Comparative analysis of enantioselective separation of novel PPAR agonists by HPLC on cellulose-and amylose-based chiral stationary phases

Chiral chromatographic resolution of a series of dual PPAR α / γ agonists was performed on the commercially available Chiralcel OD and Chiralpak AD columns in normal phase mode using a mobile phase consisting of a mixture of n -hexane, 2-propanol and trifluoroacetic acid at constant volume ratio. The selected CSPs resolved most compounds as underivatized acids without requiring time consuming analysis. Both electronic features and steric hindrance affected chiral interactions between the analytes and the CSPs. Additional information on the chiral recognition mechanism was deduced from the chromatographic behaviour of some selected methyl esters.


Introduction
Chiral chromatographic techniques have been used for the analytical and preparative separation of enantiomers for many years.Enantioselective HPLC, based on the formation of transient diastereomeric complexes between a chiral selector and a chiral selectand, is today the most widely used methodology for the separation of optical isomers both in academia and industry. 1In particular, this technique is largely employed for chiral drugs whose resolution is an issue of the greatest importance due to the strict relationship often existing between absolute configuration and pharmacological and toxicological effects.
Our interest for stereochemistry and biological activity started more than two decades ago and recently focused on the synthesis and pharmacological evaluation of new PPAR agonists.Peroxisome Proliferator-Activated Receptors (PPARs) belong to the nuclear receptors superfamily, 2,3 and they are transcription factors activated by specific ligands, which are usually lipophilic small molecules. 4Binding of these ligands results in conformational changes of the receptors that facilitate their interaction with coactivator proteins in the nucleus. 3,5,6The resulting protein complexes activate the transcription of specific target genes, resulting in the induction of intracellular signalling cascades that mediate the physiological effects of the ligands. 7,8So far, three PPAR subtypes have been described in mammals: PPARα, PPARγ, and PPARδ.All of these are targets for treatment of the metabolic syndrome, a cluster of risk factors for cardiovascular disease and diabetes including obesity, atherogenic dyslipidemia, hypertension, insulin resistance, and elevated fasting blood glucose. 3Because of the well-documented therapeutic actions of their synthetic agonists, PPARs have been the focus of intense academic and pharmaceutical research since their discovery in the early 1990s.
][11] Very recently, we reported the effects on PPARs of a series of 2-aryloxy-3-aryl-propanoic acids which can be considered chiral analogs of clofibric acid, the active metabolite of the hypolipidemic drug clofibrate.Some of these derivatives showed an agonist activity strictly related to the absolute configuration, 12 which prompted us to carefully investigate their optical purity.4][15][16] The study of the chromatographic behavior, however, was extended to all compounds of the series; due to our continuous research on the synthesis of new chiral analogs of clofibric acid as PPAR agonists, [17][18] we were interested, in fact, in a systematic investigation of the chromatographic enantioseparation of this class of drugs in relation with their structure and substitution pattern with the aim to gain useful information for method development and optimization.
In this study, therefore, the direct HPLC enantiomeric separation of a series of PPAR ligands (Figure 1) has been comparatively evaluated using the above-mentioned cellulose-and amylose-based chiral stationary phases and a mobile phase consisting of a mixture of n-hexane, 2-propanol and trifluoroacetic acid at constant volume ratio.The use of trifluoroacetic acid as an additive into the mobile phase allows to perform the chiral resolution of numerous acid analytes 19 without any derivatization which represents, on the contrary, a prerequisite for other CSPs.

Results and Discussion
Racemates 1-21, 23-25 and 27 as well as the enantiomers of compounds 1, 21, 22, 26 and 28 were synthesized according to previously reported procedures. 12,17For the analytes 22, 26 and 28, non-equimolecular mixtures of the corresponding optical isomers were injected in place of racemates.The R-enantiomers of 2, 4, 6, 7 and 8 were prepared as reported for racemates 12 starting, in this case, from commercially available (S)-phenyllactic acid (scheme 1).The Mitsunobu condensation of its ethyl ester with the suitable 4-chlorophenols afforded the Rphenoxyesters 2b, 4b and 6b-8b.The hydrolysis of these intermediates, except for 6b, afforded the desired R-acids 2, 4, 7 and 8.Alternatively, the bromo-derivative 6b was condensed with phenylboronic acid in Suzuki conditions to give the corresponding intermediate 6c whose hydrolysis afforded acid R-6.R-19 was obtained by fractional crystallization from ethyl acetate of the diastereomeric salts prepared from the corresponding racemate with (S)-1phenylethylamine (see Experimental section).The absolute configuration of the stereoisomers of 1, 22, 26 and 28 was previously assigned on the basis of the known stereochemical course of their synthetic pathways. 12,17Similarly, the absolute configuration of the R-enantiomers of 2, 4, 6, 7 and 8 was unequivocally established by Mitsunobu reaction which is known to occur with inversion of configuration. 20Only for the stereoisomers of the analytes 19 and 21 the absolute configuration was attributed by circular dichroism analysis.The CD curves of (-)-19 and (-)-21, in fact, show a negative Cotton effect around 220-250 nm as well as the R-isomers of 1 and 28 with known absolute configuration.The high structural similarity of these compounds, differing only for a moiety of the molecule which is located far away from the stereogenic center, allows to argue that, with reasonable confidence, the levo-rotatory isomers of 19 and 21 have the R-configuration too.Table 1 provides chromatographic data for the normal-phase separation of the enantiomers of compounds 1-8 on cellulose-and amylose-derived CSPs.In this series, we used derivative 1 as a reference compound and analyzed the chromatographic behavior resulting from the introduction of additional substituents in position 2 (ortho) or 2,6 (di-ortho) of the phenoxylic nucleus.
Five out of eight compounds were baseline resolved on OD column, whereas the polyhalogenated derivatives 2, 3 and 7 were poorly or not resolved.The 2-phenyl-substituted derivative 6 displayed the highest separation factor (see Figure 2); the selectivity decreased for the alkyl-substituted compounds 4, 5 and 8 but still remained higher than reference 1.

ISSN 1551-7012
Page 269 © ARKAT USA, Inc.Using the amylose-derived AD CSP, all compounds displayed a marked increment of the retention time with concomitant substantial improvement in the separation only for the polyhalogenated derivatives 2 and 7 (see Figure 3).A significant decrease of selectivity was observed for all of the other acids, whereas the 2bromo-derivative 3 still remained unresolved.Interestingly, the enantiomer elution order inverted only for the analyte 6; for all of the other analytes whose enantiomers were available, it remained unchanged differently from what often observed switching from OD to AD CSP. 14 From a general overview of the data here reported, it appears evident the high recognition ability of both the cellulose-and amylose-based CSPs towards the investigated analytes; except for the 2-bromo-derivative 3, in fact, all of them were well enantiodiscriminated at least on one of the two examined columns.These polysaccharide CSPs are semisynthetic polymers containing chains of derivatized D-(+)-glucose residues in β-1,4-linkage in cellulose and in α-1,4-linkage in amylose.These chains lie side-by-side in a helical fashion.It is well known that the binding of the analytes to these CSPs is steered by the difference in volume between the helical groove of the cellulose derivative (OD) and that of the amylose derivative (AD).The former appears to be more linear and rigid in nature, whereas the latter gives a more compact and wider helix with well-defined grooves. 21,22In both cases the polar carbamate groups are oriented inward the groove and the hydrophobic aromatic groups outward it.The former allow H-bonding and the latter π−π interactions with the analytes.These bonding types, together with dipoledipole interactions and steric effects, determine the stereogenically different fit of enantiomers into the chiral cavities. 14,22he experimental data obtained with the AD column shows longer retention times for all the compounds of this set compared to OD.This could be explained suggesting an easier accessibility of the helical cavity of the amylose-derived CSP.However, only for compounds 2 and 7 the higher capacity factors result in better selectivity and resolution probably due to a supplementary π−π interaction of their electron-poor aryloxylic rings with the phenyl rings of the tris(3,5-dimethylphenylcarbamate) of amylose.The lower electron-withdrawing effect as well as the larger size of bromine compared to chlorine could explain, instead, the lack of stereoselectivity towards the bromo-derivative 3.
The influence of a possible H-bond interaction between the phenoxylic oxygen atom of these compounds and the carbamoyl NH group of both OD and AD CSPs was also evaluated.For this purpose, the sulfur analog 1a 18 (Figure 4) was prepared and analyzed; the introduction of a sulfur in place of the ether oxygen atom produced a concomitant decrease of capacity, selectivity and resolution factors on OD column (k 1 = 2.24, α = 1.11,Rs = 2.15) compared to the isosteric analyte 1.The influence of this substitution appears less important on AD CSP; the selectivity factor towards the sulfur analog 1a, in fact, remained unchanged compared to OD in spite of the higher capacity factor (k 1 = 4.59, α = 1.11,Rs = 1.15).

Figure 4
Table 2 provides chromatographic data for the normal-phase separation of the enantiomers of compounds 9-28 on cellulose-and amylose-derived CSPs.
In this series, we analyzed the chromatographic behavior resulting from the introduction of additional substituents in different positions of the benzylic aromatic ring of 1.All orthosubstituted compounds, except for the trifluoromethyl derivative 25, were baseline resolved on OD column (as an example, see Figure 5) as well as the meta-phenoxyl compound 20.All parasubstituted analytes were poorly or not resolved with the exception of the para-phenoxylderivative 21.In any case, the alkyl-substituted analytes mostly displayed the highest separation factors.Using the amylose-derived AD CSP, all compounds displayed a marked increment of the retention time; 13 out of 20 analytes showed a concomitant substantial improvement in the enantioseparation.A significant decrease of selectivity was observed, instead, for the best resolved acids on OD column 13, 15, 17, 24 and 27 as well as for the 3-phenoxyl-derivative 20, whereas the separation factor of the 2-phenoxyl-derivative 19 remained unchanged.Only the 3phenoxyl-and 2-phenyl-substituted analytes 20 and 27 were not resolved.Interestingly, the AD CSP showed a reverse chromatographic behavior compared to OD.The para-substituted acids, in fact, turned out to be the best resolved analytes with the 4-thienyl-derivative 22 showing, by far, the highest selectivity factor.
No reverse elution order was observed for the only three analytes 19, 21 and 22 whose enantiomers were available and resolved on both cellulose-and amylose-derived CSPs.
The analysis of the chromatographic data of 9-28 shows, once again, the highly versatile and complementary properties of both the cellulose-and amylose-based CSPs towards the investigated analytes; except for the 2-trifluoromethyl-derivative 25, in fact, all of them were baseline resolved at least on one of the two examined columns.
These analytes, again, showed higher capacity factors on AD compared to OD column suggesting, also in this case, an easier accessibility of the helical cavity of the amylose-derived CSP.The benzylic moiety as well as its substituents, therefore, would be able to realize supplementary interactions with AD CSP accounting for the better enantioresolution of 13 out of 20 analytes compared to OD (as an example, see Figure 6).Interestingly, the AD column exhibited a very efficient enantiodiscrimination process towards the para-thienyl derivative 22 which might be ascribed to the strong electrostatic interaction realized by the thiophene ring with the π-cloud of the aromatic ring of the tris(3,5-dimethylphenylcarbamate) of amylose.In particular, sulfur-arene interactions are known to be strongly attractive.The only analytes displaying lower enantioselectivity factors on AD column compared to OD were the ortho-alkyl-substituted derivatives 13, 15, 17 and 24, and compounds 20 and 27 bearing the meta-phenoxyl and the ortho-phenyl substituents, respectively.These substituents could force the benzylic aromatic ring to assume a conformation not particularly favorable for interactions with the amylose-CSP.
With the aim to gain insight into the factors influencing the chiral recognition of the acids reported in this work, we analyzed the chromatographic behavior of the methyl esters 29-37, obtained by reaction of the corresponding acids with a solution of diazomethane in Et 2 O, on both the polysaccharide CSPs.These derivatives were chosen because of the different steric and electronic properties of the substituents present on both aromatic rings.Moreover, esters 32, 33 and 35-37 derived from acids which were not resolved on OD column (10, 12, 25, 26 and 28), whereas ester 29 derived from the acid 3 which was unresolved on both OD-and AD-CSPs.The experimental data reported in Table 3 shows that it is possible to perform the resolution of all methyl esters, with the exception of 35, at least on one of the two investigated columns.All the analytes were less retained than the corresponding acids with the exception of 31 and 37 on OD.This implicates that the formation of a H-bond between the carboxylic group of acid analytes and CSPs is an important feature; other interactions, however, play a crucial role in the recognition mechanism as confirmed from the better separation of some esters compared to their corresponding acids (see esters 29, 30 and 37 on OD, ester 30 on AD).One of these interactions could be realized from the methyl group of these esters with the π-cloud of the CSP aromatic rings.
To summarize, the investigated CSPs were very efficient in accomplishing chiral resolution of both acids and esters without requiring time consuming analysis.This holds true especially if one considers that no attempts were made to optimize the enantioseparation conditions.They could, therefore, be used to provide the enantiomers of many of the examined compounds even on a semi-preparative scale.As an example, the chromatographic traces on Chiralpak AD relating to the racemates and S-isomers of the potent PPAR agonists 21 and 26 are reported in Figure 7.In both cases, the high resolution allows to distinguish the presence of the R-isomer (distomer) as an impurity (lower than 1-2 %) of the biologically active S-isomer (eutomer).occur in methyl esters accounting for the better enantioresolution of some of these derivatives compared to the corresponding acids.

Experimental Section
General Mobile phase elution was made isocratically at flow 1.0 mL/min using n-hexane/2-propanol 95/5 (% v/v) or n-hexane/2-propanol/TFA 95/5/0.02for methyl esters and acids, respectively.A conditioning time of 120 minutes was applied whenever the eluent composition was changed to obtain a good reproducibility.Injected solutions (filtered through a 0.45 µm membrane filter) were obtained dissolving 2-3 mg of analytes in 2 mL of 2-propanol.The retention factor (k) was calculated using the equation k = (t r /t o )−1, where t r is the retention time of the analyte and t o is the retention time of an unretained compound.In this study t o was determined as the first disturbance of the baseline after injection.The separation factor (α) was calculated using the equation, α = k 2 /k 1 where k 1 and k 2 are the retention factors for the first and last eluted enantiomers, respectively.Resolution is calculated from the equation Rs = 2(t 2 − t 1 )/(w 1 + w 2 ), where t 1 and t 2 are the retention times of the first-and second-eluted peaks, respectively, and w 1 and w 2 are the peak widths.The determination of the enantiomeric elution order was carried out by injecting solutions of pure enantiomers under the same experimental conditions of the racemates.
Racemates 1-21, 23-25 and 27 as well as the enantiomers of compounds 1, 21, 22, 26 and 28 were synthesized according to previously reported procedures. 12,17The enantiomeric excesses of the optically active acids were >90% as determined by HPLC analysis of their methyl esters, obtained by reaction with a solution of diazomethane in Et The R-enantiomers of 2, 4, 6, 7 and 8 were prepared as reported for racemates 12 starting, in this case, from commercially available (S)-phenyllactic acid.The physico-chemical properties of these isomers were identical to those of the corresponding racemates.Chemical yields were 20-75%.The enantiomeric excesses were determined by HPLC analysis on Chiralpak AD column using the experimental conditions employed in this study (n-hexane/2-propanol/TFA 95:5:0.02as the mobile phase, flow rate 1.0 mL/min, detection 280 nm).All of the optically active acids had enantiomeric excesses > 95% except for 6 (e.e. 65%).R-19 was obtained by resolution of the corresponding racemate as follows.1.10 g of R,S-19 (2.9 mmol) and 0.39 g of (S)-1-phenylethylamine (3.2 mmol) were dissolved in 20 mL of absolute ethanol and the mixture was stirred overnight at room temperature.The solvent was evaporated under vacuo and the solid residue was crystallized from ethyl acetate.After three crystallizations, the diastereomeric salt was treated with 2N HCl and then extracted with Et 2 O.The organic layer was washed with brine, dried over Na 2 SO 4 and evaporated to dryness affording the final acid which was crystallized from CHCl 3 /n-hexane.Yield: 15%.The physico-chemical properties of this isomer were identical to those of the corresponding racemate. 12[α] D = -29 (c 0.1, MeOH); e.e. 96% (HPLC on Chiralcel OD; n-hexane/2-propanol/TFA 95:5:0.02as the mobile phase, flow rate 1.0 mL/min, detection 280 nm).

Figure 1 .
Figure 1.General structure of the investigated PPAR agonists.

Figure 2 .
Figure 2. Chromatographic trace on Chiralcel OD of racemate 6. Experimental conditions are reported in the text.

Figure 3 .
Figure 3. Chromatographic trace on Chiralpak AD of racemate 7. Experimental conditions are reported in the text.

Figure 5 .
Figure 5. Chromatographic trace on Chiralcel OD of racemate 13.Experimental conditions are reported in the text.

Figure 6 .
Figure 6.Chromatographic trace on Chiralpak AD of racemate 18. Experimental conditions are reported in the text.

Figure 7 .
Figure 7. Chromatographic traces on Chiralpak AD of the racemates 21 and 26 and their Sisomers.Experimental conditions are reported in the text.

.
Chromatography was performed on a Perkin-Elmer chromatograph, equipped with a Rheodyne 7725i model injector, a 785A model UV/Vis detector, a series 200 model pump, and NCI 900 model interface.The analyses were carried out at room temperature, with detection at 280 nm.The chromatographic data were collected on a computer running Perkin-Elmer TotalChrom version 6.3.1.0504software.Chiralcel OD and Chiralpak AD columns (4.6 mm i.d.x 250 mm) were supplied by Daicel Chemical Industries, Ltd, Tokyo, Japan; HPLC grade n- hexane and 2-propanol were purchased from Carlo Erba (Milan, Italy); trifluoroacetic acid (TFA, 99+%, spectrophotometric grade) was obtained from Sigma Aldrich (Milan, Italy).