Da

Da

b

c

a

ARRAA

KAMMP

1

iAPcdtuiat(

nf

(

0h

Journal of Chromatography A, 1249 (2012) 1– 7

Contents lists available at SciVerse ScienceDirect

Journal of Chromatography A

jou rn al h om epage: www.elsev ier .com/ locat e/chroma

esign of a microfluidic platform for monoclonal antibody extraction using anqueous two-phase system

.F.C. Silvaa,c, A.M. Azevedob,c, P. Fernandesb,c, V. Chua, J.P. Condea,b, M.R. Aires-Barrosb,c,∗

INESC Microsistemas e Nanotecnologias and IN-Institute of Nanoscience and Nanotechnology, Lisbon, PortugalDepartment of Bioengineering, Instituto Superior Técnico, Technical University of Lisbon, Lisbon, PortugalIBB – Institute for Biotechnology and Bioengineering, Centre for Biological and Chemical Engineering, Instituto Superior Técnico, Technical University of Lisbon, Lisbon, Portugal

r t i c l e i n f o

rticle history:eceived 21 March 2012eceived in revised form 23 May 2012ccepted 25 May 2012vailable online 12 June 2012

ey words:queous two-phase systemsonoclonal antibodiesicrofluidic device

rocess simulation

a b s t r a c t

The use of monoclonal antibodies (mAbs) in medical treatments and in laboratory techniques has a veryimportant impact in the battle against many diseases, namely in the treatment of cancer, autoimmunediseases and neural disorders. Thus these biopharmaceuticals have become increasingly important, rein-forcing the demand for efficient, scalable and cost-effective techniques for providing pure antibodies.Aqueous two-phase systems (ATPS) have shown potential for downstream processing of mAbs. In thiswork, an ATPS in a microfluidic platform was designed and tested for mAbs extraction. The system demon-strated the potential to be an effective tool to accelerate bioprocess design and optimization. The partitionof immunoglobulin G (IgG) tagged with fluorescein isothiocyanate (FITC) in an ATPS of polyethylene-glycol (PEG)/phosphate buffer with NaCl was investigated using a PDMS microfluidic device fabricatedusing soft lithography techniques. Different structures were tested with different values of microchannellength (3.14–16.8 cm) and flow rates of the salt (1–2 �L/min) and PEG-rich phases (0.2–0.5 �L/min). A

stable interphase between the phases was obtained and the phenomena of diffusion and of partitionof the IgG from the salt-rich phase to the PEG-rich phase were measured by fluorescence microscopy.Process simulation allowed the modeling of the IgG diffusion and partitioning behavior observed in themicrostructure. The reduction to the microscale does not greatly affect the antibody extraction yield whencompared with macroscale results, but it does reduce the operation time, demonstrating the potentialityof this approach to process optimization.

. Introduction

The need for therapeutic proteins, such as monoclonal antibod-es, for the treatment of age and society-related diseases, such aslzheimer’s, Parkinson’s, cancer, or diabetes, continues to increase.resently more than 250 of these therapeutic products are inlinical trials and hundreds more are at the preclinical stage ofevelopment [1]. Downstream processing (dsp) of therapeutic pro-eins has however failed to keep up with the throughput of thepstream stage, resulting in a production bottleneck and in escalat-

ng costs [2]. Improved dsp processes are therefore actively sought,

mong them aqueous two phase systems (ATPS), which rely onhe selective partitioning of a molecule in a biocompatible polymersalt)/polymer environment. The widespread use of ATPS has been

∗ Corresponding author at: Department of Bioengineering, Instituto Superior Téc-ico, Technical University of Lisbon, Lisbon, Portugal. Tel.: +351 218419065;

ax: +351 218419062.E-mail addresses: rabarros@ist.utl.pt, pcbarros@alfa.ist.utl.pt

M.R. Aires-Barros).

021-9673/$ – see front matter © 2012 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.chroma.2012.05.089

© 2012 Elsevier B.V. All rights reserved.

hampered by poorly understood partition mechanisms, empiricaland time-consuming method development, and the need to evalu-ate a large number of variables for process optimization [2–5].

Microscale process techniques have appeared as effective toolsfor accelerating bioprocess design in a cost-effective manner sincethe 1990s [6] nevertheless the trend to a widespread use has onlyemerged more recently [7]. This approach allows for a large numberof variables to be evaluated in a parallel process, reducing the timeand sample/reagent volumes required. Microliter scale systems canbe designed in microwell or microfluidic format. Microfluidic sys-tems process microliter quantities of fluid within channels withdimensions of tens to hundreds of micrometers, and are com-patible with continuous flow processing [8]. Continuous flow isoften the mode of operation in large-scale processing, so the useof microfluidic devices brings high-throughput, lab-scale processoptimization closer to the final large-scale production processes[9]. Given the small dimensions of the microchannels, the flow

regime in a microfluidic environment is laminar, with mixingrelying on diffusion, and interfacial forces dominating over gravita-tional forces [10]. Microfluidic devices furthermore take advantageof high surface-to-volume ratios, which minimize heat and mass

2 omato

tpmdaHCfwslflb

e[tAret[Mtcupotb

fcbfddpponomewa

2

2

(pmASPu

2

b

D.F.C. Silva et al. / J. Chr

ransfer distances, and can allow a high level of parallelization, thusroviding useful tools for gaining early information on the perfor-ance of production systems. Most of the work with microfluidic

evices has centered on the reaction/metabolite production stagend there has been relatively little work in the dsp. Titchener-ooker et al. published pioneering work on microscale dsp [11].urrent microfluidic approaches to protein purification tend toocus on chromatography, affinity capture, and electrophoresis,hich are predominantly batch systems, requiring the injection of

mall and precise amounts of sample and whose efficiency is estab-ished only at the end of the separation. ATPS is particularly suitableor continuous operation in microfluidic devices [12], where theaminar regime and the low interfacial tension favor interface sta-ilization, a key requirement in ATPS [3,13].

Although the use of ATPS in a microscale is still relatively limited,xamples in the recovery of whole cells [14–18], tagged proteins19] with recovery of 75% of activity and crude membrane pro-eins with recovery yields of 91% [20] have been reported. AnTPS-microfluidic setup was also used for the purification of bacte-iorhodopsin, a light-sensitive membrane protein, from crude cellxtract [21]. For cell separation, a PEG/dextran system was usedo extract live and dead CHO K-1 cells from a culture medium15] and to extract leukocytes from whole blood samples [16,18].

icrofluidic extraction of a genetically tagged recombinant pro-ein has also been reported using a ATPS and about 85% of theontaminants were successfully removed [19]. An overview of these of aqueous two-phase systems in microfluidics was recentlyublished highlighting the opportunities for future developmentsf using of ATPS in microfluidic devices as an automated high-hroughput platform for separation, purification and analysis ofiomolecules [12].

In this paper, a microfluidic device was designed and testedor the partition of immunoglobulin G (IgG) tagged with fluores-ein isothiocyanate (FITC) in polyethylene-glycol (PEG)/phosphateuffer with NaCl in ATPS. The PDMS microfluidic devices wereabricated using soft lithography techniques. The optimized chipesign was a three inlets/three outlets serpentine structure. Thisesign allowed the formation of a stable interphase and the com-lete extraction of IgG from the salt-rich phase to the PEG-richhases. Fluorescence microscopy image allowed the monitoringf the IgG concentration both across and along the microchan-el. The extraction process was simulated to allow the modelingf the IgG diffusion and partitioning behavior. The application oficrofluidics to ATPS has the potential of combining the process

fficiency of ATPS with the reduced times and volumes associatedith microfluidics, as well as the possibility to multiplex and par-

llel process in real downstream processes.

. Materials and methods

.1. Chemicals and biologicals

Polyethylene glycol (PEG, MW 3350 Da) and sodium chlorideNaCl) were purchased from Sigma (St. Louis), while potassiumhosphate dibasic anhydrous (K2HPO4) and sodium phosphateonobasic anhydrous (NaH2PO4) were purchased from Panreac.

rabbit anti-goat IgG conjugated with FITC was purchased fromigma as an aqueous solution with a concentration of 3.6 mg/mL.olydimethylsiloxane (PDMS) was obtained from Dow Corning,nder the trade name of Sylgard® 184.

.2. Aqueous two-phase systems preparation

Stock solutions of 50% (w/w) PEG and 40% (w/w) phosphateuffer were prepared in milli-Q water. Solutions of NaH2PO4 at 40%

gr. A 1249 (2012) 1– 7

(w/w) and K2HPO4 at 40% (w/w) were prepared in mili-Q water.The phosphate buffer at pH 7.0 was prepared by mixing 40% (w/w)K2HPO4 solution and 40% (w/w) NaH2PO4 solution with a volumeratio of 2.13 (K2HPO4/NaH2PO4). The pH value was adjusted to 7.0with the addition of 40% (w/w) K2HPO4 solution. PEG/phosphatebuffer/NaCl ATPS were prepared by weighting the proper amountsof the stock solutions of 50% (w/w) PEG, 40% (w/w) phosphatebuffer, and NaCl, to obtain the desired final concentrations. Thetotal system weight was maintained at 10 g. After stirring andallowing for settling overnight, a clear interface separating thetwo phases was formed. The two phases were separated using a2 mL syringe and stored in different containers. The IgG-FITC anti-body was mixed with the salt-rich phase to a final concentration of0.2 g/L.

2.3. Microfluidic device design and fabrication

The soft lithography technique was used to fabricate themicrofluidic structures. Soft lithography involves the transfer ofthe pattern defining the microfluidic structure to a SU-8 (a neg-ative photoresist) mold via UV light exposure through an Al/quartzmask. The height of the microchannels, defined by the thicknessof the SU-8, was kept constant at 20 �m. PDMS was mixed in a10:1 base:curing agent weight ratio and then poured over the SU-8mold. The PDMS was cured for approximately 2 h and then peeledoff the SU-8 mold. Then, holes were punched to provide access tothe inlets and outlets and, afterwards, the PDMS was placed in a UVozone cleaner for surface oxidation and then bonded to a flat glasssubstrate. Polyethylene tubes were manually inserted in the inletsand outlets to provide access to the microchannel.

2.4. Antibody partitioning in microfluidic device

Both the PEG-rich phase and the salt-rich phase containing theIgG-FITC antibody were loaded into the channels simultaneouslyusing a syringe pump (New era pump systems, Inc. NE-300 model).The salt-rich phase was loaded in the middle inlet and the PEG-rich phase, in the two side inlets. The location of the interphasesand the IgG partitioning phenomena were observed on-chip usingvisible and fluorescence imaging, respectively, obtained with a dig-ital color camera DFC300 FX from Leica. The visible images wereobtained with an exposure rate of 1.13 ms and were used to mon-itor the interphase stability. Fluorescence images were obtainedwith several different exposure rates, ranging from 130 ms to 6.2 s,and enable the tracking of the IgG-FITC antibody diffusion/partitionfrom the salt-rich phase to the PEG-rich phase. The analysis of theimages was made with ImageJ software, which allowed obtain-ing the fluorescence intensity corresponding to the characteristicfluorescence emission wavelength of FITC (525 nm) at differentlocations in the micrographs. Because FITC is bound to the IgG, thefluorescence intensity across a perpendicular cross section of themicrochannel allows the measurement of IgG diffusion at differ-ent locations along the microchannel. The slope of the fluorescenceintensity profile in the PEG-rich phase constitutes a measure ofthe progress of the IgG diffusion into the PEG-rich phase. Addi-tionally, the average of the fluorescence intensity profile in eachphase is a measure of the average amount of IgG in each phase ata particular point along the length of the extraction microchan-nel. The antibody extraction yields to the PEG phase were alsocalculated based on fluorescence intensity. For this the average flu-orescence of the salt-rich phase was determined in the beginningand the end of the microchannel. Since the concentration in the

beginning of the microchannel is known, it is also possible to deter-mine the concentration of antibody in the salt-rich phase at the endof the microchannel. The difference in the antibody concentrationin the salt-rich phase between the beginning and the end of the

omatogr. A 1249 (2012) 1– 7 3

mriy

2

artdtrcemi6tarPwE

N

N

ddsimtcictccmmmtiPoScTcrccipIromcc

Table 1Schematic of the method used for the simulation of the diffusion progressionalong the length of the microchannel (L). The time (t) is calculated based on thelength (L) attributed to the row and the velocity of the fluids. D is the diffusioncoefficient, cj;i is the concentration value of the cell at row number j and columnnumber i. lw represents the distance between two adjacent columns, equivalentto x in Eqs. (1) and (2). In this case lw is always 5 �m, and td represents the timerequired for the fluid to move between two adjacent rows. The calculation of theequations used to calculate the antibody concentration for characteristic selectedmatrix positions highlighted with circles (circled in the table) is detailed at thebottom given below the table.

D.F.C. Silva et al. / J. Chr

icrochannel shows the amount of antibody extracted to the PEG-ich phase. Then estimating the percentage of PEG-rich phase thats lost in the middle channel allows the calculation of the recoveryield of IgG in the outlet channels.

.5. Simulation of the antibody partition and diffusion

The simulation of the effect of the partition and diffusion of thentibody was made by constructing a two dimensional matrix thatepresented the position along the length and across the width ofhe microchannel. In this matrix, each row, j, corresponds to a givenistance along the length of the microchannel and each column, i,o a given distance across the microchannel cross section. Each rowepresented a section 70 �m long along the microchannel and eacholumn a section 5 �m wide across the microchannel. Therefore,ach node in this simulation represented an area of 350 �m2 of theicrochannel. The width of each phase was chosen based on exper-

mental observations, i.e., 30 �m for the middle salt-rich phase and0 �m for each of the side walls PEG-rich phases. It was assumedhat, at the beginning of the microchannel, the concentration of thentibody was 0.2 mg/mL in the salt-rich phase and zero in the PEG-ich phases. The flux of antibody from the salt-rich phase to theEG-rich phases was calculated according to Fick’s first law. Eq. (1)as used to describe the antibody diffusion in the bulk phases and

q. (2) to describe the antibody partition at the interface:

= D(

�c

x

)(1)

= D(

c1 − Kp × c2

x

)(2)

In Eq. (1), N is the solute flux, D is the diffusion coefficient (theiffusion coefficient in the interface was assumed to be equal to theiffusion coefficient in the PEG-rich phase), x is the distance of diffu-ion, �c is the molar concentration difference between two pointsn space at x distance from each other. In Eq. (2), c1 is the solute

olar concentration in phase 1, c2 is the solute molar concentra-ion in phase 2 and Kp is the partition coefficient [22]. The partitionoefficient is a measure of the distribution of a solute between twomiscible phases and it is given by the quocient between the con-entration of the solute in the top phase (PEG-rich phase) and inhe bottom phase (salt-rich phase). For this reason in the model theoncentration in the top phase was always expressed as the con-entration in the bottom phase times the partition coefficient. Theolecular diffusion in the j axis is negligible compared to the move-ent of particles caused by the convection of solutions. Therefore,olecular diffusion in the j axis was not considered in the simula-

ion. At the beginning of the microchannel, row number j = 1,theres 0.2 g/L of antibody in the salt-rich phase and no antibody in theEG-rich phases: c1;1–12 = 0, c1;12–18 = 0.2 and c1;18–30 = 0. The fluxf antibody to the adjacent cells is calculated using Eqs. (1) and (2).ince the solutions are flowing, the distance along the channel alsoorresponds to the time available for diffusion between each rows.herefore the antibody concentration in each cell is affected by theells in the row before and not by cells in the same row. Table 1epresents a small part of the simulation with the equation used toalculate the antibody concentration of a few cells. The IgG diffusionoefficient in the salt-rich phase was determined by perform-ng an experiment in which a solution identical to the salt-richhase of the ATPS was pumped into all three inlets at 0.2 �L/min.

n one of the side inlets, IgG was mixed in the solution. Fluo-escence images from several points of the microchannels were

btained. The flux of antibody in the salt-rich phase along theicrochannel from one side of the microchannel to the other was

alculated from the variation in fluorescence and was used to cal-ulate the diffusion coefficient. The IgG diffusion coefficient in

the PEG-rich phase was calculated from the fluorescent micro-graphs from the experiments. The flux of IgG in the PEG-richphase was calculated from the variance in fluorescence in thePEG-rich phase only. Using the IgG flux in the PEG phase it was pos-sible to calculate the diffusion coeficient. The diffusion coefficientsof IgG thus determined were 20 × 10−11 m2/s in the salt-rich phaseand 5.8 × 10−11 m2/s in the PEG-rich phase; the partition coeffi-cient was 2.15 [23]. The linear velocity of each of the two phaseswas determined by dividing the volumetric flow rate by the crosssection that each phase occupied in the microchannel.

3. Results and discussion

3.1. Preliminary microfluidic structure design studies

The first structure fabricated was a two inlets and two outletsstraight microchannel with 200 �m of width, 20 �m of height and3.14 cm of length. The devices were made with PDMS, an elas-tomeric polymer commonly used in microfluidics with biologicalsamples in aqueous solutions. Micron-sized features can be repro-duced by PDMS using the replica molding technique. Furthermore(i) it is optically transparent above 280 nm; (ii) it can be deformedreversibly; (iii) it is cured at low temperatures; (iv) it can be sealedreversibly or irreversibly with itself and a variety of other materi-als; and (v) it is nontoxic, unreactive, impermeable to water andpermeable to air [24].

Experiments performed with this first microchannel deviceshowed that the two phases could coexist in a laminar flow, form-ing a stratified type flow. It was however observed that there wasalmost no variation in fluorescence of either phase indicating thatpartition was negligible probably due to the low residence time.

The microchannel was then redesigned to enhance partition.The microchannel extraction structure was designed as a serpen-tine (Fig. 1a) to increase the microchannel length while keeping

a small footprint. The width of the microchannel was reduced to150 �m to decrease the diffusion distance and a three inlets/threeoutlets structure were selected (Fig. 1b). By feeding the salt-richphase through the middle inlet, the average diffusion distance can

4 D.F.C. Silva et al. / J. Chromatogr. A 1249 (2012) 1– 7

Fig. 1. (a) Schematic of the microfluidic structure for the ATPS extraction of antibodies from a salt-rich to PEG-rich phases. The total microchannel length is 16.8 cm. Theblue lines are the microchannel walls, the red dotted lines indicate the locations for the interfaces between the salt-rich (center) and PEG-rich (sides) phases, The red arrowsindicate the expected direction of antibody diffusion. (b) Detail of the inlets. (c) Detail of the outlets. (For interpretation of the references to color in this figure legend, ther

bdeflwcro

3

fSicabmePpigtnapwispiltl

eader is referred to the web version of the article.)

e further reduced by half. This three inlets/three outlets designecreased, however, the stability of the stratified type flow and,xcept in specific flow rates, the flow degenerated into droplets,orming a segmented type flow. In addition, the three outlets designed to the loss of the PEG phase trough the center channel. This

as due to the constriction of the salt-rich phase caused by theurvature of the side outlets. To cope with this the structure wasedesigned to include a narrower central outlet and smoother sideutlets (see Fig. 1c).

.2. IgG partitioning in ATPS microfluidic device

The optimal ATPS composition obtained at the macroscale (5 g)or IgG purification in 15 mL test tubes [25], which is described inection 2.2, was used as the starting condition for the microflu-dic design and operating parameters. The system has an overallomposition of 7.04% PEG, 14.37% phosphate buffer at pH 7.0,nd 10% NaCl. A pure FITC-labeled IgG is used as the model anti-ody for separation. As the solutions progress along the extractionicrochannel, the fluorescence resulting from the tag antibody is

xpected to move from the salt-rich phase to the PEG-rich phase.artition of the FITC-labeled IgG from the salt- to the PEG-richhases and diffusion in the PEG-rich phase along the microchannel

s illustrated by the fluorescence micrographs in Fig. 2. The micro-raphs were obtained at the entrance (L = 0 cm), at L = 3 cm fromhe entrance, and at the end of the main separation microchan-el (L = 16.8 cm). The images show the diffusion of the FITC-labeledntibody from the middle, salt-rich, phase to the outer, PEG-rich,hases. In Fig. 3, the fluorescence intensity profiles across theidth of the channel are quantified at different distances from the

nlets. As observed qualitatively in Fig. 2, the fluorescence inten-ity decreases in the salt-rich phase and increases in the PEG-richhase along the length of the extraction microchannel, highlight-

ng the partitioning of IgG from the central salt-rich phase to theateral PEG-rich phases. The diffusion of the IgG progresses alonghe channel in the PEG-rich phase until the concentration of FITC-abeled antibody in the PEG-rich phase reaches a plateau and the

ratio of IgG concentration in the two phases is given by the valueof the partition coefficient.

3.3. Simulation and validation of IgG diffusion in ATPSmicrofluidic device

3.3.1. Effect of microchannel lengthIn order to allow for complete diffusion, the residence time of

the diffusing species in a microfluidic device must be sufficientlylong. There are two ways to change the residence time of IgG inthe extraction microchannel: either by changing the length of themicrochannel, or by changing the flow rates of the solutions. Twoextraction channels with different lengths were studied: one with10.45 cm and the other with 16.8 cm.

Fig. 4 shows the effect of microchannel length on the diffu-sion and partition of IgG from the salt- to the PEG-rich phase,using flow rates of 1.0 and 0.2 �L/min, respectively, and the cor-responding simulation as described in Section 2.5. Fig. 4a showsthe slopes of the tangents to the corresponding PEG-rich phasefluorescence intensity profile at different distances, L, from theinlets. These slopes can be used as indication of the driving force formass transfer. Initially, this slope increases rapidly because the IgGconcentrates at the PEG-rich phase near the interphase, causing astrong IgG concentration gradient in the PEG phases. This suggeststhat at the beginning of the extraction microchannel, the diffusionacross the interface occurs faster than the regular diffusion in a sin-gle phase. After reaching a peak, the slope of the tangent decreasesgradually to zero as the IgG diffuses in the PEG-rich phase, until theIgG concentration is constant along the PEG-rich phases. The exper-imental results agree closely with the simulation and show thatthe diffusion across the PEG-rich phase is complete at a distanceof L ≈ 16.5 cm. Fig. 4b shows the values of the average fluorescenceintensity across a phase, in both PEG-rich and salt-rich phases, for

different locations L along the length of the extraction microchan-nel. It can be observed in Fig. 4b that the concentration of IgG inthe salt-rich phase and in the PEG-rich phases reach stable valuesat L ∼ 10 cm from the inlets, indicating that the partition of the IgG

D.F.C. Silva et al. / J. Chromatogr. A 1249 (2012) 1– 7 5

Fig. 2. Optical (left) and fluorescence (right) micrographs taken at the beginning (L = 0), at L = 3 cm and at the end (L = 18.8 cm) of the extraction microchannel. The salt-richp middi is the

t

bbllbPPiP

3

wodtw

hase containing IgG with a concentration of 2 mg/mL was introduced through thentroduced through the side inlets with a volumetric pumping rate of 0,2 �L/min. L

he boundaries of the microchannel are indicated with a dashed white line.

etween salt-rich phase and PEG-rich phase is essentially completeefore the end of the microchannel, in agreement with the simu-

ation. Nevertheless, the extraction microchannel length must beonger to allow for the diffusion of IgG from the interphase to theulk PEG-rich phase and for the complete recovery of IgG in theEG-rich phase at the end of the channel. If the diffusion in theEG-rich phase is not complete, IgG will be concentrated near thenterphase on the PEG-rich side leading to lower recovery due toEG-rich phase losses to the central exiting channel.

.3.2. Effect of the flow rate of the two phasesThe effect of changing the flow rates of the different phases

as evaluated in a range where segmented flow is avoided. It was

bserved that, although a defined interface was maintained for theifferent flow rates tested, mass transfer of IgG from the salt- tohe PEG-rich phase was incomplete when PEG-rich phase flow rateas increased from 0.2 up to 0.8 �L/min and salt-rich flow rate was

le inlet with a volumetric pumping rate of 1 �L/min and the PEG-rich phases weredistance from the beginning of the microchannel. In the fluorescence micrographs,

increased from 1.0 up to 2.0 �L/min, as evidenced in Fig. 5. Theseflow rates were chosen to maintain the interphase stable and tominimize PEG losses in the outlet. Contrary to the lower flow rates,as the flow rates of both phases are increased, a driving force formass transfer from the salt-rich phase, albeit decreasing after theinitial peak, still persists at the end of the extraction microchannel(Fig. 5), indicating that the diffusion of the IgG within the PEG-richphase was not complete at the exit of the microchannel. As for theaverage concentration of IgG in the PEG- and salt-rich phases, theexperimental values seem to have reached a steady value but thesimulation suggest that, with higher flow rates, the steady value isnot reached, despite being very close. This suggests that the par-tition of IgG to the PEG-rich phase (Fig. 6) is not yet completed,

despite being very close to completion. Similar results are observedas the flow rates are further increased for the PEG- and salt-richphases to 0.8 and 2.0 �L/min, respectively. These results suggestthat, as expected, for the same microchannel length, a decreased

6 D.F.C. Silva et al. / J. Chromatogr. A 1249 (2012) 1– 7

Fig. 3. Fluorescence intensity profiles across the width of the microchannel struc-ture at L = 0, 2.2, and 16.8 cm from the inlet. The profiles show the evolution of theconcentration of the FITC-tagged IgG from being exclusively in the salt-rich phase atthe beginning of the microchannel (solid) to its partitioning to the PEG-rich phasealong the channel (dashed) to its diffusion within the PEG-rich phase at the end(dotted).

Fwct70s

ig. 4. (a) Fluorescence intensity slope for the PEG-rich phase along the channel,ith detail showing the tangent of the PEG-rich phase data, and (b) average IgG

oncentration for the PEG-rich phase for the serpentine like structures with extrac-ion microchannels 10.45 cm and 16.8 cm of length. The ATPS used was made with.04% of PEG, 14.37% of phosphate salt and 10% of NaCl. The volumetric flow was.2 �L/min for both PEG-rich phases and 1 �L/min for the salt-rich phase. All dataeries were normalized.

Fig. 5. Normalized data of the progression of the fluorescence intensity slope in thePEG-rich phase for the extraction microchannel with L = 16.8 cm. The volumetricpumping rates are indicated.

Fig. 6. Normalized data of the fluorescence intensity average in (a) the salt-rich phase and (b) the PEG-rich phases. The volumetric pumping rates areindicated.

omato

rfctdHptidmortrp

3

assp

caswl16Pemt

srcubd

4

waIploomoPrptbida

[[[[

[

[

[[

[

[[

[

[

[

[

D.F.C. Silva et al. / J. Chr

esidence time resulting from higher flow rates does not allow forull diffusion of the IgG within the PEG-rich bulk phase, and foromplete partition of the IgG from the salt to the PEG-rich phase,hough it is very close to completion, because of the shorter averageistance for diffusion in the salt phase and the partition coefficient.owever, the incomplete diffusion of IgG within the PEG-rich bulkhase will mean that there will be a greater concentration of IgG inhe area near the salt-rich phase. Since the loss of PEG-rich phasen the microchannel occurs next to the interface, this means that arop of the recovery yield will inevitably occur. For these experi-ents, the recovery yields calculated were of 54% using flow rates

f 1.0 �L/min and 0.2 �L/min for the salt-rich and PEG-rich phases,espectively; 41% using flow rates of 2.0 �L/min and 0.8 �L/min forhe salt-rich and PEG-rich phases, respectively; and 46% using flowates of 1.5 �L/min and 0.4 �L/min for the salt-rich and PEG-richhases respectively.

.4. Macro and microscale comparison

Extraction of IgG in ATPS at macroscale was already optimizednd reported by the authors in previous studies [26]. The authorshowed that the driving force for IgG partition was the ionictrength. By increasing the NaCl concentration from 0 to 15%, theartition of IgG shifted from the bottom to the top phase.

Macroscale partition assays were performed in 15 mL graduatedentrifuge tubes. All system components were thoroughly mixednd phases were left to separate at 25 ◦C. To ensure complete phaseeparation the systems were further centrifuged; phase volumesere determined and samples from the top and bottom were ana-

yzed by HPLC. The recovery yields obtained at the macroscale with0% of NaCl in the ATPS ranged from 25% to 60% for pH between

and 8, respectively [26]. However, if the molecular weight ofEG was increased from 3350 to 6000 Da, a decrease in the recov-ry yield to 11.3% was observed at pH 7, showing that the PEGolecular weight has an important influence in the partition of

he antibody in ATPS [25].The change from macro to microscale does not appear to have a

ignificant influence in the extraction yield as concluded from theesults of the previous section. This is an expected result since theomposition of the ATPS has not been greatly varied. However these of microscale devices has greatly reduced the diffusion timey increasing the interface surface area to phase volume ratio andecreasing the diffusion distance.

. Conclusions

A microfluidic device with a three inlets/three outlets designas fabricated using soft-lithography techniques based on PDMS

nd successfully demonstrated for the extraction of IgG with ATPS.t was possible to establish a stable, laminar flow of the two aqueoushases in the microfluidic network. Partition studies with FITC-

abelled IgG performed in PEG/phosphate ATPS showed a decreasef the fluorescence of the salt-rich phase, indicating partioningf the IgG–PEG-rich phase, as observed in macro-scale experi-ents. Microchannel length and flow rates for efficient diffusion

f IgG in PEG/phosphate ATPS were identified. The diffusion intoEG-rich phase was complete for a length of 16.8 cm with flowates of 0.2 �L/min and 1.0 �L/min, respectively in the PEG-richhase and salt-rich phase. Furthermore, it was demonstrated thathe reduction to the microscale does not greatly affect the anti-

ody extraction yield when compared with macroscale results, but

t does reduces the operation time. A mathematical model waseveloped to simulate the partition behavior of the FITC-labelledntibody, showing good agreement with experimental results.

[

[

gr. A 1249 (2012) 1– 7 7

The application of ATPS to the recovery of biomolecules stillrequires a somewhat trial-and-error approach since the behav-ior of a given biomolecule in the partition environment is unique.This is due to the chemical and physical interactions relatedto surface hydrophobicity, charge, and size that are involvedin the partitioning process and that gain particular importancein a microfluidic environment. Thus, the use ATPS in an inte-grated microfluidic system for microscale protein purificationusing an automated high-throughput platform for cell separation,biopharmaceuticals purification and analysis could potentially pro-vide a means to speed up the process optimization of ATPS indsp.

The ATPS chip developed for IgG extraction has great poten-tial in biotechnology as an example of an application of microscalelab-on-a-chip devices as a generic technological platform capa-ble of monitoring and optimizing the pharmaceutical productionprocess, and, in particular, provides a demonstration of minia-turized dsp in the case of protein drugs such as therapeuticantibodies.

Acknowledgements

The authors acknowledge funding from FCT through theAssociated Laboratories IN – Institute of Nanoscience andNanotechnology and IBB – Institute of Biotechnology and Bioengi-neering. A. Azevedo and P. Fernandes acknowledge FCT for fundingunder program Ciência 2007.

References

[1] M.C. Via, Monoclonal antibodies: pipeline analysis and competitive assessment,in: Insight Pharma Reports, Cambridge Healthtech Institute, Needham, MA,2006.

[2] A.M. Azevedo, A.G. Gomes, P.A.J. Rosa, I.F. Ferreira, A.M.M.O. Pisco, M.R. Aires-Barros, Sep. Purif. Tecnhol. 65 (2009) 14.

[3] A.M. Azevedo, P.A.J. Rosa, I.F. Ferreira, M.R. Aires-Barros, Trends Biotechnol. 27(2009) 240.

[4] P.A.J. Rosa, A.M. Azevedo, M.R. Aires-Barros, J. Chromatogr. A 1141 (2007) 50.[5] L.N. Mao, J.K. Rogers, M. Westoby, L. Conley, J. Pieracci, Biotechnol. Prog. 26

(2010) 1662.[6] T. Chován, A. Guttman, Trends Biotechnol. 20 (2002) 116.[7] P. Fernandes, F. Carvalho, M.P.C. Marques, Recent Patents on Biotechnology 5

(2011) 160.[8] G. Whitesides, Nature 442 (2006) 368.[9] M.S. Thomsen, B. Nidetzky, Eng. Life Sci. 8 (1) (2008) 40.10] R.L. Hartman, K.F. Jensen, Lab Chip 9 (2009) 2495.11] N.J. Titchener-Hooker, P. Dunnill, M. Hoare, Biotechnol. Bioeng. 100 (2008) 473.12] S. Hardt, T. Hahn, Lab Chip 12 (2012) 434.13] K. Nam, W. Chang, H. Hong, S. Lim, D. Kim, Y. Koo, Biomed. Microdev. 7 (3)

(2005) 189.14] M. Yamada, V. Kasim, M. Nakashima, J. Edahiro, M. Seki, Biotechnol. Bioeng. 88

(2004) 489.15] K.-H. Nam, W.-J. Chang, H. Hong, S.-M. Lim, D. Lim, Y.-M. Koo, Biomed. Microdev.

7 (2005) 189.16] J. Soohoo, G. Walker, Biomed. Microdev. 11 (2009) 323.17] M. Tsukamoto, S. Taira, S. Yamamura, Y. Morita, N. Nagatani, Analyst 134 (2009)

1994.18] Y.S. Huh, S.J. Jeon, E.Z. Lee, H.S. Park, W.H. Hong, Korean J. Chem. Eng. 28 (2011)

633.19] R.J. Meagher, Y.K. Light, A.K. Singh, Lab Chip 8 (2008) 527.20] R. Hu, X. Feng, P. Chen, M. Fu, H. Chen, L. Guo, B.-F. Liu, J. Chromatogr. A 1218

(2011) 171.21] Y.S. Huh, C.-M. Jeong, H.N. Chang, S.Y. Lee, W.H. Hong, T.J. Park, Biomicrofluidics

4 (2010) 14103.22] E.L. Cussler, Diffusion Mass Transfer in Fluid Systems, Cambridge University

Press, Cambridge, 2009.23] P.A.J. Rosa, A.M. Azevedo, I.F. Ferreira, J. de Vries, R. Korporaal, H.J. Verhoef, T.J.

Visser, M.R. Aires-Barros, J. Chromatogr. A 1162 (2007) 103.24] J.C. McDonald, D.C. Duffy, J.R. Anderson, D.T. Chiu, H. Wu, O.J.A. Schueller, G.M.

Whitesides, Electrophoresis 21 (2000) 27.25] A.M. Azevedo, P.A.J. Rosa, I.F. Ferreira, M.R. Aires-Barros, J. Biotechnol. 132

(2007) 2011.26] P.A.J. Rosa, A.M. Azevedo, S. Sommerfeld, M. Mutter, M.R. Aires-Barros, W.

Bäcker, J. Biotechnol. 139 (2009) 306.