (ASTEC) ChiraDex (native β-CD) and ChiraDex Gamma (native γ-CD) (from Merck) Ultron ES-CD (native β-CD) and Ultron ES-PhCD (phenylcarbamoylated β-CD) ( from Shinwa ) α-Cyclodextrin β-Cyclodextrin γ-Cyclodextrin n = 0, m = 6 n = 1, m = 7 n = 2, m = 8 O O OH O OH O H O O OH OH HO O O OH OH HO O O OH HO HO O O HO HO OH O O HO HO OH O HO HO OH n 2 3 6 [OH] m lower rim upper rim m, number of glucose units [OH] 2m Figure 14.17 Structure of cyclodextrins and trade names of corresponding CSPs. 698 ENANTIOMER SEPARATIONS CD-based CSPs are truly multi-modal with regards to both elution conditions and chiral recognition mechanisms [18, 106]. They can be operated in RP, NP, and polar-organic modes. The molecular recognition mechanism was shown to differ for the different elution modes. It is commonly understood for the RP mode using aqueous or hydro-organic mobile phases (acetonitrile-buffer or methanol-buffer mixtures) that lipophilic solutes interact with CD selectors by inclusion complexa- tion (Fig. 14.18b). A two-step (hydrophobic) mechanism has been proposed [107]: (1) penetration of the hydrophobic part of the analyte molecule into the CD cav- ity and (2) release of associated solvent (water) molecules from the analyte and CD molecules, followed by complex stabilization through hydrophobic interaction. Moreover hydrophilic interactions with hydroxyl groups at the upper and lower rim (hydrogen bonding, dipole–dipole interactions) may take place and positively con- tribute to complex stabilities. These combined effects plus conformational changes upon complexation (i.e., induced fit) can sometimes lead to extraordinary complex stabilities. As in RPC, the concentration of organic solvent in the mobile phase can be used to control retention (for 1 ≤ k < 10). Solvent strength for CD-based CSPs increases in similar fashion as for achiral RPC: water < methanol < ethanol < propanol ∼ acetonitrile < tetrahydrofuran The extent of analyte inclusion generally depends on the size of the CD cavity. According to this size-fit concept of inclusion complexation, higher affinity and greater enantioselectivity for the CSP-analyte pairs generally occur for the CD that gives the best match in terms of the size of hydrophobic portions of the solute with the CD cavity. Substituted phenyl, naphthyl, and heteroaromatic rings can conveniently be accommodated in a β-CD cavity, while larger analytes such as steroids fit preferentially into γ -CD, and smaller analytes prefer α-CD. This structure-binding relationship can be a helpful initial guide for column selection. In the polar-organic mode (e.g., acetonitrile-methanol mixtures) the inner CD cavity is blocked by solvent molecules, so that inclusion complexation of lipophilic residues becomes thermodynamically unfavorable. Hence, solutes with O OH N reversed-phase mode (inclusion complexation) H OH N polar-organic & normal-phase mode (a)(b) O C C Figure 14.18 Molecular recognition mechanisms for cyclodextrin columns in polar-organic or normal-phase modes (a) and reversed-phase mode (b). Adapted from [108]. 14.6 CHIRAL STATIONARY PHASES AND THEIR CHARACTERISTICS 699 hydrophilic groups bind to the polar surface of the CD (either upper or lower rim; see Fig. 14.18a) [108]. The polar hydroxyls are surrounded by a chiral environment, with enantioselectivity resulting from differences in the strength of these polar interactions (hydrogen bonding, dipolar interactions) for the two enantiomers. Solutes with more than one polar functional group, one of which is located at or close to the stereogenic center, are particularly amenable for cyclodextrin columns and the polar-organic mode—with great promise for a successful separation. Bulky groups near the stereogenic center facilitate the enantiorecognition process. Since many chiral drugs are polar, this elution mode can be quite useful—especially when RP and NP modes fail to resolve the enantiomers [109]. These polar-organic mobile phases for CD-based CSPs are typically composed of 0–15% methanol in acetonitrile plus 0.001– 1.2% glacial acid plus 0.001–1.2% triethylamine and, for example, have been employed for enantiomer separations of β-blockers [110] . In the NP-mode (e.g., hexane or heptane mixtures with polar solvents such as alcohols), the internal CD cavity is also occupied by solvent molecules so that inclusion complexation does not occur. Interaction instead takes place at the polar surface of the CD or CD derivative. Only aromatic-derivatized CDs such as (R) or (S)−1-(1-naphthyl)ethyl carbamates have been shown to be effective CSPs for enantiomer separation in the normal-phase mode. In that case polar interactions such as hydrogen bonding and dipole–dipole interactions are supported by π–π interactions, which seem to become crucial in this mode. Since inversions of elution orders have been observed when the configuration at the stereogenic center of the rim substituent was reversed, it can be concluded that enantioselective analyte binding is primarily controlled by the attached 1-(1-naphthyl)ethyl carbamate groups, rather than by the supporting CD cavity. In recent years the importance of CD-based CSPs appears to be declining. Their applications overlap those of the more powerful macrocyclic antibiotics, which usually generate higher levels of enantioselectivity. Furthermore their preparative application is limited by a low sample-loading capacity. However, the compatibility of RP and polar-organic modes with the use of ESI-MS remains an important advantage for CD bonded CSPs. 14.6.5 Macrocyclic Antibiotic CSPs Inspired by the stereoselective inclusion capabilities of macrocyclic cyclodextrins, Armstrong and coworkers have investigated other (more effective) macrocyclic nat- ural compounds based on inclusion-complexation properties. From this research a new important CSP class was developed: the macrocyclic antibiotic CSPs. The first described CSP of this class was vancomycin-modified silica, introduced by Armstrong in 1994 [111]. Several structural analogues of this glycopeptide antibi- otic were subsequently found to be powerful chiral selectors with complementary enantioselectivity (different enantioselectivities and elution orders), leading to CSPs based on vancomycin [111], teicoplanin [112], ristocetin A [113], and the aglycone of teicoplanin [114], all of which have been commercialized by Astec under the tradenames Chirobiotic V, Chirobiotic T, Chirobiotic R, and Chirobiotic TAG. Chirobiotic V and T are available with older (V1, T1) as well as newer (V2, T2) bonding chemistry and/or silica support. These CSPs have a broad range of application—which includes very polar compounds such as underivatized amino 700 ENANTIOMER SEPARATIONS acids; these are today among the most powerful CSPs available. They have been described in several reviews [115–118]. Antibiotic selectors possess considerable structural complexity (Fig. 14.19). They share a common heptapeptide aglycone core, with aromatic residues that are bridged to each other to form a basket-like shape with shallow pockets for inclusion complexation and surface-confined carbohydrate moieties (Fig. 14.20). Inclusion complexation is often driven by polar interactions, notably for carboxylic acid-containing solutes: triple hydrogen-bonding of the carboxylic terminus (sup- ported by other hydrogen bonding interactions), as well as π –π and hydrophobic interactions (RP mode only). Together with the enantioselectivity that originates from multiple stereogenic centers, the heterogeneous multifunctionality and struc- tural specificity of the glycopeptides provides a variety of potentially stereoselective binding options that appear to be the origin of enantioselectivity for a variety of pharmaceuticals. Apart from their favorable structural features, the broad applicability of the macrocyclic antibiotics CSPs may also arise in part from their multimodal usage, which comprises RP, PO, and NP modes (Fig. 14.21) [111, 118, 119]. The polar-organic mode is recommended as first choice, if the solute has more than one polar functional group (which applies to many chiral drugs), and if at least one of these groups is located at or close to the stereogenic center. Since hydrophobic interactions are disrupted by the PO mode, polar interactions such as ionic and dipole interactions as well as hydrogen bonding are of primary importance for enantioselectivity. A typical starting mobile phase is composed of methanol plus 0.1% acetic acid and 0.1% triethylamine as additives for the vancomycin CSP (additives should be increased to 1% for each for the teicoplanin CSPs). The ratio of acetic acid to amine is a key variable for adjusting enantioselectivity, while the total amount of the additives at given ratio mainly adjusts retention. If the solute cannot be eluted in the PO mode with about 1% of additives, the compound is too polar, and the RP mode is recommended. Typical starting conditions for the RP mode are mobile phases composed of THF-20 mM ammonium nitrate pH-5.5 (10:90; v/v) for vancomycin CSP, and methanol-0.1% triethylammonium acetate pH-4.1 (20:80; v/v) for the teicoplanin CSP. More detailed method-development procedures are provided in the Chirobiotic Handbook [119]. In the RP mode, inclusion complexation may be driven or strongly supported by hydrophobic interactions, while multiple hydrogen bonds as well as ionic and dipole interaction contributions may be responsible for the stereoselective alignment of the solute in the binding cleft of the macrocyclic selector. If solutes cannot be sufficiently retained in the PO mode with as little as 0.01% of additives, or if solute solubility is limiting, the NP mode is a viable alternative. Typical starting conditions are a mixture of hexane-ethanol (80:20; v/v). Some other common elution conditions for different elution modes can be inferred from the caption of Figure 14.21. Overall, the use of different elution modes adds to the broad applicability of the macrocyclic antibiotic CSPs, and offers a tremendous flexibility of these CSPs for method development. It is further remarkable that these CSPs can be changed from one mode to the other without irreversible changes in performance. From a statistical evaluation by the column supplier [120], which was later confirmed by researchers from industry [121], it was found that the highest success rate can be achieved with the polar-organic mode (∼40%) or the reversed-phase 14.6 CHIRAL STATIONARY PHASES AND THEIR CHARACTERISTICS 701 N H HN N H N H N H NH O O O O O O O O O O O O NH 2 OH OH OH NH NH 2 OH Cl HO Cl OH HO OH HO H 3 C H 3 C OH Vancomycin A B C O N H N H N H Cl N H H O O H H OH HO O OH O O O HN O OH O O OH Cl OH OH NH O HO HO HO H H O H 3 C HO HO HO O O N H O NH 3 O O OH OH R A B C D Teicoplanin O O N H N H N H N H O OH O O O O N H O HO OHCH 3 O NH 2 O O HO OH H H H H H O O O OH OH O O O O HO HO HO HO OH OH O HO H OH CH 3 O NH 2 HO H 3 C O O OH OH OH OH A B C D Ristocetin A (a) (b) (c) O O CH 3 CH3 H 3 C CH 3 Figure 14.19 Structures of glycopeptide antibiotics: (a) Vancomycin (1 sugar moiety; 3 inclu- sion cavities A, B, C; molecular weight ≈ 1449 Da; pKa, 2.9, 7.2, 8.6, 9.6, 10.4, 11.7; pI, 7.2); (b) teicoplanin (molecular weight ≈ 1885 Da; 3 sugar moieties; R, decanoic acid residue; 4 inclusion cavities, A, B, C, D); (c) ristocetin A (molecular weight ≈ 2066 Da; 6 sugar moieties; 4 inclusion cavities A, B, C, D). 702 ENANTIOMER SEPARATIONS (a) (b) Figure 14.20 Basket-like shape and molecular recognition mechanism of vancomycin. X-ray crystal structure of a complex of vancomycin with N α , N ω -diacetyl-L-Lys-D-Ala-D-Ala. (a) Side view; (b) top view. X-ray crystal structure image was generated with SYBYL molecular modeling software (Tripos, St. Louis, MO) from fractional coordinates extracted from the Brookhaven protein databank (http://www.rcbs.org/pdb/). mode (∼40%), while normal-phase conditions appear to be less useful (∼5%). In this context it should be noted, however, that many of the reported ‘‘RP’’ separations (with hydro-organic mobile phases) were actually HILIC-type (i.e., aqueous NP separations; Section 8.6), as pointed out by Wang et al. [122]. Actual RP separations can be obtained on the Chirobiotic phases with mobile phases that contain less than 20% organic solvent, while—for higher-% organic—polar interactions may be reinforced by the low dielectric constant media, and retention increases with increase in the organic solvent (suggesting HILIC behavior). Although different macrocyclic antibiotic selectors have a close structural resemblance, they feature somewhat complementary enantioselectivity (Fig. 14.22), which is often helpful for method development. When a particular glycopeptide antibiotic CSP gives marginal or no separation, there is a fair chance that one of the other antibiotic CSPs can provide baseline separation. Subtle differences in the CSP structure and binding properties may be responsible for their complementary separation profiles; these have been attributed to distinct end-to-end distances of 14.6 CHIRAL STATIONARY PHASES AND THEIR CHARACTERISTICS 703 CHNHC O CH 3 NO 2 NO 2 N O NHCCH 2 NHCCH 2 Br OO 01530 time (min) MeOH/AcOH/ TEA MeOH/ NH 4 AcO MeOH/TFA/ NH 4 OH HO OH H N C(CH 3 ) 3 OH Terbutalin α= 1.38 α= 1.30 α= 1.30 (a) (b) 01326 time (min) CHCN(C 2 H 5 ) 2 CH 3 O O Devrinol N H N Br O CHCH 2 CH 3 CH 3 O Bromocil O O OH Cl O Coumachlor RP PO (c) NP Figure 14.21 Separations by different elution modes with a macrocyclic antibiotic (vancomycin) column. (a) Reversed-phase (RP) mode; (b) polar-organic (PO) mode; (c) normal-phase (NP) mode. Experimental conditions: (a) 10:90 acetonitrile/1% tri- ethylammonium acetate buffer, pH-7; (b) mobile phases from left to right: 100:0.3:0.2 MeOH/AcOH/TEA; 99:1 MeOH/20 mM NH 4 AcO; 100:0.05:0.05 MeOH/TFA/NH 4 OH; (c) 50:50 hexane/2-propanol. Adapted from [111] and [119]. 704 ENANTIOMER SEPARATIONS Ristocetin CSP Teicoplanin CSP Vancomycin CSP O NH CH 3 O O OH N H N H O O CH 3 8.75 12.4 5.92 14.8 Time (min)Time (min) Time ( min ) Time ( min ) Time ( min ) Time (min) 10.5 7.88 5.61 11.2 3.73 5.14 4.70 4.91 Figure 14.22 Complementarity of separation by different macrocyclic antibiotic columns, exemplified by enantiomer separations of Z-Ala (top) and 5-methyl-5-phenylhydantoin (bottom). Experimental conditions: Ristocetin A, 20:80 methanol/0.1% triethylammonium acetate buffer, pH-4.1; Teicoplanin, 20:80 methanol/1% triethylammonium acetate buffer, pH-4.1; vancomycin, 10:90 methanol/1% triethylammonium acetate buffer, pH-4.1. Adapted from [124]. the C-shaped aglycone (which becomes smaller in the order vancomycin, ristocetin, teicoplanin), and its helical twist (which increases in the same order), as well as the influence of various substituents and the sugar moieties [117]. In general, the broad applicability of the Chirobiotic CSP family includes chiral acids, bases, as well as amphoteric and neutral compounds—with the Chirobiotic T and V showing superior performance over Chirobiotic R [121]. Chirobiotic V especially shows enantiorecognition potential for neutral solutes, amides, esters, and amines (including aminoalcohols and cyclic amines). Chirobiotic T can separate amino alcohols, underivatized and N-derivatized amino acids, and (di)peptides [116]; Chirobiotic R can be used for chiral acids such as hydroxy acids, substituted aliphatic acids, and other acids [117]. In this connection the enantiomer-separation ability of the teicoplanin CSP for underivatized natural and synthetic amino acids deserves particular attention [125]. With the teicoplanin CSP, α-values between 1.2 and 2.7 were reported for native amino acids, with baseline resolutions (R s > 1.5) for all but His (R s = 0.8). Amino-acid enantiomer separations on teicoplanin CSP are best carried out with plain ethanol-water (or methanol-water) mobile phases, but for acidic or basic amino acids a 0.1% triethylammonium buffer should be included in the mobile phase. An important study addressed the role of the carbohydrate moieties on the chiral-recognition capability of the teicoplanin-based CSP [114]. For a chro- matographic comparison with the native teicoplanin-based CSP, a corresponding teicoplanin aglycone (TAG) analogue in which all the sugar moieties were chemically cleaved off was prepared and bonded to silica. While the overall retention of the 14.6 CHIRAL STATIONARY PHASES AND THEIR CHARACTERISTICS 705 HO CH 2 HO C H NH 2 COOH (a) (b) A g l y cone CSP Teicoplanin CSP 0 4 8 12 (min) 0 4 8 12 16 20 24 (min) Figure 14.23 Enantiomer separation of D, L-DOPA on native teicoplanin CSP (a)and teicoplanin CSP after removal of sugars (aglycone) (b). Conditions: mobile phase, 60:40 methanol/triethylammonium acetate, pH-4.1; UV detection, 254 nm; temperature, 22 ◦ C; flow rate, 1 mL/min. Reprinted with permission from [114]. TAG CSP was quite similar as compared to the native teicoplanin CSP, the enan- tiorecognition profiles were significantly altered for most of the test compounds. Notably the sugar units considerably reduced enantioselectivity for amino acids (Fig. 14.23), which clearly indicates that the active chiral distinction site is located in the aglycone part of the CSP. For other solutes the opposite trend may be observed. Chirobiotic phases are increasingly popular for the enantiomeric analysis of drugs and other xenobiotics in biofluids because of (1) their compatibility with these aqueous sample matrices and (2) their ideal ESI-MS compatibility in both reversed-phase and polar-organic modes [104, 123]. Drawbacks include the lack of CSPs that allow for a predictable reversal of elution order—when needed for the analysis of a trace-level enantiomer. Despite the claimed preparative applicability of these phases, their sample-loading capacity is limited, for the same reason as outlined for protein phases: a relatively low concentration of binding sites as compared to polysaccharide- and low-molecular-weight brush-type CSPs [61]. On the other hand, promising enantiomer separations of glycopeptide-type CSPs under SFC conditions have been demonstrated [126]. Comprehensive screens in this mode revealed that of a set of more than 100 chiral test solutes (including heterocycles, profens, β-blockers, sulfoxides, N-blocked amino acids and natural amino acids), more than 90% of these enantiomers could be resolved on commercial glycopeptide-type CSPs employing sub-/super-critical carbon dioxide–methanol as mobile phase, with low amounts of acidic and basic additives. . found to be powerful chiral selectors with complementary enantioselectivity (different enantioselectivities and elution orders), leading to CSPs based on vancomycin [111], teicoplanin [112], ristocetin. antibiotic columns, exemplified by enantiomer separations of Z-Ala (top) and 5-methyl-5-phenylhydantoin (bottom). Experimental conditions: Ristocetin A, 20:80 methanol/0.1% triethylammonium acetate buffer,. commonly understood for the RP mode using aqueous or hydro-organic mobile phases (acetonitrile-buffer or methanol-buffer mixtures) that lipophilic solutes interact with CD selectors by inclusion