N

IB

a

AA

KPCCG

1

mcpchb

sf2ff

catuoaam(st

0d

Journal of Chromatography A, 1216 (2009) 4201–4205

Contents lists available at ScienceDirect

Journal of Chromatography A

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

ew 18-l process-scale counter-current chromatography centrifuge

an Sutherland ∗, Peter Hewitson, Svetlana Ignatovarunel Institute for Bioengineering, Brunel University, Uxbridge, Middlesex UB8 3PH, UK

r t i c l e i n f o

rticle history:vailable online 9 December 2008

eywords:

a b s t r a c t

A new Dynamic Extractions Maxi-counter-current chromatography (CCC) centrifuge with a column vol-ume of 18-l has been installed in the Advanced Bioprocessing Centre at Brunel. This instrument has fourtimes the capacity of the 4.6-l Maxi-CCC centrifuge which has been operating robustly for 3 years. Tests

rocess scaleounter-current chromatographyCCUESS

using the model sample system benzyl alcohol and p-cresol with a heptane:ethyl acetate:methanol:water(HEMWat) phase system (1.4:0.1:0.5:1.0) show that resolution is almost double with this new high capac-ity device. Commissioning tests with a mixture of caffeine, KD = 0.21; ferulic acid, KD = 0.82; umbelliferone,KD = 1.2 and vanillin, KD = 1.49 using a HEMWat phase system of 1:1.5:1:1.5 on the 9-l column show thatresolutions equivalent to analytical instruments will be possible using the full 18-l capacity. They alsoshow that predictable scale-up from simple test tube tests is feasible with knowledge of the stationary

lanne

phase retention for the p

. Introduction

The universal acceptance by industry of counter-current chro-atography will depend on whether it can be seen as a

ompetitive/complementary separation system in its tool chest ofrocesses. Predictable scale-up has been demonstrated [1,2], butomparative studies at the process scale where 30 kg of crude areandled in a few days for tox trials are required before industry wille convinced of the competitiveness of the process.

At Brunel we have been performing dedicated research into thecale-up of counter-current chromatography and have invested inacilities like the Advanced Bioprocessing Centre (opened in April006 by Dr. Yoichiro Ito of the National Institutes of Health, USA,ounder of the technology) to house both pilot and process scaleacilities in its hazards laboratory.

The 4.6-l capacity Dynamic Extractions (Slough, UK) Maxi-CCCentrifuge (R = 300 mm; d = 10 mm) has been running routinelynd reliably now for 3 years. A number of commercial manufac-uring runs, involving repeatability studies, have been performednder contract, so very little has been published. Typical examplesf process scale separations using the 4.6-l Maxi-CCC centrifugere the production of glucoraphanin, a cancer chemoprotectivegent derived from broccoli seeds [3,4] and honokiol, an iso-

eric compound isolated from Magnolia officianalis Rehd. et Wils.

Chinese name Houpu) from the Magnoliaceae family [5] whichubsequently formed the basis for a new synthesized product andreatment regime for cancer biotherapy [6]. Sample loadings on the

∗ Corresponding author. Tel.: +44 1895 266 920; fax: +44 1895 274 608.E-mail address: ian.sutherland@brunel.ac.uk (I. Sutherland).

021-9673/$ – see front matter © 2008 Elsevier B.V. All rights reserved.oi:10.1016/j.chroma.2008.11.097

d process scale run.© 2008 Elsevier B.V. All rights reserved.

4.6-l centrifuge have typically been 5% of the coil/column volume(230 ml) and have varied from 100 to 500 mg/ml giving through-puts of up to 8 kg a day when working with the bobbins in paralleland 4 kg/day in series working a 2-shift 16-h day.

This paper describes the new 18-l Maxi-CCC centrifuge givingpreliminary results on its potential for process scale separations.

2. Design features of the new 18-l Maxi-CCC centrifuge

A new Maxi-CCC centrifuge with a total column volume of18-l has been installed by Dynamic Extractions in the hazard’slaboratory of the Advanced Bioprocessing Centre (ABC) at BrunelUniversity (Fig. 1). The 4.6-l and 18-l Maxi-CCC centrifuges aremounted side by side with a common liquid handling controlsystem which can operate each centrifuge in parallel or seriesdepending on the required resolution. Column capacities of 2.3,4.6, 9 and 18 l are now available to cope with separations requiringdifferent resolution. Liquid handling flow rates can vary between0–3000 ml/min for parallel operation and 0–1500 ml/min for seriesoperation. The 18-l centrifuge is the one on the right of Fig. 1 andis shown here from the control room with its shroud raised to giveeasy all round access to the centrifuge itself. One feature of the newcentrifuge is that it has a smaller footprint than the 4.6-l centrifugedespite having four times the capacity. The dimensions in metres(width, depth and height) of the 4.6-l centrifuge with its shroud are1.5 × 1.8 × 1.9 and of the 18-l centrifuge are 1.3 × 2.0 × 1.6. Note that

these dimensions are with the shrouds closed for full containment.The 4.6-l centrifuge has access doors opening at the back and frontrequiring an additional footprint area, while the 18-l centrifuge hasa shroud that can be raised for all round access, which requires aminimal extra footprint area.

4202 I. Sutherland et al. / J. Chromatogr

Fig. 1. The hazards laboratory in the Brunel Advanced Bioprocessing Centre asvttt

ttEmh

aiaeSih

3

3

fwwc(vcwS

Fa

iewed from the control room showing the 4.6-l Maxi centrifuge on the left andhe 18-l Maxi centrifuge on the right with its shroud raised. The liquid handling sys-em is positioned between the two centrifuges and can be switched between thewo from the control consol shown in the foreground.

Another unique feature of this centrifuge is its belt drive sys-em. All counter-current J type centrifuges built before 2008 haveraditionally had gear driven planets which are very noisy. Dynamicxtractions are currently the only manufacturer to have introduceduch quieter belt drives to their range of CCC instruments which

ave resulted in significant noise reductions.The new casing and shroud arrangement gives much better

ccess to the centrifuge for routine servicing and changing out fly-ng leads. Also there is a viewing window on each side of the shroudnd windows through to the “Tail” and “Head” ends of the tubing onach bobbin if stroboscopic visualisation is required. Fig. 2 showsvetlana Ignatova setting up the centrifuge during the commission-ng process. Note the tubing-viewing window just below her rightand.

. Experimental

.1. Reagents and materials

All solvents used were of analytical grade and purchasedrom Fisher Chemicals (Loughborough, UK), deionised wateras purified using a Purite Select Fusion (Thame, UK) pureater system (0.4 �m filter). Benzyl alcohol (99%) and p-

resol (4-methylphenol, 99%) were supplied by Sigma–AldrichGillingham, UK). GUESS (“generally useful estimate of sol-

ent systems for CCC” after Friesen and Pauli [7]) compoundsaffeine (C) and umbelliferone (U) were supplied by Fisher,hile ferulic acid (F) and vanillin (V) were supplied by

igma.

ig. 2. Close up of the 18-l Maxi rotor being set up by Dr. Svetlana Ignatova givingn idea of its size.

. A 1216 (2009) 4201–4205

3.2. Apparatus

The 18-l Maxi-CCC centrifuge was supplied by Dynamic Extrac-tions. It has two bobbins, symmetrically mounted, of 9-l capacityeach which can be operated in parallel or series or as individualbobbins. The centrifuge parameters are given in Table 1 for com-parison with other ABC centrifuges at the analytical (Mini) andlaboratory/preparative (Midi) scale. The tubing bore and the rotorradius are identical to the 4.6 l centrifuge. The centrifuge is designedto rotate up to 850 rpm (242 × g) but its normal operating speed is600 rpm for most solvent systems. The centrifuge was rotated in adirection which placed the “head” of the column at the centre andthe “tail” at the periphery as recommended in [8].

3.3. Operating system

The operating system is controlled by a purposed-built com-puter controlled liquid handling system (Armen Instrument,Vannes, France) which continuously monitors system pressures,flows and flow paths. There are two 6 head pumps as shown inFig. 1, the left hand one being the elution pump and the right handone being the sample pump. If the heads are numbered E1 to E6and S1 to S6 for the elution and sample pumps respectively, thenheads E1, E3 and E5 can be pneumatically activated to pump theupper phase as the mobile phase while the other heads are with-drawn and dormant, and heads E2, E4 and E6 can be selected forthe lower phase being the mobile phase in a similar way. For thesample, pump heads S1, S3 and S5 are arranged to pump sample inupper phase, while S2, S4 and S6 are arranged to pump sample inlower phase. For the sample pump all six heads can operate simul-taneously to load the sample continuously for a given volume in a50:50 mixture of upper and lower phase. The display (Fig. 3) allowsthe operator to select which operation procedure to use (isocraticor continuous extraction), whether to operate in parallel or series,whether to operate in reverse phase or normal phase and whetherto select the 4.6-l or 18-l centrifuge. The operator gets a UV spec-trophotometer display of the chromatogram which allows “mouse”selected fraction collection remotely.

3.4. Operating procedure

The pumps were initially primed with upper and lower phasesand all lines of the pumping system were flushed with phase sys-tem to ensure there was no trapped air. The rotor was rotatedslowly (14 rpm) such that the head was at the centre of the bob-bin. The coils/columns were filled from periphery to centre withupper phase in parallel (1000 ml/min for 30 s, then 2000 ml/minfor 30 s, then 3000 ml/min for 7 min). All air in the column, beingthe lighter phase, was expelled from the head. The rotation speedwas then increased to 600 rpm (121 × g). The columns where equi-librated in parallel, with the lower phase flowing at 1600 ml/min(800 ml/min through each column). Once break through had beenobserved, the flow was switched to series through the columns andthe flow rate adjusted to 850 ml/min. Injection was carried out oncea stable UV trace was observed. The sample was injected directlyonto the column using the sample pump at a flow rate similar tothe mobile phase flow rate. Fractions were collected every minuteto allow analysis of stationary phase stripping and for HPLC analysis.At the end of each run the columns were emptied with compressedair (4 bar). The coils/columns were rotated such that the head wasat the periphery and the tail was at the centre. Compressed air was

applied to the centre of the column and the resultant pump-out col-lected from the periphery of each column separately into measuringcylinders to confirm stationary phase retention in each one.

Running conditions for the benzyl alcohol/p-cresol separa-tion were similar to that reported for the 4.6-l centrifuge [2]:

I. Sutherland et al. / J. Chromatogr. A 1216 (2009) 4201–4205 4203

Table 1Design parameters of the Mini, Midi and Maxi centrifuges at Brunel.

Centrifuge Rotor radius (mm) Bobbin Tubing bore (mm) Length (m) Volume (ml)

Mini 50 1 0.8 35.8 18Midi 110 1 4 36.3 456Midi 110 1 + 2 4 72.6 912Maxi 300 1 10 29.3 2,300MNN

su2ipm

3

ba

C2t6

(

Fs

axi 300 1 + 2ew Maxi 300 1ew Maxi 300 1 + 2

peed 600 rpm, flow 850 ml/min, sample mass 57 g, sample vol-me 920 ml (20% column volume) and concentrations of 42 and0 mg/ml for benzyl alcohol and p-cresol, respectively. This resulted

n a resolution Rs = 0.71. The percentage column volume for the sam-le in the 18-l Maxi-CCC was 5%. Running conditions for the GUESSixture were: speed 600 rpm, flow 150 ml/min, column volume 9 l.

.5. Preparation of phase and sample systems

A hydrophobic phase system (HEMWat-4A) was used for theenzyl alcohol/p-cresol separation, containing n-heptane, ethylcetate, methanol and water with volume ratios of 1.4:0.1:0.5:1.0.

For the benzyl alcohol/p-cresol separation on the 18-l Maxi-CC centrifuge, 1100 ml of the sample (42 g/l benzyl alcohol,

0 g/l p-cresol made up in HEMWat-4A lower phase) was loadedhrough the sample pump at 850 ml/min, to give a total loading of8.2 g.

For the GUESSmix separation, a more polar phase systemHEMWat-15, 1:1.5:1:1.5, v/v/v/v) was used. Due to the large quan-

ig. 3. The operating system in series and descending (reversed phase) mode with loweome stationary phase stripping (the noise on the trace) before sample components elute

10 58.6 4,60010 114.6 9,00010 229.2 18,000

tities of phase system required to carry out the various runs theupper and lower phases were made up separately [9].

For the GUESSmix separation, 70 ml of the sample (1.05 g/l C,0.67 g/l F, 0.80 g/l U and 0.63 g/l V made up in HEMWat-15 lowerphase) was loaded through the sample pump at 150 ml/min, to givea total load of 0.22 g.

3.6. Theoretical model

The theoretical model developed by de Folter and Sutherland[10] was used. Partition coefficients of the compounds with thevalue of stationary phase retention are used as input parameters.The number of equivalent mixing and settling transfers is estimatedto give the “best fit” to the results.

4. Results and discussion

Two test systems were used. The first was the standard binarytest mixture of benzyl alcohol and p-cresol used in previous

r phase mobile flowing from head (centre) to tail (periphery). The UV trace shows.

4204 I. Sutherland et al. / J. Chromatogr. A 1216 (2009) 4201–4205

F resol on (a) the 4.6-l centrifuge [2] and (b) the 18-l centrifuge. Running conditions for bothc ple mass 57 g; stationary phase retention 47.4% for the 4.6-l centrifuge. Sample volume1 tion of 66% prior to injection falling to 31% at the end of the run.

sscltwt4ao

f(uiawvku7amca4rp

Fu1t

ig. 4. Plot of HPLC peak area against time for a separation of benzyl alcohol and p-centrifuges: speed 600 rpm (121 × g); flow 850 ml/min. Sample volume 920 ml; sam100 ml, sample mass of 68.2 g for the 18-l centrifuge with a stationary phase reten

tudies [2,11] and the second was a mixture from the GUESS testystem proposed by Friesen and Pauli [7]. Fig. 4a and b shows theomparative separation of benzyl alcohol and p-cresol in the 4.6-Maxi and 18-l Maxi, respectively. As the 18-l centrifuge has fourimes the capacity of the 4.6-l one, an increase in resolution of 2×ould be expected. Whereas the chromatogram in Fig. 4a shows

hat benzyl alcohol and p-cresol are only partially resolved on the.6-l centrifuge, they are completely resolved on the 18-l centrifugend the ratio of resolution was found to be 1.85 when a similar massf sample was injected.

Four components from the GUESS mix [7], caffeine (C), KD = 0.21;erulic acid (F), KD = 0.82; umbelliferone (U), KD = 1.2 and vanillinV), KD = 1.49 were run on one of the 18-l Maxi’s bobbins with a vol-me of 9-l. The U and V compounds were only partially resolved

n Friesen and Pauli’s original paper [7] after 6 h of an optimisednalytical run. The same experimental conditions were used hereith flow rate increased from 1.5 to 150 ml/min (100×) and column

olume increased from 324 ml to 9-l (28×). Sample volume wasept approximately the same as used by Friesen and Pauli [7]. Theysed a sample loop of 5 ml [1.5% of column volume (CV)]. We used0 ml (0.8% CV); sample concentrations were in the same order atround 3–4 g/l. Fig. 5 shows the HPLC fraction analysis of the GUESSixture run on the 9-l new Maxi column. It can be seen that C is

ompletely resolved from F (Rs = 1.96, fractions 24–32, 100% puritynd recovery), F is partially resolved from U (Rs = 0.82, fractions0–49, 98.5% purity, 86.2% recovery) and U and V are only partiallyesolved (Rs = 0.56) as in the Friesen and Pauli paper. The final com-ound eluting (V) can be shaved (fractions 62–75) to give 98.7%

ig. 5. HPLC fraction analysis of GUESS compounds caffeine (C), ferulic acid (F),mbelliferone (U) and vanillin (V). Running conditions: speed 600 rpm, flow50 ml/min, sample volume 70 ml, sample mass 0.22 g, mean stationary phase reten-ion 60.4%.

Fig. 6. HPLC Chromatogram from Fig. 5 compared to the theoretical prediction basedon test tube partition coefficients and the 60.4% retention of the stationary phase.

purity and 88.7% yield as determined by HPLC. Umbelliferone (U)is difficult to get in pure form. Fraction 56 was 99.2% pure but onlygives a 13.5% yield. Combining fractions 55–57 gives a 97.9% puritywith an increased yield of 38.9%. Fig. 6 shows the chromatogramfrom the sum of all compounds, which in fact mimics the UV trace(not shown) compared to a theoretical prediction based on partitioncoefficients and the measured stationary phase retention [10]. Themodel can then be used to predict whether it would be worth doingan 18-l run to resolve F and U. The model prediction (not shown)shows that F and U would be resolved but that U and V would stillnot quite be completely resolved, but competitive with analyticalseparations [7]. Note that due to the higher g-field, separation timeswere considerably shorter than in Friesen and Pauli’s original studywith an increase in throughput from 0.62 mg/h (3.75 mg in 6 h) to176 mg/h (0.22 g in 1.25 h).

These results show that volumetric scale-up is both feasible andpredictable from simple partition tests and knowledge of stationaryphase retention. Now higher resolution (˛ factors greater than 1.5)separations are possible at large scale using high-performance CCC.

5. Conclusions

The new 18-l Maxi-CCC centrifuge with column lengths four

times longer than its sister 4.6-l centrifuge has been demonstratedto give the predicted increase in resolution expected. Tests usingmixtures of compounds ranging in partition coefficients from 0.21to 1.49 demonstrate elution behaviour predicted from test tubepartition tests and chromatograms from the literature once the

atogr

rstt

A

(1ctFtFLii

23 (2000) 2259.[9] I. Garrard, L. Janaway, D. Fisher, J. Liq. Chromatogr. Relat. Technol. 30 (2007)

I. Sutherland et al. / J. Chrom

etention for a given process scale run is known. This significantlyimplifies the scale-up process putting the emphasis on optimisinghe separation factors analytically using high throughput screeningechniques for solvent selection.

cknowledgements

The authors would like to thank the Research Councils EPSRCgrant ref: GR/R03143/01) and BBSRC (grant refs: BBD524583/1;92/SBRI9675) for their support of research on the scale-up ofounter-current chromatography and helping to equip the cen-re. Thanks also to Brunel University for its support via Westocus, SRIF2 and HEIF4 which has helped fund the development of

he Advanced Bioprocessing Centre and the Maxi-CCC centrifuges.inally the authors are extremely grateful to Dynamic Extractionstd. and in particular Dr. Philip Wood and Lee Janaway for their ded-cation in producing such an innovative design of centrifuge whichs the first of its kind in the world.

[[

. A 1216 (2009) 4201–4205 4205

References

[1] I.A. Sutherland, J. Chromatogr. A 1151 (2007) 6.[2] P. Wood, S. Ignatova, L. Janaway, D. Keay, D. Hawes, I. Garrard, I.A. Sutherland, J.

Chromatogr. A 1151 (2007) 25.[3] K.L. Wade, K.K. Stephenson, F.E. Chou, J. Chromatogr. A 996 (2003)

85.[4] D. Fisher, I.J. Garrard, R. van den Heuvel, I.A. Sutherland, F.E. Chou, J.W. Fahey, J.

Liq. Chromatogr. Relat. Technol. 28 (2005) 1913.[5] L. Chen, Q. Zhang, G. Yang, L. Fan, I. Garrard, S. Ignatova, D. Fisher, I.A. Sutherland,

J. Chromatogr. A 1142 (2007) 115.[6] Y. Luo, Y. Xu, L. Chen, H. Luo, C. Peng, J. Fu, H. Chen, A. Peng, H. Ye, D.-C. Xie, A.

Fu, J. Shi, S. Yang, Y. Wei, J. Chromatogr. A 1178 (2007) 160.[7] J.B. Friesen, G.F. Pauli, J. Liq. Chromatogr. Relat. Technol. 28 (2005) 2777.[8] I.A. Sutherland, J. Muytjens, M. Prins, P. Wood, J. Liq. Chromatogr. Relat. Technol.

151.10] J. de Folter, I.A. Sutherland, J. Chromatogr. A 1216 (2009) 4218.11] S. Ignatova, P. Wood, D. Hawes, L. Janaway, D. Keay, I. Sutherland, J. Chromatogr.

A 1151 (2007) 20.