Oo

Aa

Ab

a

ARRA

KFEMPS

1

aamkmialedc

fooctb

1d

Separation and Purification Technology 68 (2009) 9–23

Contents lists available at ScienceDirect

Separation and Purification Technology

journa l homepage: www.e lsev ier .com/ locate /seppur

ptimization of the mobile phase composition for preparative chiral separationf flurbiprofen enantiomers

ntónio E. Ribeiroa, Nuno S. Gracaa, Luís S. Paisa,∗, Alírio E. Rodriguesb

Laboratory of Separation and Reaction Engineering, School of Technology and Management, Braganca Polytechnic Institute, Campus de Santa Apolónia,partado 1134, 5301-857 Braganca, PortugalLaboratory of Separation and Reaction Engineering, Faculty of Engineering, University of Porto, Rua Dr. Roberto Frias s/n, 4200-465 Porto, Portugal

r t i c l e i n f o

rticle history:eceived 14 August 2008eceived in revised form 27 March 2009ccepted 31 March 2009

eywords:lurbiprofen

a b s t r a c t

This work presents the experimental and simulation results obtained for the optimization of the mobilephase composition for the preparative separation of flurbiprofen enantiomers by liquid chromatogra-phy using an amylose-based chiral stationary phase (Chiralpak AD). The experimental work carried outincludes solubility and adsorption isotherm measurements and pulse and breakthrough experimentsunder preparative conditions. The simulation work predicts the operation of a simulated moving bed(SMB) system for the separation of flurbiprofen enantiomers to compare the productivity and solventconsumption performances, for the different mobile phase compositions and using the experimental data

nantiomer separationobile phase composition

reparative chromatographyimulated moving bed

obtained. This paper presents a new and different case study (flurbiprofen) of the one recently reportedby the authors (ketoprofen enantiomers [A. Ribeiro, N. Graca, L. Pais, A. Rodrigues, Preparative separa-tion of ketoprofen enantiomers: choice of mobile phase composition and measurement of competitiveadsorption isotherms, Sep. Purif. Technol. 61 (2008) 375–383]), to clearly show that the optimization ofthe mobile phase composition for preparative chiral separation requires an individualized study, since

ned e

different results are obtai

. Introduction

Flurbiprofen [(R,S)-2-(2-fluoro-4-biphenyl)-propionic acid], a 2-rylpropionic acid derivative (profen), is a chiral non-steroidalnti-inflammatory drug (NSAID) (Fig. 1). This drug has a phar-acological action similar to other drugs of this class, such as,

etoprofen, ibuprofen or naproxen. This chiral drug is still beingarketed worldwide as a racemic mixture, although the increas-

ng number of published studies referring that R(−)-Flurbiprofennd S(+)-Flurbiprofen pure enantiomers have distinct pharmaco-ogical activities. In this way, the chiral resolution of flurbiprofennantiomers can promote the development of two new therapeuticrugs which have distinct profiles and/or which are more pharma-ologically safe.

Currently, there is a strong interest in enantioseparation of pro-ens, mainly the flurbiprofen enantiomers. This interest is basedn the fact that his R enantiomer has been referred as a promoter

f efficient inhibition on the development of varied forms of can-er in human beings [2], such as, the prostate cancer [3–5] andhe colon cancer [6,7]. The most recent application area of flur-iprofen enantiomers has been described in numerous clinic and

∗ Corresponding author. Tel.: +351 273 303 090; fax: +351 273 313 051.E-mail address: pais@ipb.pt (L.S. Pais).

383-5866/$ – see front matter © 2009 Elsevier B.V. All rights reserved.oi:10.1016/j.seppur.2009.03.049

ven for enantiomers systems of the same family.© 2009 Elsevier B.V. All rights reserved.

pharmacological reports, in which the R enantiomer is referred asan important hypothetical drug used to minimize the progressionof Alzheimer disease [8–11].

Liquid chromatography is now the most accepted method forchiral separations, not only in the direct way, using chiral station-ary phases, but also in the indirect way, by using chiral derivatizingreagents [12]. Examples of flurbiprofen enantiomers separation,using a Chiralpak AD stationary phase as chiral selector, can befound in the literature. Patel et al. used a hexane/isopropyl alco-hol/trifluoracetic acid (90/10/0.05%, v/v) mobile phase compositionand obtained a selectivity value of 1.7 [13,14]. A similar mobilephase composition (95/5/1) was used by Booth and collabora-tors, using the same stationary phase [15]. At a preparative scale,some examples can be found on both the enzymatic resolutionof R-Flurbiprofen [16–18] and S-Flurbiprofen [19–21] enantiomers.However, as far as our knowledge, there are no published studiesrelated to the preparative separation of flurbiprofen enantiomersby liquid chromatography.

The objective of this paper is to study the effect of mobilephase composition on the preparative separation of flurbiprofen

enantiomers by chiral liquid chromatography. Experimental resultsobtained for different mobile phase compositions will be presentedand discussed, including solubility measurements, elution (pulses)and frontal (breakthroughs) chromatographic experiments, andthe measurement of the equilibrium binary adsorption isotherms.

10 A.E. Ribeiro et al. / Separation and Purific

FR

Atp

2

UawCLoiotpai

raflMagu11

anwumeasteleatte

3

3

tibga

ig. 1. Chemical structure of flurbiprofen enantiomers: (a) S(+)-Flurbiprofen and (b)(−)-Flurbiprofen.

fterwards, the comparison of the performance of the prepara-ive separation of flurbiprofen enantiomers is carried out, using theredictions for fixed bed and simulated moving bed operation.

. Experimental

All the analysis were performed on a Jasco HPLC system with anV-1575 multiwavelength detector set at 260 nm, equipped withpreparative cell (1.0 mm). Two chiral chromatographic columnsere used with the same adsorbent material (Chiralpak AD, Daicelhemical Industries Ltd., Japan) and the same dimensions (250 mm× 4.6 mm ID). These two columns have different particle size:ne column, with a particle size of 10 �m, was used for analyt-cal purposes (measurement of enantiomers concentrations); thether, with a particle size of 20 �m, was used in the prepara-ive chromatographic experiments (adsorption–desorption steps,ulses and breakthrough experiments). It must be pointed out thatparticle size of 20 �m is normally used for preparative separations,

ncluding SMB operation.The measurements of the binary adsorption isotherms were car-

ied out at 23 and 35 ◦C, using an Eldex CH-150 column oven andthermostatic water bath for solvents. Analytical grade racemic

urbiprofen was purchased from Merck (Darmstadt, Germany).ethanol, ethanol and n-hexane (Fluka, Buchs, Switzerland) were

ll HPLC grade. Trifluoracetic acid (TFA) was spectrophotometricrade. If nothing is said in contrary, all mobile phase compositionssed in this work include 0.01% of the TFA modifier. For example,000 mL of a 10% ethanol/90% n-hexane mixture is prepared adding00 mL ethanol, 900 mL n-hexane and 100 �L of TFA.

The gravimetric method for solubility measurements and thedsorption–desorption method used in the experimental determi-ation of binary adsorption isotherms are described in the previousork [1]. In the adsorption–desorption method, the preparative col-mn is saturated with a large amount of feed solution (racemicixture), with known concentration of both enantiomers, until

quilibrium is achieved. The column is then completely regener-ted with eluent. The eluted volume, resulting from this desorptiontep, is collected and analyzed, in order to measure each enan-iomer concentration. A mass balance will allow the evaluation ofach enantiomer concentration retained in the particle, in equi-ibrium with its known concentration in the feed solution. Thentire adsorption isotherm measurement will require a set ofdsorption–desorption experiments, using different feed concen-rations. The concentration of each flurbiprofen enantiomer inhe feed (racemic) and eluted solutions was evaluated by HPLC,quipped with the analytical column described before.

. Modeling, correlation and simulation

.1. Binary adsorption equilibrium data

After experimental determination, the fitting of the adsorp-

ion measurements to an adsorption isotherm model is advisedn order to allow the simulation and prediction of the adsorptionehavior and a better understanding of the overall chromato-raphic separation process. In this study, it is presented the binarydsorption equilibrium data of Chiralpak AD/enantiomer mixtures

ation Technology 68 (2009) 9–23

(R(−)Flurbiprofen and S(+)Flurbiprofen). The adsorption equilib-rium measurements yield data of adsorbed phase concentrations ofR(−)Flurbiprofen and S(+)Flurbiprofen as functions of their solutionphase concentrations. These data were fitted with three relativelysimple binary isotherm equations, Langmuir and its modified ver-sions (see Table 1 for the expressions). The isotherm parametersobtained are listed in Table 1 as well. These isotherm model param-eters can be estimated using a Levenberg–Marquardt algorithm forthe minimization of the sum of squares of the residues, SQ (seeTable 1, Eq. (7)) or, in order to compare models with a differentnumber of parameters, the standard deviation, SD (see Table 1, Eq.(8)).

The system selectivity and its dependence on both enantiomerconcentrations can be evaluated by ˛ = (q∗

2/C2)/(q∗1/C1), where q∗

iis

the concentration of enantiomer i retained in the particle, in equi-librium with its concentration in the liquid phase, Ci. The subscripti = 1, 2, represents, respectively, the less (R(−)-Flurbiprofen) and themore (S(+)-Flurbiprofen) retained enantiomer.

3.2. Simulation of fixed bed and SMB operation and performance

The simulation of fixed bed operation under non-linear prepar-ative conditions can be carried out by using a linear driving forcemodel. Model equations include the mass balance equations, theequilibrium isotherms models, the initial and the boundary condi-tions, and can be found in the previous work [1].

The simulation of SMB operation and performance was carriedout by using the findings published by Morbidelli and co-workers,who developed a complete design of the binary countercurrentseparation processes by SMB chromatography in the frame of equi-librium theory, assuming that mass transfer resistances and axialdispersion are negligible, and that the adsorption equilibria canbe described through a variable selectivity modified Langmuirisotherm [22]. The SMB performance can be evaluated by defin-ing the complete separation regions and through the performanceparameters of productivity and solvent consumption. A separa-tion region is the area of possible SMB internal flow rates thatallows 100% pure products (pure extract, only containing the moreretained enantiomer; and pure raffinate, only containing the lessretained enantiomer). The performance parameters are evaluatedat the vertex of each separation region, since it represents the bestoperating conditions in terms of system productivity and solventconsumption for a given feed concentration:

Pr = ε

Nct∗ (�3 − �2)(CF1 + CF

2 ) (9)

� = PrNct∗

ε= (�3 − �2)(CF

1 + CF2 ) (10)

SC = 1CF

1 + CF2

(1 + �1 − �4

�3 − �2

)(11)

where ε is the bed porosity, Nc the total number of columns in theSMB unit, t* the switch time interval, � j = vj/us the ratio betweenfluid and solid interstitial velocities in section j of the equivalenttrue moving bed operation, and CF

ithe feed concentration of enan-

tiomer i.Although it represents a simplified approach (no axial disper-

sion or mass transfer resistances and equivalence to the ideal truemoving bed operation), the equilibrium theory model allows astraightforward prediction of SMB performance and is very use-

ful for comparative studies as the one carried out in this work(to compare the performances obtained for different mobile phasecompositions). For more information concerning SMB modelingand simulation, through the equilibrium theory and other moreprecise SMB models, see Refs. [22–29].

A.E. Ribeiro et al. / Separation and Purification Technology 68 (2009) 9–23 11

Table 1Binary adsorption isotherm models and criteria parameters used to fitting experimental data.

Model Enantiomer 1: R(-)-Flurbiprofen Enantiomer 2: S(+)-Flurbiprofen

Langmuir, LG3 q∗1 = Qb1C1

1 + b1C1 + b2C2(1) q∗

2 = Qb2C2

1 + b1C1 + b2C2(2)

Linear + Langmuir, LLG4 q∗1 = mC1 + Qb1C1

1 + b1C1 + b2C2(3) q∗

2 = mC2 + Qb2C2

1 + b1C1 + b2C2(4)

Modified linear + Langmuir, LLG5 q∗1 = m1C1 + Qb1C1

1 + b1C1 + b2C2(5) q∗

2 = m2C2 + Qb2C2

1 + b1C1 + b2C2(6)

Criteria parameters

Sum of square of the residues, SQ SQ =M/2∑j=1

[(q∗T1j

− q∗E1j

)2 + (q∗T

2j− q∗E

2j)2](7)

S

√1

M , m1, ma s, res

4

4

badofliwt

hTpur

TSd(

S

S

FuTo

tandard deviation, SD SD =M − N

SQ

is the number of experimental points; N the number of estimated parameters; mnd q∗T are the experimental and model equilibrium stationary phase concentration

. Results and discussion

.1. Solubility measurements

At first, the evaluation of the flurbiprofen enantiomers solu-ility was carried out in three pure solvents (n-hexane, ethanolnd methanol). These solubility measurements were performed inuplicate, and at two different temperatures (23 and 35 ◦C). Thebtained results, presented in Table 2, indicate, as expected, thaturbiprofen enantiomers have increasing solubility with the ris-

ng of temperature. It can be also observed that solubility increaseshen the solvent is changed from n-hexane to ethanol and from

his to methanol.On a second stage, the dependency of solubility on the alco-

olic content of an ethanol/n-hexane mixture was investigated.

he results shown in Fig. 2 are consistent with the ones presentedreviously in Table 2. The flurbiprofen enantiomers have high sol-bility values in solvents with a high polar composition. Theseesults are similar to the ones previously obtained for ketoprofen

able 2olubility (S) of racemic flurbiprofen in three pure solvent compositions and twoifferent temperatures, expressed in mass (g) of solute per unit mass (kg) of solventon a solute-free basis).

olvent composition 100% n-hexane 100% ethanol 100% methanol

olubility (S) T = 23 ◦C 7.2 538.7 685.1T = 35 ◦C 9.7 795.5 991.8

ig. 2. Effect of the alcoholic content (ethanol/n-hexane-based solvent) in the sol-bility (S) of racemic flurbiprofen at two different temperatures (closed circles for= 35 ◦C; open circles for T = 23 ◦C) expressed in mass (g) of solute per unit mass (kg)f solvent (on a solute-free basis).

(8)

2, Q, b1 and b2 are the model parameters, C is the feed solution concentration, q∗E

pectively.

[1] and ibuprofen [30]. Additionally, it must be pointed out that amobile phase with a high polar content is referred as an advantagein preparative chromatography, because allows higher racematesolubility but also because it presents lower retention times [31].The solubility measurements of flurbiprofen enantiomers on amethanol/n-hexane-based solvent were not carried out due to itsconsiderable immiscibility range (between 6% and 60% of methanolin n-hexane at 23 ◦C) [32,33].

4.2. Elution chromatography

Several experiments of elution chromatography (pulses) wereperformed on different ethanol/n-hexane and methanol/n-hexanemixtures, in order to characterize the system selectivity. Six levelconcentration samples (between 0.05 and 4.0 g/L) were preparedand injected using two different volume loops (100 �L and 1 mL).The obtained results are presented in Fig. 3 for the ethanol/n-hexane mixtures and in Fig. 4 for the methanol/n-hexane system.High selectivity values can be observed for a composition of 5%ethanol/95% n-hexane. However, this mobile phase compositionexhibits very high retention times and very low solubility values,which represents a clear disadvantage under preparative operation.For an ethanol/n-hexane-based mobile phase, a 10/90 compositionrepresents, as it will be stressed out later, a reasonable compromisebetween selectivity, retention time and solubility.

For methanol/n-hexane-based mobile phases, due to the immis-cibility referred above, flurbiprofen enantioseparation is notpossible in the range between 6 and 60% of methanol. Due tothe very low solubility values, experiments with less than 6% ofmethanol were also not carried out, since they are not attractive forpreparative separation. In spite of high solubility, the use of a mobilephase with a high methanol content presents low selectivity, as itcan be observed in Fig. 4.

4.3. Binary adsorption isotherm measurements and modeling

The binary adsorption isotherms for flurbiprofen enantiomersin Chiralpak AD stationary phase were determined experimentallyat 23 ◦C for four different ethanol/n-hexane mobile phase compo-sitions (10/90, 20/80, 40/60 and 100/0) and for pure methanol. Theexperimental results were fitted using the models presented inTable 1. The correspondent model parameters, obtained with the

Levenberg–Marquardt algorithm, are shown in Table 3.

Figs. 5 and 6 present the experimental and model predictions,showing a very good agreement for the modified linear + Langmuirmodel (LLG5), although good agreements are also obtained for thesimpler Langmuir (LG3) and linear + Langmuir (LLG4) models. The

12 A.E. Ribeiro et al. / Separation and Purification Technology 68 (2009) 9–23

Fig. 3. Experimental elution profiles of flurbiprofen enantiomers in six different ethanol/n-hexane mobile phase compositions: 5/95, 10/90, 20/80, 40/60, 70/30 and 100/0.Racemic flurbiprofen concentrations in six different levels: 0.05, 0.2, 0.5, 1.0, 2.0 and 4.0 g/L; preparative column (particle diameter of 20 �m); UV detection at 260 nm; flowrate of 1 mL/min; temperature of 23 ◦C; injection volumes of 100 �L and 1 mL.

A.E. Ribeiro et al. / Separation and Purification Technology 68 (2009) 9–23 13

Fig. 3. (Continued ).

Fig. 4. Experimental elution profiles of flurbiprofen enantiomers in two different methanol/n-hexane mobile phase compositions: 70/30 and 100/0. Racemic flurbiprofenconcentrations in six different levels: 0.05, 0.2, 0.5, 1.0, 2.0 and 4.0 g/L; preparative column (particle diameter of 20 �m); UV detection at 260 nm; flow rate of 1 mL/min;temperature of 23 ◦C; injection volumes of 100 �L and 1 mL.

14 A.E. Ribeiro et al. / Separation and Purification Technology 68 (2009) 9–23

Table 3Estimated model parameters for flurbiprofen adsorption isotherms for different mobile phase compositions at 23 ◦C.

Model M N m1 m2 Q b1 b2 SQ SD

10% ethanol/90% n-hexaneLG3 22 3 – – 216.6 9.461 × 10−3 1.291 × 10−2 1.7393 0.3026LLG4 4 0.6117 100.2 1.429 × 10−2 2.295 × 10−2 1.3314 0.2720LLG5 5 0.3296 8.038 × 10−5 183.1 9.003 × 10−3 1.564 × 10-2 1.1245 0.2572

20% ethanol/80% n-hexaneLG3 22 3 – – 176.2 7.584 × 10−3 9.846 × 10−3 0.7305 0.1961LLG4 4 0.6201 45.53 1.569 × 10−2 2.701 × 10−2 0.3767 0.1447LLG5 5 0.7178 0.8341 25.91 2.603 × 10−2 4.130 × 10−2 0.3014 0.1331

40% ethanol/60% n-hexaneLG3 22 3 – – 241.1 4.132 × 10−3 5.069 × 10−3 0.7268 0.1956LLG4 4 0.1747 164.3 4.992 × 10−3 6.407 × 10−3 0.7252 0.2007LLG5 5 0.2630 2.935 × 10−5 183.7 3.863 × 10−3 6.782 × 10−3 0.6484 0.1953

100% ethanolLG3 22 3 – – 287.3 2.906 × 10−3 3.283 × 10−3 0.4948 0.1614LLG4 4 0.6284 15.75 1.307 × 10−2 2.271 × 10−2 0.4411 0.1565LLG5 5 0.6826 0.7460 4.895 3.876 × 10−2 5.613 × 10−2 0.4033 0.1540

1

5

ro

cTei

Fae

00% methanolLG3 22 3 – –LLG4 4 0.3531LLG5 5 0.2703 1.241 × 10−

esults also clearly show a decrease of selectivity with the increasef the alcoholic content.

As it is experimentally well known, the selectivity factor for

hiral systems decreases with the increase of feed concentration.he modified linear + Langmuir model (LLG5), as well as the lin-ar + Langmuir model (LLG4), can predict this dependency, as its shown in Fig. 7 (for ethanol/n-hexane compositions) and Fig. 8

ig. 5. Comparison between model (modified linear + Langmuir model, LLG5) and experimt 23 ◦C in four different ethanol/n-hexane mobile phase compositions: 10/90, 20/80, 40/6nantiomer; closed circles for experimental concentration of the S(+)-Flurbiprofen enant

159.6 6.713 × 10−3 8.214 × 10−3 0.5705 0.173367.38 1.066 × 10−2 1.467 × 10−2 0.5396 0.1731120.9 6.307 × 10−3 1.118 × 10−2 0.3714 0.1478

(for pure methanol), which present the three-dimensional plotsfor the selectivity factor and its dependency on both enantiomersconcentrations. These results also predict the observed decrease in

selectivity with the increase of the alcoholic content. As an exam-ple, for a 20 g/L of each enantiomer, selectivity is 1.34, 1.28, 1.21 and1.12 for compositions of 10/90, 20/80, 40/60 and 100/0 ethanol inn-hexane, respectively; using a pure methanol mobile phase, selec-

ental results for the equilibrium adsorption isotherms of flurbiprofen enantiomers0 and 100/0. Open circles for experimental concentration of the R(−)-Flurbiprofen

iomer; solid lines for adsorption isotherm model.

A.E. Ribeiro et al. / Separation and Purification Technology 68 (2009) 9–23 15

Fig. 6. Comparison between model (modified linear + Langmuir model, LLG5) andettt

tn

4

o

Fe

xperimental results for the equilibrium adsorption isotherms of flurbiprofen enan-iomers at 23 ◦C in pure methanol. Open circles for experimental concentration ofhe R(−)-Flurbiprofen enantiomer; closed circles for experimental concentration ofhe S(+)-Flurbiprofen enantiomer; solid lines for adsorption isotherm model.

ivity is 1.20, similar to the one obtained for the 40% ethanol/60%-hexane mobile phase composition.

.4. Frontal chromatography

Different experiments of frontal chromatography were carriedut with the purpose of testing the selected binary adsorption

ig. 7. Selectivity factor three-dimensional plots and its dependency on both flurbiprofthanol/n-hexane mobile phase compositions: 10/90, 20/80, 40/60 and 100/0.

Fig. 8. Selectivity factor three-dimensional plot and its dependency on both flur-biprofen enantiomers concentrations (modified linear + Langmuir model, LLG5) forpure methanol.

isotherm models. Three ethanol/n-hexane-based mobile phasecompositions were tested (10/90, 40/60 and 100/0), using as feedsolution a 40 g/L flurbiprofen racemic mixture. Fig. 9 presents

the experimental results obtained (points) and the simulationpredictions (lines). The experimental data was obtained by recov-ering and analyzing samples at different times of the saturationand regeneration steps. The simulation predictions were carriedout by using the linear driving force model for fixed bed chro-

en enantiomers concentrations (modified linear + Langmuir model, LLG5) for four

16 A.E. Ribeiro et al. / Separation and Purification Technology 68 (2009) 9–23

Fig. 9. Saturation (adsorption) and regeneration (desorption) curves for total feed concentration of 40 g/L (racemic mixture) at 23 ◦C in three ethanol/n-hexane mobile phasecompositions: 10/90, 40/60 and 100/0. Comparison between experimental (points) and simulation (lines) results. Closed and open circles for the R(−)-Flurbiprofen andS(+)-Flurbiprofen enantiomers, respectively. Flow rate: 0.5 mL/min. Model parameters used: ε = 0.4, Pe = 3500, St = k� = 1000, and modified linear + Langmuir (LLG5) modelp

mcvcs

dflhaac

arameters (see Tables 1 and 3).

atography and the modified linear + Langmuir model (LLG5). Itan be concluded that the selected adsorption model describesery well the fixed bed chromatographic behavior in the wholeoncentration range and for all the mobile phase compositionstudied.

Additionally, frontal chromatography experiments using fiveifferent feed concentrations (1, 10, 20, 30 and 40 g/L of racemic

urbiprofen) were performed using the same 10% ethanol/90% n-exane mobile phase composition. Fig. 10 also presents a goodgreement between experimental data and simulation predictions,lthough with small discrepancies at the end of the regenerationurve for the more retained component.

4.5. SMB chromatography

The performance of flurbiprofen preparative chiral separationby SMB technology was predicted using the equilibrium theorymodel and the equilibrium adsorption data obtained experimen-tally and presented before. Fig. 11 presents the separation regionsobtained for four ethanol/n-hexane mobile phase compositions

(10/90, 20/80, 40/60 and 100/0), using the linear + Langmuir (LLG4)adsorption model. These predictions are presented for two limitsituations: one for a racemic feed concentration of 0.1 g/L (lowconcentrations, linear behavior); the other for 25 g/L (high con-centrations, non-linear behavior). Fig. 11 shows that high retention

A.E. Ribeiro et al. / Separation and Purification Technology 68 (2009) 9–23 17

Fig. 10. Saturation (adsorption) and regeneration (desorption) curves for five total feed concentrations (1, 10, 20, 30 and 40 g/L, racemic mixtures), at 23 ◦C and using a 10%ethanol/90% n-hexane mobile phase composition. Comparison between experimental (points) and simulation (lines) results. Closed and open circles for the R(−)-Flurbiprofenand S(+)-Flurbiprofen enantiomers, respectively. Flow rate: 0.5 mL/min. Model parameters used: ε = 0.4, Pe = 3500, St = k� = 1000, and modified linear + Langmuir (LLG5) modelparameters (see Tables 1 and 3).

18 A.E. Ribeiro et al. / Separation and Purification Technology 68 (2009) 9–23

Fig. 10. (Continued ).

F massc + CF

2

tmhccsddtfpehteph

ioceda�

ig. 11. SMB separation regions predicted under the equilibrium theory (negligibleompositions (10/90, 20/80, 40/60 and 100/0) and two total feed concentrations (CF

1

imes are obtained when using high hydrocarbon contents, whicheans that the separation regions are progressively displaced for

igher values of �2 and �3 (from the bottom-left to the upper-rightorner, in the �3 × �2 plane) with the increase of the n-hexaneontent (decrease of the ethanol content). Another main conclu-ion is that the dimension of the separation regions progressivelyecreases with the decrease of the n-hexane content. This pre-iction can be explained by the experimental data obtained forhe adsorption isotherms and correspondent selectivity valuesor the different mobile phase compositions. Therefore, the besterformance (bigger separation region) is obtained with a 10%thanol/90% n-hexane composition in both situations of low andigh feed concentrations. These findings are very different fromhe ones recently reported for the separation of the ketoprofennantiomers, where the best system performance is obtained forure ethanol, instead of an ethanol/n-hexane mixture with highydrocarbon content [1].

Fig. 12 stresses out the conclusions taken before, by present-ng the system productivity and solvent consumption for SMBperation as a function of feed concentration. For a given feed

oncentration, system productivity and solvent consumption werevaluated by Eqs. (10) and (11), at the vertex of the correspon-ent separation region and using the critical values for both �1nd �4, predicted by the equilibrium theory model [22]; that is,1,min, �2,vertex, �3,vertex, and �4,max, representing the best oper-

transfer resistances and axial dispersion) for four ethanol/n-hexane mobile phase): 0.1 g/L (linear range) and 25 g/L (non-linear range).

ating conditions for SMB performance. Fig. 12 clearly shows thatbetter productivity is obtained for mobile phases with a lowalcoholic content, while solvent consumption does not differ sig-nificantly.

4.6. Effect of the acidic modifier content

The addition of acidic or basic modifiers to the mobile phase isa common practice in liquid chromatography using chiral station-ary phases, since it increases selectivity and decreases retention(chromatographic tail). The trifluoracetic acid, TFA, is generally usedto decrease the retention of acidic compounds, which can be verysignificant in normal phase operation (high hydrocarbon content)[34–38].

For all experimental results presented until now in this worka mobile phase containing 0.01% TFA was used. In order to studyif this is enough, it was carried out experiments of pulse andfrontal chromatography using a higher TFA content (0.1%) withthe elected 10% ethanol/90% n-hexane mobile phase composition.It should be pointed out that the maximum allowed content is

referred to be 0.5% by the chiral stationary phase manufacturer.Figs. 13 and 14 present the results obtained and compare themto the ones obtained using 0.01% TFA. Observing these figures, itcan be concluded that the use of a higher TFA content only slightlydecrease retention time of both enantiomers and do not change the

A.E. Ribeiro et al. / Separation and Purification Technology 68 (2009) 9–23 19

Fig. 12. Prediction of productivity and solvent consumption of SMB operation using the equilibrium theory (negligible mass transfer resistances and axial dispersion) as afunction of feed concentration for four ethanol/n-hexane mobile phase compositions (solid line for 10/90, traced line for 20/80, dotted line for 40/60 and semi-dotted line for100/0).

F /90% na .5, 1.02 d 1 m

obm

4

oswgp

ig. 13. Experimental elution profiles of flurbiprofen enantiomers in a 10% ethanolnd 0.1% TFA. Racemic flurbiprofen concentrations in six different levels: 0.05, 0.2, 060 nm; flow rate of 1 mL/min; temperature of 23 ◦C; injection volume of 100 �L an

verall behavior of the chromatographic process. Therefore, it cane concluded that no better performances are obtained with a TFAodifier content higher than 0.01%.

.7. Effect of temperature

In order to predict the effect of temperature on the performance

f the flurbiprofen enantiomers preparative separation, the set ofolubility and adsorption measurements and pulse experimentsere reproduced at a higher temperature (35 ◦C). The chromato-

raphic experiments were carried out using the elected mobilehase composition (10% ethanol/90% n-hexane/0.01% TFA).

-hexane mobile phase composition and using two acidic modifier contents: 0.01%, 2.0 and 4.0 g/L; preparative column (particle diameter of 20 �m); UV detection atL.

Table 2 and Fig. 2 show that, as expected, higher solubilityvalues are obtained for higher temperatures. The pulse experi-ments presented in Fig. 15 show that, as also expected, the use ofhigher temperatures decreases retention time of both enantiomers,but also decreases selectivity. For a better understanding of theinfluence of temperature on the adsorption behavior and on thechromatographic process performance, the adsorption isotherms

were experimentally measured for the elected mobile phase com-position at 35 ◦C. The results obtained are presented in Fig. 16and compared with the ones obtained at 23 ◦C. The correspon-dent adsorption model parameters using the LG3, LLG4, and LLG5adsorption models presented earlier in this work are presented in

20 A.E. Ribeiro et al. / Separation and Purification Technology 68 (2009) 9–23

Fig. 14. Saturation (adsorption) and regeneration (desorption) experimental curves for total feed concentration of 40 g/L (racemic mixture) at 23 ◦C using a 10% ethanol/90%n-hexane mobile phase composition and two TFA contents: circles for 0.01% and squares for 0.1% TFA. Closed circles and squares for the R(−)-Flurbiprofen and open circlesand squares for the S(+)-Flurbiprofen enantiomers, respectively. Flow rate: 0.5 mL/min.

F /90% n2 .5, 1.02

Tasp

TE

M

1

ig. 15. Experimental elution profiles of flurbiprofen enantiomers in a 10% ethanol3 and 35 ◦C. Racemic flurbiprofen concentrations in six different levels: 0.05, 0.2, 060 nm; flow rate of 1 mL/min; injection volumes of 100 �L and 1 mL.

able 4. Fig. 16 corroborates that the increase of system temper-ture (from 23 to 35 ◦C) implies a decrease in both retention andelectivity. This decrease influences the performance of the SMBrocess, as can be predicted by using the equilibrium theory model

able 4stimated model parameters for flurbiprofen adsorption isotherms for a 10% ethanol/90%

odel M N m1 m2

0% ethanol/90% n-hexaneLG3 22 3 – –LLG4 4 0.5926LLG5 5 0.6086 0.5638

-hexane/0.01% TFA mobile phase composition and at two different temperatures:, 2.0 and 4.0 g/L; preparative column (particle diameter of 20 �m); UV detection at

and the adsorption model parameters obtained by fitting the pre-vious experimental adsorption measurements. Fig. 17 shows thatthe separation region for 35 ◦C is located at lower values of �2 and�3 (due to its lower retention times) and is smaller (due to lower

n-hexane/0.01% TFA mobile phase composition and at 35 ◦C.

Q b1 b2 SQ SD

250.3 6.950 × 10−3 9.284 × 10−3 0.9223 0.2203106.0 1.072 × 10−2 1.699 × 10−2 0.7334 0.2019107.5 1.034 × 10−2 1.709 × 10−2 0.7221 0.2061

A.E. Ribeiro et al. / Separation and Purification Technology 68 (2009) 9–23 21

Fig. 16. Comparison between model (modified linear + Langmuir model, LLG5) and experimental results for the equilibrium adsorption isotherms of flurbiprofen enantiomersin a 10% ethanol/90% n-hexane/0.01% TFA mobile phase composition and at two different temperatures: 23 and 35 ◦C. Open circles for experimental concentration of theR(−)-Flurbiprofen enantiomer; closed circles for experimental concentration of the S(+)-Flurbiprofen enantiomer; solid lines for adsorption isotherm model.

Fig. 17. SMB separation regions predicted under the equilibrium theory (negligible mass transfer resistances and axial dispersion) in a 10% ethanol/90% n-hexane/0.01% TFAmobile phase composition, at two different temperatures (23 and 35 ◦C) and two total feed concentrations (CF

1 + CF2 ): 0.1 g/L (linear range) and 25 g/L (non-linear range).

Ff

ig. 18. Prediction of productivity and solvent consumption of SMB operation using theunction of feed concentration in a 10% ethanol/90% n-hexane/0.01% TFA mobile phase co

equilibrium theory (negligible mass transfer resistances and axial dispersion) as amposition and at two temperatures: 23 (solid line) and 35 ◦C (traced line).

2 Purific

sbtabio

5

ctftbvtasets

tcmadooTfscd

mtf

sbl(

A

(naPm

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

2 A.E. Ribeiro et al. / Separation and

electivity) than the one obtained at 23 ◦C. This result is obtained foroth low (0.1 g/L) and high (25 g/L) feed concentrations and justifieshe decrease on system productivity predicted for SMB operationnd presented in Fig. 18. From the results presented before, it cane concluded that, despite higher solubility, there is no advantages

n using a higher operating temperature, since it leads to a decreasen system productivity.

. Conclusions

The separation of enantiomers by liquid chromatography usinghiral stationary phases commonly uses alcohol/hydrocarbon mix-ures as mobile phase. A high hydrocarbon content is usuallyollowed, since it frequently presents better selectivity. However,he use of a high hydrocarbon content also implies low solu-ility of the racemate and high retention times. Although notery relevant at an analytical scale, this represents an impor-ant drawback at preparative and production scales. Low solubilitynd high retention times imply low system productivity and higholvent consumption for frontal and SMB chromatography. Thisxplains why the optimization of the mobile phase composi-ion can lead to different results at analytical and preparativecales.

For the preparative separation of flurbiprofen enantiomers,his work shows that a 10% ethanol/90% n-hexane mobile phaseomposition is the best choice, since it represents a good compro-ise between selectivity, retention time and solubility. The use ofhigher alcoholic content leads to worse results because of the

ecrease on system selectivity. This is a different conclusion to thene obtained in a previous work for the preparative separationf ketoprofen enantiomers, where pure ethanol was chosen [1].hese results revealed that there are no predictive general rulesor the optimization of mobile phase composition at preparativecale and that an individualized study must be carried out, sinceompounds of the same family (profen enantiomers) can lead toifferent solutions.

The results presented in this work also support that a low TFAodifier content (0.01%) is enough and that the operation at room

emperature (23 ◦C) is a good choice, since no advantages wereound using higher operating temperatures (35 ◦C).

The modeling and simulation tools used in this work prove to beuitable for prediction, since very good agreements were obtainedetween experimental and simulation results, not only at low (ana-

ytical scale, linear range), but also at high feed concentrationspreparative scale, non-linear range).

cknowledgements

Financial support by the Portuguese R&D foundation FCTFundacão para a Ciência e a Tecnologia) and European Commu-ity through FEDER (project POCI/EQU/59738/2004), is gratefullycknowledged. The authors wish to thank Simão P. Pinho (Bragancaolytechnic Institute) for the support on the solubility measure-ents.

eferences

[1] A. Ribeiro, N. Graca, L. Pais, A. Rodrigues, Preparative separation of ketopro-fen enantiomers: choice of mobile phase composition and measurement ofcompetitive adsorption isotherms, Sep. Purif. Technol. 61 (2008) 375–383.

[2] D. Groot, E. Vries, H. Groen, S. Jong, Non-steroidal anti-inflammatory drugs to

potentiate chemotherapy effects: from lab to clinic, Crit. Rev. Oncol. Hematol.61 (2007) 52–69.

[3] M. Zemskova, W. Wetcher, S. Bashkirova, C. Chen, R. Reiter, M. Lilly, Geneexpression profiling in R-flurbiprofen-treated prostate cancer: R-flurbiprofenregulates prostate stem cell antigen through activation of AKT kinase, Biochem.Pharmacol. 72 (2006) 1257–1267.

[

[

ation Technology 68 (2009) 9–23

[4] P. Andrews, X. Zhao, J. Allen, F. Li, M. Chang, A comparison of the effectivenessof selected non-steroidal anti-inflammatory drugs and their derivatives againstcancer cells in vitro, Cancer Chemother. Pharmacol. (March) (2007).

[5] E. Quann, F. Khawaja, K. Zavitz, D. Djakiew, The aryl propionic acid R-flurbiprofen selectivity induces p75NTR-dependent decreased survival ofprostate tumor cells, Cancer Res. 67 (2007) 3254–3262.

[6] W. Wetcher, E. Murray, D. Kantoci, D. Quiggle, D. Leipold, K. Gibson, J.McCracken, Treatment and survival study in the C57BL/6J-APCMin/+ (Min)mouse with R-flurbiprofen, Life Sci. 66 (2000) 745–753.

[7] S. Grösch, K. Schilling, A. Janssen, T. Maier, E. Niederberger, G. Geisslinger,Induction of apoptosis by R-Flurbiprofen in human colon carcinoma cells:involvement of p53, Biochem. Pharmacol. 69 (2005) 831–839.

[8] C. Szekely, J. Thorne, P. Zandi, M. Ek, E. Messias, J. Breitner, S. Goodman, Nons-teroidal anti-inflammatory drugs for the prevention of Alzheimer’s disease: asystematic review, Neuroepidemiology 23 (2004) 159–169.

[9] R. Mohs, C. Kawas, M. Carrillo, Optimal design of clinical trial for drugs designedto slow the course of Alzheimer’s disease, Alzheimer’s Dement. 2 (2006)131–139.

[10] G. Wilcock, Disease-modifying treatments for Alzheimer’s disease: a perspec-tive based on experience with R-flurbiprofen, Alzheimer’s Dement. 2 (2006)150–152.

[11] P. McGeer, E. McGeer, NSAIDs and Alzheimer’s disease: epidemiological, animalmodel and clinical studies, Neurobiol. Aging 28 (2007) 639–647.

12] J. Bojarski, H. Aboul-Enein, A. Ghanem, What’s new in chromatographic enan-tioseparations, Curr. Anal. Chem. 1 (2005) 59–77.

[13] B. Patel, J. Valentova, A. Hutt, Chromatographic separation and enantiomericresolution of flurbiprofen and its major metabolites, Chromatographia 55(2002) 135–142.

[14] B. Patel, J. Valentova, A. Hutt, Stereospecific analysis of flurbiprofen and itsmajor metabolites in plasma and urine by chiral-phase liquid chromatography,Chromatographia 57 (2003) 7–18.

[15] T. Booth, I. Wainer, Mechanistic investigation into the enantioselective separa-tion of mexiletine and related compounds, chromatographed on an amylosetris(3,5-dimethylphenylcarbamate) chiral stationary phase, J. Chromatogr. A741 (1996) 205–211.

[16] G. Duan, C. Ching, Preparative scale enantioseparation of flurbiprofen by lipase-catalysed reaction, Biochem. Eng. J. 2 (1998) 237–245.

[17] H. Zhang, X. Wang, C. Ching, R-stereopreference analysis of Lipase Novozym435 in kinetic resolution of flurbiprofen, Chirality 19 (2007) 245–249.

[18] L. Qian, S. Chen, B. Shi, Preparation of enantiopure (R)-flurbiprofen catalysedby a newly isolated Bacillus cereus C71, Biocatal. Biotransf. 25 (1) (2007) 29–34.

[19] E. Lee, H. Won, H. Ro, Y. Ryu, B. Chung, Preparation of enantiomerically pure (S)-flurbiprofen by an esterase from Pseudomonas sp. KCTC 10122BP, J. Mol. Catal.B: Enzym. 26 (2003) 149–156.

20] G. Shin, K. Lee, T. Kim, H. Shin, Y. Lee, Lipase-catalyzed production of opti-cally active (S)-flurbiprofen in aqueous phase reaction system containing chiralsuccinyl �-cyclodextrin, J. Mol. Catal. B: Enzym. 33 (2005) 93–98.

21] G. Shin, K. Lee, Y. Lee, Fed-batch production of (S)-flurbiprofen in lipase-catalysed dispersed aqueous phase reaction system induced by succinyl�-cyclodextrin and its extractive purification, J. Mol. Catal. B: Enzym. 37 (2005)109–111.

22] M. Mazzotti, G. Storti, M. Morbidelli, Optimal operation of simulated mov-ing bed units for nonlinear chromatographic separations, J. Chromatogr. A 769(1997) 3–24.

23] L.S. Pais, J.M. Loureiro, A.E. Rodrigues, Modeling, simulation and operation of asimulated moving bed for continuous chromatographic separation of 1,1′-bi-2-naphthol enantiomers, J. Chromatogr. A 769 (1997) 25–35.

24] L.S. Pais, J.M. Loureiro, A.E. Rodrigues, Separation of enantiomers of a chiralepoxide by simulated moving bed chromatography, J. Chromatogr. A 827 (1998)215–233.

25] D.C.S. Azevedo, A.E. Rodrigues, Design of a simulated moving bed in the pres-ence of mass-transfer resistances, AIChE J. 45 (1999) 956–966.

26] C. Migliorini, A. Gentilini, M. Mazzotti, M. Morbidelli, Design of simulatedmoving bed units under nonideal conditions, Ind. Eng. Chem. Res. 38 (1999)2400–2410.

27] G. Biressi, O. Ludemann-Hombourger, M. Mazzotti, R.M. Nicoud, M. Morbidelli,Design and optimisation of a simulated moving bed unit: role of deviationsfrom equilibrium theory, J. Chromatogr. A 876 (2000) 3–15.

28] A.E. Rodrigues, L.S. Pais, Design of SMB chiral separations using the concept ofseparation volume, Sep. Sci. Technol. 39 (2004) 245–270.

29] L.S. Pais, V.G. Mata, A.E. Rodrigues, Simulated moving bed and relatedtechniques, in: G. Cox (Ed.), Preparative Enantioselective Chromatography,Blackwell Publishing, Oxford, UK, 2005, p. 176 (Chapter 7).

30] S. Gracin, A. Rasmunson, Solubility of phenylacetic acid, p-hydroxyphenylaceticacid, p-aminophenylacetic acid, p-hydroxybenzoic acid, and ibuprofen in puresolvents, J. Chem. Eng. Data 47 (2002) 1379–1383.

31] K. Lynam, R. Stringham, Chiral separations on polysaccharide stationary phasesusing polar organic mobile phases, Chirality 18 (2006) 1–9.

32] J. Sørensen, W. Arlt, Liquid–Liquid Equilibrium Data Collection: Binary Systems,in DECHEMA, Chemistry Data Series, vol. V, Part 1, 1979.

33] A. Trejo, P. Yanez, R. Rincón, Liquid–liquid coexistence curves for binary sys-tems: methanol + cyclohexane + several isomers of hexane, J. Chem. Eng. Data51 (2006) 1070–1075.

34] J. Blackwell, Effect of acidic mobile phase additives on chiral selectivity forphenylalanine analogs using subcritical fluid chromatography, Chirality 11(1999) 91–97.

Purific

[

[

A.E. Ribeiro et al. / Separation and

35] Y. Ye, R. Stringham, Effect of a mobile phase acidic additives on enan-tioselectivity for phenylalanine analogs, J. Chromatogr. A 927 (2001) 47–52.

36] Y. Ye, B. Lord, R. Stringham, Memory effect of mobile phase additives inchiral separations on a Chiralpak AD column, J. Chromatogr. A 945 (2002)139–146.

[

[

ation Technology 68 (2009) 9–23 23

37] Y. Ye, B. Lord, L. Yin, R. Stringham, Enantioseparation of amino acids on apolysaccharide-based chiral stationary phase, J. Chromatogr. A 945 (2002)147–159.

38] N. Matthijs, M. Maftouh, Y. Heyden, Screening approach for chiral separationof pharmaceuticals IV. Polar organic solvent chromatography, J. Chromatogr. A1111 (2006) 48–61.