Miniaturizing free-flow electrophoresis - A critical review

977 Electrophoresis 2008, 29, 977–993 Dietrich Kohlheyer Jan C. T. Eijkel Albert van den Berg Richard B. M. Schasfoort MESA1 Institute for Nanotechnology, University of Twente, Enschede, The Netherlands Received September 28, 2007 Revised November 23, 2007 Accepted November 24, 2007 Review Miniaturizing free-flow electrophoresis – a critical review Free-flow electrophoresis (FFE) separation methods have been developed and investigated for around 50 years and have been applied not only to many types of analytes for various biomedical applications, but also for the separation of inorganic and organic substances. Its continuous sample preparation and mild separation conditions make it also interesting for online monitoring and detection applications. Since 1994 several microfluidic, miniaturized FFE devices were developed and experimentally characterized. In contrast to their large-scale counterparts microfluidic FFE (m-FFE) devices offer new possibilities due to the very rapid separations within several seconds or below and the requirement for sample volumes in the microliter range. Eventually, these m-FFE systems might find application in so-called lab-on-a-chip devices for real-time monitoring and separation applications. This review gives detailed information on the results so far published on m-FFE chips, comprising its four main modes, namely free-flow zone electrophoresis (FFZE), free-flow IEF (FFIEF), free-flow ITP (FFITP), and free-flow field-step electrophoresis (FFFSE). The principles of the different FFE modes and the basic underlying theory are given and discussed with special emphasis on miniaturization. Different designs as well as fabrication methods and applied materials are discussed and evaluated. Furthermore, the separation results shown indicate that similar separation quality with respect to conventional FFE systems, as defined by the resolution and peak capacity, can be achieved with m-FFE separations when applying much lower electrical voltages. Furthermore, innovations still occur and several approaches for hyphenated, more integrated systems have been proposed so far, some of which are discussed here. This review is intended as an introduction and early compendium for research and development within this field. Keywords: Free-flow electrophoresis / Free-flow isoelectric focusing / Free-flow isotachophoresis / Microfluidic DOI 10.1002/elps.200700725 1 Introduction Since the introduction of free-flow electrophoresis (FFE) in the 1960s [1], this separation method has found a permanent position among analytical and preparative methods in biochemistry and chemistry for the separation of, e.g., cells, organelles, peptides, proteins, inorganic, and organic compounds [2–5]. In FFE, analytes are separated continuously in Correspondence: Dietrich Kohlheyer, MESA1 Institute for Nanotechnology, University of Twente, P. O. Box 217, NL-7500AE Enschede, The Netherlands E-mail: [email protected] Fax: 131-53-489-3595 Abbreviations: ASS, acetylsalicylic acid; FFE, free-flow electrophoresis; ì-FFE, microfluidic free-flow electrophoresis; FFFSE, free-flow field step electrophoresis; FFIEF, free-flow IEF; FFITP, free-flow ITP; FFZE, free-flow zone electrophoresis; LOC, lab-ona-chip; SPR, surface plasmon resonance  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim an electrical field applied perpendicular to a thin pressuredriven carrier electrolyte flow between two insulating plates, as shown in Fig. 1. The sample mixture is injected into the carrier electrolyte flow and with increasing residence time the differently charged components split up into diverging lanes which can be collected at various device outlets. Commercially available large-scale FFE systems (e.g., BD Diagnostics, Germany) were originally developed as standalone sample clean-up devices, e.g., as a prefractionation step prior to 2-DE or other detection and identification methods. Size-reduced mini-FFE systems were developed later and were coupled to MS and LC enabling continuous separation with online detection possibilities [6, 7]. This combination of FFE as a continuous separation method with online detection would allow for real-time monitoring of analytes, such as patients samples, reaction products, and more. These sizereduced systems are still complicated and slow in operation and downscaling would be very promising. Eventually such microfluidic FFE (m-FFE) systems might find application in small portable devices and point-of-care (POC) tools [8]. With www.electrophoresis-journal.com 978 D. Kohlheyer et al. Figure 1. Illustrative principle of FFZE and main components. the upcoming trend of miniaturization and the development of lab-on-a-chip (LOC) systems [9], also microfluidic, chipbased FFE systems have been developed and demonstrated. The miniaturization of FFE implies several advantages especially considering sample volume and separation speed. In contrast to the tens of milliliters of sample consumed by conventional large-scale FFE devices, m-FFE systems require only tens of nanoliters up to hundreds of microliters of sample. This is especially interesting in clinical analysis where often only low sample volumes are available. Furthermore, instead of residence times of up to tens of minutes, mFFE devices separate within several seconds. Scaling laws predict a 100-fold increase in speed for a ten-fold (linear) downscaling of an FFE experiment [10]. Such short analysis times would be very beneficial in POC devices. The shallower separation chambers of several micrometers enable good heat dissipation allowing higher electrical field strengths necessary for rapid separations. According to theory, the quality of the separation, as defined by the resolution does Electrophoresis 2008, 29, 977–993 not decline with reduced separation geometry [11], as shown for CIEF. Due to the mentioned advantages of miniaturizing FFE and due to the availability of new fabrication technologies, the interest in this field has recently grown rapidly indicated by the relatively large number of m-FFE publications in 2005 and in 2006. This review has its focus mainly on m-FFE systems, while for a more general overview the reader is referred to the recent review by Pamme [12] on continuous flow separations in microfluidic devices. This review gives detailed information about the recent developments in the field of m-FFE systems including various modes of FFE. The different technological solutions and developments are critically analyzed and compared with respect to design, fabrication methods, and separation quality. Furthermore, detection methods and hyphenation aspects are discussed. In this paper, we will first describe the common separation modes of FFE and discuss basics of the related theory. In Section 3 we will describe different technological approaches that have been applied for FFE microchips. In Section 4 recent separation results are shown and hyphenation and detection aspects are discussed in Section 5. 2 FFE modes and related theory Most of the modes of standard CE [13] can be applied in FFE as well. The four common ones are free-flow zone electrophoresis (FFZE), free-flow IEF (FFIEF), free-flow ITP (FFITP), and free-flow field step electrophoresis (FFFSE) [14, 15]. 2.1 FFZE FFZE implies the usage of a carrier electrolyte flow of constant composition of pH and electrical conductivity in which components are separated according to their mobility (determined by the charge-to-size ratio; Fig. 2a). It has been found Figure 2. Modes of FFE: (a) FFZE, (b) FFIEF, (c) FFITP, and (d) FFFSE.  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com Microfluidics and Miniaturization Electrophoresis 2008, 29, 977–993 that there is a direct correlation between CZE as used for analytical applications and FFZE as used for preparative ones. The theory of this correlation can be found in refs. [16, 17]. In FFZE the analyte to be separated is deflected linearly under a constant angle determined by the electrical field strength, the analyte mobility, and the flow velocity. The migration distance d of the analyte moving through the electrical field is given by d ¼ mp Et (1) where mp is the apparent electrophoretic mobility, E is the electrical field strength, and t is the residence time of molecules in the separation chamber [18]. The goal of FFZE is to achieve high resolution, which is reached by analytes migrating in narrow zones with sharp boundaries. However, several phenomena negatively influence the separation quality. The following sources of band broadening are common in FFZE: the width of the injected sample stream leads to a band SD sINJ, and further types of band broadening are diffusional broadening (sD), hydrodynamic broadening (sHD), electrodynamic broadening (sED), Joule heating (sJH), and electromigration dispersion (sEMD) [4, 19]. Not all sources of band broadening are discussed here and the reader is referred to the literature for more details. The variance s2T of a separated analyte band is then given by the sum of the variances of all broadening contributing factors: s2T ¼ s2INJ þ s2D þ s2HD þ s2ED þ s2JH þ s2EMD (2) The variance due to the sample injection width wi is usually expressed by [20] s2INJ ¼ w2i 12 (3) Reducing the sample injection width is often easier in microfluidic systems than in larger fluidic systems for example by the precise control of the neighboring laminar flow streams causing hydrodynamic focusing of the sample as for example shown in ref. [21]. The variance caused by lateral diffusion is directly related to the residence time t of the analyte in the separation chamber, and can be expressed by s2D ¼ 2Dt (4) where D is the analyte diffusion coefficient [22]. Short residence times, as usually involved in m-FFE systems therefore reduce this effect. The parabolic flow profile, which is caused by the pressure-driven flow leads to unequal velocity regions inside the separation chamber. The analytes flowing near to the top and the bottom plate spend more time in the electrical field, and thus are deflected more than analytes flowing near the  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 979 chamber center. This effect is called hydrodynamic broadening and leads to a crescent-shaped deformation of the sample. The variance due to hydrodynamic band broadening equals s2HD ¼ h2 t 2 2 E mp 105D (5) where h is the chamber height [22]. Equation (5) shows that a shallower separation chamber, as used in m-FFE devices does strongly reduce the effect of hydrodynamic band broadening. The applied electrical field together with an electrical double layer present at the separation chamber surfaces leads to EOF inside the separation chamber. Normally, in an open channel EOF results in a plug-shaped flow profile from the anode towards the cathode. However, in many FFE systems the EOF generated at the upper and lower cover plates results in a hydrodynamic counterflow in the center of the chamber due to a high transport resistance in the direction of the flow (e.g., due to incorporated membranes). This hydrodynamic flow perpendicular to the separation axis often results in crescent-shaped sample deformation comparable to that caused by hydrodynamic band broadening. This effect is generally referred to as electrodynamic band broadening. As discussed later, not all m-FFE systems are subject to this type of band broadening, therefore we will not discuss the theory in more detail here and one is referred to ref. [3]. As the electrical current heats the separation medium (Joule heating) inside the FFE device a temperature gradient is established between the two cover plates with the maximum in the center. This temperature change leads to a viscosity decrease in the center locally affecting the analyte mobility, and thus leading to band broadening. In contrast to actively cooled conventional FFE systems the shallower separation region of m-FFE systems favors fast heat dissipation reducing this type of sample distortion and usually also much lower electrical currents are involved. Experiments have confirmed that m-FFE separations could be performed without noticeable influence of Joule heating applying electrical field strengths of up to 60 V/mm [22, 23] with an exception given by ref. [24]. Zones of different electrical conductivity in sample and carrier electrolyte will lead to electromigration dispersion [4, 19]. This type of distortion results from different migration velocities of the analytes in zones having different electrical field strengths. This effect can be minimized by choosing appropriate buffer systems with similar conductivities for sample and carrier electrolyte. An interesting approach was shown by Fonslow and Bowser [22] in which the authors derived an analog of the van Deemter equation describing the separation in FFZE. They showed that linear velocity, electric field, and migration distance must all be considered to optimize bandwidth and resolution. www.electrophoresis-journal.com 980 D. Kohlheyer et al. Electrophoresis 2008, 29, 977–993 The separation resolution Rs of two adjacent peaks is defined by Rs ¼ d1
d2 2ðsT1 þ sT2 Þ (6) Assuming that the band broadening is dominated by diffusion and inserting Eqs. (1) and (4) one can derive (assuming equal diffusion coefficients) Rs ¼ ðmt1
mt2 ÞEt pffiffiffiffiffiffiffi 4 2Dt (7) Substituting the residence time t with L/v, where L is the length of the separation chamber in the flow direction and v is the linear flow velocity, and substituting E with Veff/W, where Veff is the effective separation voltage (the effective voltage utilized for separation) and W is the separation chamber width, one obtains Rs ¼ rffiffiffiffiffiffiffi ðmt1
mt2 Þ L pffiffiffiffiffiffi 2 4 2D |fflffl{W zfflffl} |fflfflfflfflfflffl{zfflfflfflfflfflffl} Analytes V eff pffiffi v |{z} (8) Device Tunable geometry parameters This equation shows that when scaling down FFZE devices, the separation resolution is independent of the device size (assuming a constant device aspect ratio L/W) and is dependent only on the applied separation voltage and the linear flow velocity. In the derivation of this equation many assumptions were made and it is just intended to guide the reader and to demonstrate that resolution eventually is independent of the actual device size and does not have to suffer from downscaling. Of course at the same time the separation time L/v decreases proportionally with downscaling. 2.2 FFIEF In FFIEF, the used carrier electrolyte is composed of a mixture of ampholytes, which lead in the presence of the applied voltage and a natural pH gradient (mostly achieved by low and high pH anodic and cathodic electrode electrolytes), to the formation of a linear pH-gradient perpendicular to the flow direction. Sample components migrate within this pH gradient due to the electrical field until they reach the point where their pI is equal to the local pH value of the buffer, where they become neutrally charged and focus (Fig. 2b). Unlike the linear separation technique FFZE where the SD of the band increases as the separation continues, the SD of a sample fraction bandwidth in FFIEF stays constant once equilibrium is reached. Band broadening, e.g. due to diffusion, is continuously counteracted since species leaving the equilibrium zone become charged again and migrate back to the location of zero charge [25]. Therefore, the sample injection width and band broadening factors as discussed for FFZE play a minor role in FFIEF. However, the low solubility of analytes at their pI and therefore precipitation often leads to nonideal and distorted focusing.  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim The following differential equation, valid under steady state conditions, describes the equilibrium conditions between simultaneous electrophoretic and diffusional mass transport during IEF, dðCmEÞ d dC ¼ D dx dx dx (9) where C is the analyte concentration at position x in the separation channel, m is its mobility at that point, E is the electrical field strength, and D is the diffusion coefficient [26]. Useful parameter to express the quality of an equilibrium gradient separation system such as an IEF system, include the SD of the peak width s (Eq. 10), minimum pI value that can be resolved D(pI)min (Eq. 12), and peak capacity n (Eq. 14). The solution to the differential Eq. (9) at final steady state gives a Gaussian concentration distribution, with an SD expressed by sffiffiffiffiffi D (10) s¼ pE where D is the diffusion coefficient, E is the electrical field strength, and p¼
dm dðpHÞ dðpHÞ dx (11) where d(pH)/dx is the pH gradient and dm/d(pH) the mobility slope of the analyte [26]. A way to express the resolving power of IEF is given by Eq. (12) which expresses the minimum difference in pI of two species that still can be separated (peak distance 3s) [26]. sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi DðdðpHÞ=dxÞ DðpIÞmin ¼ 3 (12) E ð
dm=dðpHÞÞ By replacing d(pH)/dx with DpH/L, where L is the length of the pH gradient and DpH is the total difference of the applied pH gradient, and replacing E with Veff/L, where Veff is the voltage drop utilized for separation, one can rewrite Eq. (12) to sffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffi DpH D (13) DðpIÞmin ¼ 3 V eff
dm |fflfflffl{zfflfflffl} |fflfflfflffl{zffldpH fflfflffl} Device Analyte This approach has been utilized for CIEF by Das and Fan [11]. One can see from Eq. (13) that only DpH and Veff are device-related parameters while the others depend on the analyte. It thus becomes clear that the resolution D(pI)min is independent of the device dimensions but increases with total applied voltage, a result similar to that for FFZE. Assuming ideal separation conditions, www.electrophoresis-journal.com Microfluidics and Miniaturization Electrophoresis 2008, 29, 977–993 981 such as no Joule heating one can therefore conclude that microfluidic FFIEF systems can achieve comparable resolution to their larger counterparts. Obviously, a good resolution is favored by a high separation voltage and a narrow pH gradient. The separation time, however, benefits from downscaling, since it decreases linearly with the chamber length. A common method to express the number of peaks that can be resolved is the peak capacity n, defined by n¼ L 4 s (14) where L is the total length of the pH gradient [25]. Also the peak capacity is theoretically independent of the device dimensions as it can be rewritten to n¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dm=dpH DpHV eff |fflfflfflfflfflfflffl{zfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflffl {16D zfflfflfflfflfflfflfflffl } Device (15) Analytes 2.3 FFITP In FFITP, the sample is introduced between leading and terminating buffers respectively having the highest and the lowest mobility of their ions with respect to the mobility of the analytes. During FFITP separation the analytes form adjacent regions according to their descending electrophoretic mobility (Fig. 2c). For more theory about ITP the reader is referred to refs. [27–29]. 2.4 FFFSE A fourth mode is FFFSE in which an electrical field step gradient is built up by introducing a less conductive buffer in the center of the separation chamber, enclosed by more conductive buffers on the sides. The analytes to be separated move relatively fast through the center zone with high electrical field strength until they reach the boundary with the high ionic buffer concentration and lower electrical field strength. This field step results in a drastic reduction of the analyte electrophoretic velocity and the components therefore become concentrated and focus [3, 30]. 3 Device technology A conventional large-scale FFE separation device, as illustrated in Fig. 3, consists of two insulating plates (often glass or acrylic glass (polymethyl methacrylate, PMMA)) which are separated by a membrane spacer (e.g., cellulose nitrate [7]) of up to several hundred micrometers in thickness. This spacer defines the separation chamber height and is conductive for an electrical current but due to its internal structure acts as a confinement for the pressure-driven fluid inside the separation chamber. In this configuration, it spatially defines three  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Figure 3. Illustration of a disassembled conventional FFE separation unit: 1, top plate; 2, bottom plate; 3, cooling plate; 4, electrode; 5, membrane spacer; 6, electrode electrolyte inlet; 7, electrode electrolyte outlet; 8, electrode connection; 9, sample 1 carrier electrolyte inlet; and 10, fraction collector. regions: the separation region, the anode bed, and the cathode bed. Electrodes are placed outside the separation region avoiding the disturbance of generated gas bubbles and minimizing the interference of chemical side products migrating into the separation region. Continuous electrolyte flow inside the electrode beds ensure stable electrical properties by flushing away generated oxygen, hydrogen bubbles, and chemical side products. Usually the device is placed on an actively cooled plate in order to efficiently remove heat generated by the electrical current. To some extent this principle is applicable for smaller FFE systems as well, using traditional micromachining techniques such as injection molding and milling [6, 7, 31]. However, when the lateral feature size decreases to millimeters and the separation chamber heights decrease to tens of micrometers the fabrication of the necessary thin cellulose membranes sheets and their manual handling including device assembly is not practicable anymore. Applying modern cleanroom fabrication technologies new manufacturing principles for FFE microdevices have therefore been developed whereby the developments mainly focused on alternatives to conventional membranes. Generally, three aspects are thereby of main importance: (i) efficient removal of gas and chemical side products formed during electrolysis with no crosscontamination between the separation chamber and the side beds; (ii) a stable electrical field over the separation chamber, and (iii) good voltage efficiency. The voltage efficiency ZV is defined by ZV ¼ V eff V total (16) where Vtotal is the total applied voltage and Veff is the voltage effectively utilized for separation. In conventional large-scale FFE devices the width of the membranes is much smaller www.electrophoresis-journal.com 982 D. Kohlheyer et al. than the actual width of the separation chamber, resulting in a negligible voltage loss across the membranes when the separation voltage is applied. In contrast to this, in m-FFE chips the membrane-replacing structures often have widths of the same size as the separation chamber causing a large extent of voltage loss, resulting in low voltage efficiency. Unfortunately, reducing the width of the membranes or similar structures to achieve low electrical resistance normally leads to structures with low hydrodynamic resistance as well, possibly causing fluid leakage towards the electrode side beds or vice versa. In general, as shown in Fig. 4, four different technological approaches were demonstrated in m-FFE systems trying to fulfill the mentioned criteria: (i) Electrodes were placed in open side reservoirs with membrane-like structures of high hydrodynamic resistance isolating the separation chamber. (ii) Electrodes were placed in closed side beds with a continuous electrode flow, with a membrane equivalent or different structures to shield the separation chamber. (iii) Electrodes were integrated into the separation chamber with no additional structures. (iv) The electrodes were electrically and mechanically isolated from the separation chamber. Table 1 briefly compares the papers using these different approaches which are discussed in the following sections. Comparisons should be taken with care, since many aspects have to be considered that are not always simple to compare in the different approaches. 3.1 Open electrode side beds with membrane equivalent The implementation of open electrode side beds to place the electrodes allows for an easy ventilation of gas products formed during electrolysis. This design, however, results in a Figure 4. Illustration of various design categories applied for mFFE systems. (a) Open electrode beds with membrane-like structure or structures of a similar function; (b) closed electrode beds with membrane-like structures; (c) electrodes inside the separation chamber; and (b) mechanically and electrically insulated separation chamber.  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Electrophoresis 2008, 29, 977–993 pressure gradient from the separation chamber towards the electrode side beds which can cause fluid leakage. This fluid leakage has to be compensated by proper shielding of the separation chamber with membrane equivalent structures with high hydrodynamic resistance. Since there is no need for a precise control of electrolyte fluid streams the working setup can be simplified. However, for longer separation times a continuous refreshment of the electrode solutions should be implemented, to avoid electrical conductivity changes. Zhang and Manz [32] developed an FFE chip using PDMS and Pyrex glass. In their design, they used an array of narrow side channels connecting the separation chamber with the electrode side beds. These side channels were of high hydrodynamic resistance acting as a membrane equivalent structure. Gas bubbles were efficiently hindered from entering the separation chamber and eventually left through the open electrode side beds. The design allowed for relatively fast flow rates and furthermore, the usage of glass and PDMS made the application of high voltages possible. This resulted in an FFZE separation of fluorescent components and labeled amino acids with residence times below 2 s. The same chip design was used later to demonstrate rapid FFIEF by Xu et al. [33]. They showed the 400-fold focusing of Angiotensin II at its pI within 430 ms. Despite the very short separation times, more than 90% of the applied voltage was lost across the side channels. The high hydrodynamic resistance of the side channels, necessary for rapid operation, in a trade off therefore caused a tremendous increase in the electrical resistance of the side channels, making the device less efficient in terms of applied voltage. Applying FFZE in this device, the low voltage efficiency turned out not to be a limiting factor, especially when considering it as a proof-of-principle device. However, in FFIEF a large part of the pH gradient was established inside the side channel membranes and not in the separation chamber due to carrier ampholytes migrating into the side channels. Therefore, only the part of the pH gradient which formed inside the separation chamber was available for IEF. This effect significantly limited the separation capacity. Janasek et al. [34] applied a comparable PDMS chip to demonstrate for the first time FFITP of fluorescein and Eosin G. In conventional gel electrophoresis, a dense swollen hydrogel made from cross-linked acrylamide monomer is used as a separation matrix. Similar gels were used in microfluidic systems as so called ion-bridges or salt-bridges (e.g., used in electroosmotic pumps [35]) making it also a suitable membrane material for m-FFE. This principle was applied by several groups in miniaturized FFE devices to form conductive membranes. Kohlheyer et al. [21] presented a m-FFE glass chip with photopolymerized acrylamide membranes. This device was fabricated by using two wafers of Borofloat glass, one containing the 15 mm high separation chamber as well as inlet and outlets, as shown in Fig. 5. The glass wafers were bonded by direct glass wafer bonding with no need for an additional www.electrophoresis-journal.com Microfluidics and Miniaturization Electrophoresis 2008, 29, 977–993 983 Table 1. Rough comparisons of different FFE design approaches Category (related to Fig. 4) Author Year of publication Principle a Xu, Zhang and Janasek [32–34] Kohlheyer et al. [21, 23] Albrecht and Jensen [24] de Jesus et al. [37] 2003, 2006 Side channels 4.07 2006, 2007 1994, 1996 Acrylamide membranes Acrylamide membranes Acrylamide membranes Side channels 2005 2006 a a a b 2006 2006 c Raymond et al. [18, 20] Fonslow and Bowser [36] Fonslow et al. [22, 38] Kobayashi et al. [39] Lu et al. [40] c Song et al. [41] 2006 d Janasek et al. [42] 2006 b b b 2003 2004 Chamber width (mm) Removal of bubbles and side products Stable electrical field, separation Comments 4.5 o o 3.5 63a) 1 1 1 15a) 1 1 Reduced pH gradient in FFIEF Low mechanical stability Joule heating 18 69 1 1 10 60a) o 2 Side channels 10 50 2 2 Deeper electrode beds Shallow side banks Integrated electrodes 10 91 1 1 56.5 95a) 1 1 n.a. 1 Diffusion potential Electrostatic induction 1 Voltage efficiency (%) 100 0.05 n.a. n.a. 1 4.4 50 n.a. 1 Laborious electrode connections Low break-down voltage Bubbles distorted separation Precise flow control required Relatively large, ø 100 mm wafer sized Only low diffusive analytes, long residence times Ampholyte less IEF Still under investigation a) Estimated by the author; (2, inefficient; o, fair; 1, efficient). Figure 5. Illustration of the m-FFE device during FFZE: 1, glass plates; 2, separation chamber; 3, sample inlet; 4, sheath flow inlet; 5, electrode wire; 6, fraction outlet; 7, acrylamide membrane. Reproduced with permission of The Royal Society of Chemistry from [21]. intermediate silicon layer as used for example by Fonslow and Bowser [36]. The electrodes were placed in open reservoirs to allow the ventilation of gas. Depending on the used  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim separation chamber width, between 60 and 40% of the applied voltage was utilized across the separation region. The stability of the membranes turned out to be the limiting factor. At elevated fluid velocities and especially during FFIEF the membranes eventually broke. An improved version of this FFE chip was developed and used to achieve an increased separation resolution in FFIEF [23]. Albrecht and Jensen [24] demonstrated a FFIEF PDMS device with integrated functionalized photopolymerized acrylamide membranes, as shown in Fig. 6. These membranes of low and high pH respectively provided buffering capacity during IEF to confine a pH gradient between pH 3 and 9 with no need for additional sheath flow streams as used, e.g., by Kohlheyer et al. [21]. Furthermore, an integrated cooling system was used to reduce the negative effect of Joule heating at high electrical field strengths. Considering the width of the membranes, the utilized voltage drop across the separation channel is estimated by us to be around 15% of the total applied voltage. Albrecht and Jensen [24] did not report on low mechanical stability of their membranes. de Jesus et al. [37] fabricated a low-cost m-FFE chip using laser printing toner as a structural material on a glass substrate. Furthermore, they used a polymerized hydrogel to fill the electrode reservoirs shielding the www.electrophoresis-journal.com 984 D. Kohlheyer et al. Electrophoresis 2008, 29, 977–993 Figure 7. Principal FFE layout used by Raymond et al. [18]. A highly dense side channel array acted as membrane between the separation chamber and electrode beds. Reproduced with permission from ref. [18], copyright American Chemical Society. Figure 6. Top view (a) shows the PDMS device with the sample channel bordered by left and right porous material regions (crosshatched areas) and anode and cathode, respectively. Silicone sealant (solid array) is used to form the reservoirs for the anolyte and catholyte buffers, as well as to hold the platinum electrodes in place. Reproduced with permission from ref. [24], copyright Wiley-VCH Verlag GmbH & Co. KGaA. separation chamber. The electrodes were placed inside two electrolyte filled syringes connected to the side openings via polyethylene tubing. Gas formation inside the syringes did not enter the side openings nor caused disturbance of the separation. de Jesus et al. achieved a similar voltage efficiency as Kohlheyer et al. [21]. 3.2 Closed electrode side channels The usage of closed electrode channels requires additional flow streams ensuring an effective removal of produced gas bubbles and chemical side products, and to ensure stable electrical properties. Structures or obstacles also have to be implemented to avoid gas bubbles from entering the separation chamber. In addition, an exact flow balancing of the fluid flow streams in the electrode and separation channels becomes important to avoid cross-contamination of fluids between the electrode channels and the separation chamber. Raymond et al. [18] were the first to develop a m-FFE chip, and it operated with closed side channels. This FFE microchip incorporated a separation region, inlet and outlet channels, electrode beds, and a dense array of microchannels which acted as a membrane, separating the electrodes from the actual separation chamber, as shown in Fig. 7. This liquid-filled channel array formed a high hydrodynamic resistance for the pressure-driven fluid but could conduct the electrical current. This was a very elegant solution, since all channel features could be fabricated in one standard siliconetching step. The chip was used to separate fluorescent markers as well as amino acids during FFZE.  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim The same device was used two years later to separate FITC-labeled proteins and amino acids with FFZE and was further characterized [20]. The major drawback of this system was the low breakdown voltage of 100–200 V due to the use of silicon as chip material, limiting the device in its separation power. Due to low electrical field strength, long residence times of up to 1 min were required to achieve good separations. Although the side channels formed a high hydrodynamic resistance, fluid flow from the side beds into the separation chamber, negatively influencing the separation, was still observed. Considering the width of the separation chamber (10 mm) and of the side channels arrays (261 mm), the estimated actual voltage efficiency was around 80%. Fonslow and Bowser [36] presented an FFE chip fabricated from two glass plates with an intermediate layer of amorphous silicon (a-Si) for anodic bonding. This device contained closed electrode channels, in which gas bubbles were flushed out by a pressure-driven flow passing the integrated microfabricated gold electrodes. These electrode side channels were separated by connecting side channel arrays comparable to those shown in Fig. 7. Due to the side channel arrays and its electrical resistance, 50% of the applied voltage was utilized across the separation region in this device. Although the device could withstand higher voltages and fluid pressure, the successful separation of several fluorescent dyes was eventually distorted by gas bubbles inside the side channels, since they could not be removed efficiently anymore. Fonslow et al. [38] subsequently improved their FFE concept as mentioned above and presented a device with a shallow separation region (20 mm deep) and roughly four times deeper closed electrode beds, completely avoiding side channel arrays or membrane equivalents. The flow rate through the electrode channels was now significantly higher, effectively removing electrolysis products without disrupting the flow pattern in the shallow separation channel as shown in Fig. 8 [38]. As a complicating consequence, this approach www.electrophoresis-journal.com Electrophoresis 2008, 29, 977–993 Figure 8. (a) A top view of the m-FFE mask of Fonslow et al. [38] with the following features: 1, separation buffer inlet; 2, sample inlet; 3, electrode buffer inlets; 4, Au electrodes in electrode channels; 5, separation channel; 6, electrode buffer outlets; and 7, separation buffer outlet. (b) A side view of the electrode and separation channels etched into the bottom glass wafer and bonded to the top glass wafer. Reproduced with permission from ref. [38], copyright American Chemical Society. required a precise control over the flow rates and more equipment. In the new design, 91% of the applied voltage was utilized across the separation channel where it actually impacts the separation. Although this multiple-depth m-FFE device required more fabrication steps, it could be operated continuously at electric fields (in the separation channel) as high as 589 V/cm, a four-fold improvement over their previous design. They now found that neither Joule heating, nor electrolysis product formation, was a limiting factor when applying high separation potentials. The device was applied to separate several fluorescent standards with FFZE. The chip was additionally used for a more detailed study of FFZE investigating several parameters affecting separation resolution in order to find optimum separation conditions [22]. A third approach to confine the separation region was achieved by Kobayashi et al. [39] who implemented parallel shallow side banks (20 mm) in-between the separation channel (30 mm) and the electrode beds (30 mm). Similar to the side channels discussed before, these shallower regions have a higher hydrodynamic resistance than the separation chamber, thus leading the carrier electrolyte flow through the separation region. Kobayashi et al. [39] presented a 100 mm diameter wafer-sized Pyrex FFE device where they implemented this so-called bank shape design. The efficiency of the banks shape of the m-FFE was enough to prevent dispersion to the separation chamber of bubbles generated at the electrodes. 3.3 Integrated electrodes inside the separation chamber The direct integration of electrodes into the separation chamber allows for easy flow control and chip layout. However, in order to avoid the disturbance of gas bubbles one has  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Microfluidics and Miniaturization 985 to operate the device with voltages below the electrolysis potential. Even then the occurrence of electrode reactions cannot be excluded. As shown in Fig. 9, Lu et al. [40] used a microfabricated glass chip with integrated vertical gold electrodes directly inside the separation channel. They demonstrated FFIEF of subcellular organelles with voltages applied below the potential required for electrolysis, therefore completely avoiding gas formation inside the channel. However, due to the low electrical field strength available for separation, long residence times of up to several minutes were required. Due to this, most likely only components of low diffusivity, such as cell fractions and organelles could be separated. Song et al. [41] presented a microfabricated FFE chip for IEF without the usage of carrier ampholytes. Instead of applying an external electrical field required to run the pIbased sorting the field is generated by the diffusion of buffer ions in situ, at the liquid junction between two laminar flows within the microfluidic channel. However, the separation could be enhanced by applying an additional low voltage of 1.3 V using integrated electrodes as demonstrated. Figure 9. (A) Schematic of field flow IEF of organelles. (B) Photograph of the microfabricated device before final assembly with (C) enlarged view of the fractionation end of the device. The device consists of electroplated gold electrodes and microfluidic channels formed in a photopatternable epoxy. Reproduced with permission from ref. [40], copyright American Chemical Society. www.electrophoresis-journal.com 986 D. Kohlheyer et al. 3.4 Mechanically and electrically insulated separation chamber A different approach to isolate the electrodes from the separation chamber was investigated by Janasek et al. [42]. Janasek and co-workers faced the question, if it is possible to achieve a stable electrical field across the FFE separation chamber by electrostatic induction, like in an ideal DC capacitor. A glass FFE chip was fabricated containing integrated aluminum electrodes shielded from the separation chamber by 146 mm wide glass walls. If an electrical field is applied over an insulator, a rearrangement of the mobile charges in the liquid close to the wall of the insulator will take place. For example, if a negative potential is applied at the electrode, the amount of positive charge in the double layer adjacent to the wall will increase due to the decrease in the local electrical potential as illustrated, e.g., in the paper of Schasfoort et al. [43]. After a certain charging time it would be expected that the entire electrical field drop will occur over the capacitances formed by the glass walls and the electrical double layer, and that no electrical field would be present in chamber any more (see Fig. 10). As a result separations would be impossible. This might be the explanation of the phenomenon, which was observed in the presented FFE device under stagnant conditions, when no flow was applied. However, under flow conditions a stable FFE separation was observed, with which FFITP of fluorescein was demonstrated while no electrical current was measured [42]. The authors concluded that the flow is counteracting the accumulation of charges at the glass walls, but a clear explanation of the phenomenon was not given. More detailed experiments are currently performed, to understand this phenomenon in more detail. Comparable to earlier devices, the voltage efficiency was approximately 50%. Figure 10. Principle of electrostatic induction by charge displacement caused by dipole orientation. Lower panel: equivalent circuit diagram. The capacitor C1 and the resistor R1 for the dielectric barrier are in the range of pF and TV, respectively; C2 and R2 for the liquid compartment are in the range of tens of fF and a few V, respectively. Reproduced with permission of The Royal Society of Chemistry from [42].  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Electrophoresis 2008, 29, 977–993 3.5 Device technology – concluding remarks Several technological approaches for m-FFE devices have been demonstrated in literature over the past few years, in which the main challenges were to avoid separation interference due to gas bubbles, to establish a stable electrical field, and to optimize voltage efficiency. It is difficult to choose a favorite solution, since many aspects have to be compared. A brief comparison of the different approaches can be found in Table 1. An FFE chip design with open electrode reservoirs (Fig. 4, category a) allows indeed for easy ventilation of gas products. However, such an approach puts higher demands on a proper shielding of the separation chamber. This shielding structure, usually a kind of membrane substitution has to be of low electrical resistance in order to achieve good voltage efficiency but also of high hydrodynamic resistance to prevent for example fluid leakage from the separation chamber towards the open electrodes reservoirs. Therefore, several groups implemented so-called side channel arrays which acted as a membrane allowing electrical current flow but preventing fluid flow. Using integrated side channel arrays was an elegant solution since they can be fabricated usually in one step. However, the main drawbacks are a poor voltage efficiency and especially in FFIEF the formation of a pH gradient inside the side channels. As a positive side effect, it has been reported, that with the open characteristics of side channels rather than closed membranes electrodynamic band broadening could be avoided. Another technological solution to isolate the separation chamber from the open electrode reservoirs was the implementation of gel membranes, usually of acrylamide. This more laborious implementation of conductive acrylamide membranes has proven to be an efficient method with good voltage efficiency. Acrylamide gel membranes enable high hydrodynamic resistance with low electrical resistance and a relatively small width which seems very promising for further miniaturization. However, the mechanical stability of the membranes can be the limiting factor in some designs. Generally, it should be noted that a continuous replacement of the electrolyte solutions is of importance for longer separation times, in order to avoid changes of the solutions electrical properties. A continuous refreshment of the solutions surrounding the electrodes seems straightforward when implementing electrodes into closed channels parallel to the separation chamber (see also Fig. 4, category b). Additional pumping equipment is then required to achieve flows in order to remove gas bubbles efficiently. To avoid crosscontamination between the main flow inside the separation chamber and the electrode electrolyte flows, a proper flow balancing in this technological approach becomes more important. In these devices also side channel arrays have been used, but of lower hydrodynamic resistance. However, disturbance of the electrical field was reported due to gas bubbles moving along the electrodes and evenwww.electrophoresis-journal.com Electrophoresis 2008, 29, 977–993 tually entering and blocking the interconnecting side channels. Furthermore, without proper fluid and pressure balancing fluid flow through the side channels occurred. Not many results have been published on the usage of shallower side banks, although this design seemed to work well. The device we reviewed that used this method was relatively large with 100 mm in diameter. Instead of using any membrane substitution the usage of closed deep electrode channels was found to be an efficient method to remove gas bubbles without scarifying voltage efficiency. This method seems very promising, although it requires more and precise control of flow rates. However, it has to be investigated, if the design is also applicable for other FFE methods such as FFIEF. Electrodes can also be placed directly inside the separation chamber as shown by some groups (see Fig. 4, category c). As a result however, only voltages below the electrolysis potential can be applied. This usually leads to long separation times of up to several minutes making this approach only suitable for low diffusivity substances. The usage of a diffusion potential rather than applying an external electrical field forms an interesting method for pI-based sorting, although the resolution seemed to be low. The technological approach where a mechanical and electrical isolation of electrodes from the separation chamber (here electrostatic induction, see also Fig. 4, category d) is used, has to be studied more intensively and a clear explanation and understanding of the phenomenon should be found, but might offer an interesting alternative to other methods. 4 Separation results As mentioned, the four standard modes of FFE include zone electrophoresis, IEF, ITP and, field step electrophoresis. All modes have been applied in m-FFE devices, generally as proof-of-principle. In this section, we review the separation results obtained. 4.1 FFZE Raymond et al. [18] were the first to publish results on FFZE applying their microfabricated FFE device. The potential of the system was demonstrated by the separation of three rhodamine-B isothiocyanate-labeled amino acids, namely lysine, glutamine, and glutamic acid as shown in Fig. 11. A residence time of 73 s was needed to separated the components when 50 V were applied, which resulted in an electrical field strength (estimated by the authors) of 3 V/mm (I = 7.5 mA). Raymond et al. [18] found that the separation resolution was controllable by varying the side bed electrical conductivity, which was mainly due to the large contribution of the side channels to the total electrical resistance. Higher side bed conductivity resulted in an improved resolution, as the voltage drop across the separation chamber increased. EOF did  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Microfluidics and Miniaturization 987 Figure 11. Intensity plot of three labeled amino acids continuously separated by FFZE with a residence time of 73 s when 50 V were applied. Reproduced with permission from ref. [18], copyright American Chemical Society. not significantly affect the experimental baseline widths. This was probably a consequence of using side channels rather than closed membranes for the isolation of the separation chamber. However, this more open characteristic of the side channels resulted in flow from the side beds into the separation chamber, affecting separation. Eventually the bubbles created during electrolysis filled up the side beds resulting in a loss of the electrical field. It was reported, that no apparent Joule heating was observed. For a more detailed separation study, Raymond et al. [20] used the same chip design for the continuous separation of high molecular weight compounds with FFZE. In fact, they separated a mixture of FITC-labeled HSA, bradykinin, and ribonuclease A. An electrical field strength of 100 V/cm and a residence time of 62 s was needed to fully separate these components. Furthermore, the authors investigated different sources of band broadening and found that, initial bandwidth, diffusion, and hydrodynamic band broadening were the main contributors to the band broadening. Despite some problems, Raymond et al. [20] reported a peak capacity of 8 bands/cm. This was in the same range as reported for conventional systems (10 bands/cm), indicating that miniaturized FFE devices could achieve similar separation results. They also demonstrated that the chip is capable of separating more complex samples such as diluted rat plasma and trypic digests of bovine cytochrome C and melittin. Kobayashi et al. [39] demonstrated FFZE of two native proteins, cytochrome C and myoglobin. The residence time for a complete separation was 10 s when 2 kV was applied (estimated by the authors: E = 33 V/mm, I = 0.7 mA). The separated analytes were collected at different outlets and analyzed offline by RP HPLC. The authors applied a hydroxymethylcellulose (HPMC) coating in order to reduce the glass www.electrophoresis-journal.com 988 D. Kohlheyer et al. surface charge, trying to minimize EOF. Although the EOF was suppressed, the coating efficiency was not stable during longer separations. The device was also used for a theoretical study on the temperature distribution and Joule heating [44]. Zhang and Manz [32] presented a m-FFE chip for highspeed separations of analytes. With their improved chip design, they achieved a full separation of the fluorescent dyes rhodamine 110 and fluorescein within 75 ms. The total residence time (sample flow rate was 2 nL/s) in the separation chamber was 2 s when 1.75 kV was applied (E = 13.5 V/mm, I = 0.14 mA). As shown in Fig. 12, the negatively charged fluorescein (left stream) is slightly deflected towards the negative electrode, which was caused by EOF, as Zhang and Manz [32] concluded. The chip has also been used for the separation of FITC-labeled amino acids in both aqueous and binary media. FFZE of a mixture of fluorescent dyes, namely fluorescein, rhodamine 110, a rhodamine 110 impurity, and rhodamine 123 was presented by Fonslow and Bowser [36]. With the application of 515 V (I = 310 mA), resulting in an electrical field strength inside the separation chamber of 25.9 V/ mm, the three dyes as well as the impurity clearly separated with a residence time of 9.6 s before detection. In agreement with previous publications, Joule heating was found not to be significant at the studied field strengths. Fonslow et al. [38] further developed their original FFE chip design in order to apply higher voltages and improve voltage efficiency. As before, they separated fluorescein, rhodamine 110, rhodamine 110 impurity, and rhodamine 123 by FFZE. With a linear flow velocity of 5 mm/s and electrical field strength up to 58.6 V/mm an improved and more stable separation compared to their previous device was observed. With separation voltages of over 645 V Joule heating became Figure 12. FFZE of rhodamine 110 and fluorescein with a flow velocity of 6 mm/s. The resulting electrical field strength was 135 V/cm with an induced current of 140 mA. The right separated stream was rhodamine-110 and the left fluorescein. Reproduced with permission from ref. [32], copyright American Chemical Society.  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Electrophoresis 2008, 29, 977–993 significant. The same chip was used for a more theoretical study to optimize bandwidth and separation resolution in FFZE [22]. Kohlheyer et al. [21] presented FFZE of fluorescein and rhodamine B. Both analytes were separated within a total residence time of 3.3 s when 180 V was applied (E = 25 V/ mm, I = 50 mA). The linear flow velocity was 3 mm/s. The precise control of two sheath flow streams enabled positioning of the separated components within the separation chamber. Separated components could be steered towards different outlets by varying the sheath flow rates without altering the separation voltage, as shown in Fig. 13. This steering technique can be used in future devices to purify a sample mixture in such a way that only the separated component of interest is steered to a specific outlet. Steering of fluorescently labeled, easy visible components seems straightforward. However, it is questionable how to control the precise position in more realistic label-free applications. In order to demonstrate the feasibility of their FFE chip, de Jesus et al. [37] separated a mixture of the ionic dyes bromophenol, brilliant blue and crystal violet. The separation was carried out at a linear flow velocity of 0.54 mm/s with a total residence time of 18.2 s. 300 V were applied to achieve an electrical field strength inside the separation chamber of app. 11.5 V/mm. During 3 h of evaluation under a maximum current of 706 mA a stable separation was observed. 4.2 FFIEF The first to publish results on microfluidic FFIEF were Xu et al. [33]. The authors demonstrated the IEF of fluorescently labeled angiotensin by a factor of 400 with 430 s focusing Figure 13. Steering of rhodamine B and fluorescein bands in FFZE. The sample stream is hydrodynamically focused by two parallel sheath flow streams. By adjusting the flow rates F1 and F2 the sample injection position is shifted. Reproduced by permission of The Royal Society of Chemistry from [21]. www.electrophoresis-journal.com Electrophoresis 2008, 29, 977–993 time. They applied 1750 V to realize an electrical field strength of 13.5 V/mm, as shown in Fig. 14. To generate the required pH gradient an ampholyte solution was applied. However, due the technical layout, as discussed before, only a part of the pH gradient was available for separation since the low-pH and high-pH ampholytes migrated into the side channels. The chip could be operated between 7.7 and 30 V/ mm before Joule-heating became significant and bubbles appeared. Continuous IEF of large molecules such as sub-cellular organelles was presented by Lu at al. [40]. In order to avoid electrolysis low voltages were applied. However, due to this procedure, long residence times were necessary in order to reach stable focusing. The method is therefore suited to separate components with low diffusion. Among other results, FFIEF of stained mitochondria from a cell lysate was shown. For this purpose an ampholyte-based pH gradient (pH 3–6) was utilized and 2 V were applied with a residence time of 6 min. Especially with its mild voltage conditions this method is suitable for more sensitive biological components. Using the same device as used for FFZE, Kohlheyer et al. [21] applied FFIEF for the separation of several fluorescent IEF markers. IEF markers with pI0 s at pH 4.5, 5.5, 7.6, and 8.7 were separated and focused. With a residence time at the point of measurement of 3 s (linear flow velocity 2 mm/s) and the application of 20 V (approximately 10 V/mm inside the separation chamber) all components fully focused at their pI0 s within a 500 mm wide pH gradient (pH 3–10). The full pH gradient was available for separation since outer sheath flows of high pH and low pH, respectively, formed a border for arranging ampholytes. Kohlheyer et al. [23] reported later an improved FFIEF chip based on their previous publication. Here, a preseparated ampholyte buffer was used which reduced the electrical current and the focusing time. FFIEF of various Figure 14. IEF of fluorescently labeled angiotensin II: Ang II = 10 mM, U = 1750 V; E = 135 V/cm. Reproduced with permission of The Royal Society of Chemistry from [33].  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Microfluidics and Miniaturization 989 IEF markers ranging from pI 3 to 10 and the protein HSA was demonstrated, as shown in Fig. 15. The authors determined, that the device is capable of separating analytes with a minimum difference in pI of D(pH) = 0.23. This resulted in a theoretical peak capacity of 29 peaks within 1.8 mm for a pH gradient pH 2.5–11.5. With optimized pI markers, commonly used for calibration, the peak capacity of conventional FFE systems is between 50 and 70 peaks within 70 mm separation width (personal communication, Dr. Gerhard Weber, BD Diagnostics). In terms of resolution, this would mean for a pH gradient (pH 2.5–11.5) a D(pI)min between 0.13 and 0.18 pH units. To achieve this, in the conventional FFE system usually voltages of up to 2 kV have to be applied, while Kohlheyer et al. applied only 150 V, clearly indicating the potential advantages of microfluidic FFIEF devices. In practice, the resolution is reduced when separating proteins and peptides [45]. This comparison shows that with m-FFIEF one can achieve comparable and even better resolution with applying much lower voltages. Instead of using sheath flow streams to confine the pH gradient, Albrecht and Jensen [24] used functionalized gel membranes with low and high pH buffering capacity, respectively. FFIEF of several fluorescent IEF markers (pI 3.5, 5.1, 7.2, and 7.6), with 200 V separation voltage and a residence time of 14 s, was shown. Although a lower resolution than in the publication of Kohlheyer et al. [21] was achieved, the device involved less control over fluid streams Figure 15. FFIEF of seven fluorescent IEF markers: When 150 V voltage (I = 50 mA) was applied, the markers (pI 4, 5.1, 6.2, 7.2, 8.1, 9, and 10.3) fully separated within less than 2 s. The sample flow rate was 0.4 mL/min (v = 2 mm/s). The apparent kinks in the fluorescent tracer paths are caused by merging multiple photographs. Reproduced with permission from ref. [23], copyright American Chemical Society. www.electrophoresis-journal.com 990 D. Kohlheyer et al. Electrophoresis 2008, 29, 977–993 and pumps. In contrast to previous results Albrecht and Jensen [24] were the first to report the need for a thermal cooling system to optimize their separation resolution. Therefore, a thermoelectric element was used to cool the top of the FFE device. A possible explanation is the lower heat dissipation of the chosen material (PDMS) compared with glass devices [46]. Recently, Albrecht et al. [47] also reported a cascaded FFIEF device based on their earlier published results. The streams of separated components from a first FFIEF separation chamber were guided into a second FFIEF stage resulting in an improved separation resolution. Their FFIEF device was used to focus native model proteins, denatured proteins, as well as protein complexes. Song et al. [41] demonstrated successful FFIEF of two different pI markers. This device operated differently than other FFIEF systems, since neither ampholytes nor an external voltage has to be applied. The electrical potential gradient was achieved by diffusion potential in a separation channel of around 100 mm in width. 4.3 FFITP Janasek et al. [34] were the first to report on microfluidic FFITP, and demonstrated the method by separating fluorescein, acetylsalicylic acid (ASS), and Eosin G, as shown in Fig. 16. Since in ITP the bands flow adjacent to each other, for a clear experimental validation ASS was chosen as a nonfluorescent spacer analyte between the two fluorescent components, fluorescein and Eosin G. For the experiment the electrical field strength was 21 V/mm. At higher field strength movement of the analytes into the side channels was observed. As a proof-of-principle, Janasek et al. [42] applied FFITP of fluorescein to investigate their principle of electrostatic induction for FFE. They successfully focused fluorescein with the application of 150 V, which corresponds to an electrical field strength of 18 V/mm inside the separation chamber, and a total flow rate of 20 mL/min. Figure 16. FFITP separation and concentration of fluorescein, ASS, and eosin G displayed as an assembly of single micrographs over the whole chamber. Reproduced with permission from ref. [34], copyright American Chemical Society. 4.4 FFFSE To the best of our knowledge, no results on microfluidic FFFSE have been published until now. To demonstrate the principle of FFFES, the authors separated a mixture of rhodamine B and fluorescein using their m-FFE device. For this purpose, the sample stream was hydrodynamically focused in-between two sheath flow streams of higher electrical conductivity. As shown in Fig. 17, this resulted in an electrical field step when 150 V were applied. The negatively charged fluorescein migrated towards the anode and eventually concentrated (approximately four-fold) at the field step due to a decrease in its mobility. The neutrally charged, nonmigrating Rhodamine B showed only a slight band broadening maintaining its centered position.  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Figure 17. FFFSE demonstrated in a microfluidic device of the authors. (a) Mixture of rhodamine B and fluorescein continuously flowing with no voltage applied. (b) When 150 V were applied the negatively charged fluorescein migrated and focused due to the sudden decrease in electrical field strength (Kohlheyer et al., unpublished result). www.electrophoresis-journal.com Electrophoresis 2008, 29, 977–993 4.5 Separation results – concluding remarks As shown by various publications, the four know separation modes, FFZE, FFIEF, FFITP, and FFFSE, have successfully been demonstrated in microfluidic devices. A wide spectrum of analytes has been separated including cell fragments, organelles, proteins, amino acids, fluorescent dyes, and markers. The results shown until now, which can be considered as proof-of-principle experiments, are difficult to compare in terms of resolution, bandwidth, and peak capacity, since too many parameters differ from experiment to experiment. Obviously, also predicted by scaling laws, separation times were drastically reduced by miniaturization, making it more interesting for online monitoring tools rather than conventional systems meant for sample preparation. Samples were fully separated within several seconds and below. Furthermore, sample flow rates of several microliters per minute and below were applied. Generally one can say that Joule heating, a major concern in conventional FFE systems seems to be less relevant on a microfluidic scale. Due to the lack of lateral separation length higher electrical field strengths have to be applied in order to achieve comparable separation resolution and peak capacity as on the large scale. This is possible, because due to the lower electrical currents involved (typically in the microampere range) less heat is generated, and because of the higher surface-tovolume ratios that favor fast heat dissipation. It would be beneficial for the field to investigate in more detail scaling effects by matching experiments performed on the microand macroscale. Optical detection methods, usually involving large fluorescent microscopes and optical equipment, until now have been sufficient for first characterizations. However, as proposed already by Raymond et al. [20], the major advantage of a m-FFE device is the possibility to integrate it more easily in a complex analysis system. Such an integration for online monitoring applications of separated analytes is not a straightforward task, as will be discussed in the following section. 5 Detection and hyphenation Size-reduced conventional mini-FFE devices have already been implemented into hyphenated analytical systems. Such systems can guide developers of integrated m-FFE systems, and we, therefore, mention a few examples. Chartogne et al. [7] coupled an FFE device to CIEF as a cleaning step to remove interfering ampholyte substances prior to ESI-MS to improve the MS sensitivity. This CIEFFFE-ESI-MS system was successfully realized for three model proteins (myoglobin, carbonic anhydrase I, and b-lactoglobulin B). Furthermore, Mazereeuw et al. [6] realized a system in which FFE was used as an online separation step of fluorescent compounds in a biochemical detection system. This system was based on the interaction of the affinity protein streptavidin with biotin, in which biotin acted as an  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Microfluidics and Miniaturization 991 analyte. This FFE detection system was coupled to HPLC for the bioanalysis of biotin in human urine. In the field of m-FFE systems not much progress has been shown on the implementation of FFE with other separation or detection systems into so-called micro total analysis systems (m-TAS) or LOC devices [8, 9, 48, 49]. In fact, until now only propositions for more integrated systems have been published. Zhang and Manz [32] concluded from their investigations, that m-FFE presents a potentially powerful tool in proteomics studies. They proposed coupling m-FFE with electrospray MS as an online automated analytical tool in proteomics studies. By moving the chip perpendicular to the outlets separated components spray from respective channels and analyte streams are then received by the MS in scanning mode. Although they found that only little modifications are necessary for this detection approach, to our knowledge no results have been published until now. Instead of scanning the entire chip, the capability to steer the analyte by hydrodynamic focusing during FFZE separation, as shown by Kohlheyer et al. [21], might offer an interesting alternative to scan separated components through a single chip outlet connected to an ESI-MS system. A step towards this approach was the coupling of a glass microchip for CE to ESI-MS as, e.g., successfully demonstrated by Hoffmann et al. [50] and others [51]. Furthermore, Zhang and Manz [32] proposed the use of m-FFE combined with a microreactor, for on-chip synthesis and separation of products. Using microchip CE, on-chip synthesis and separation of products was shown by Belder et al. [52]. In this initial study, the device was successfully applied for testing biocatalysts created by directed evolution of enzymes. The use of m-FFE instead of microchip CE could enable high throughput and continuous screening of this reaction rather than the analysis of discrete plugs achieved by microchip CE. An integrated system approach combining m-FFE and surface plasmon resonance (SPR) detection was proposed by Kohlheyer et al. [53]. As illustrated in Fig. 18, a thin gold detection area is integrated into the FFE separation chip, located after the separation chamber. A monochromatic ppolarized light beam is coupled into the chip device and reflected under total internal reflection at the gold surface. Refractive index changes, near the gold surface, e.g., caused by immobilization of molecules on the gold surface lead to local changes in the reflected light intensity. These light intensity changes can be monitored for the whole detection region simultaneously by a CCD camera. For more information about SPR and its theoretical background one is referred to the review of Homola [54]. For experiments the proposed chip is placed inside an SPR imaging instrument (IBIS Technologies, The Netherlands) to monitor immobilization of separated fractions. Eventually this generated fraction pattern is used in a second step to study biomolecular interactions in the search for specific biomarkers present in, e.g., patients sample. Collected fractions of immobilized lanes, which positively react with specific antibodies, will be anawww.electrophoresis-journal.com 992 D. Kohlheyer et al. Electrophoresis 2008, 29, 977–993 as reviewed in this paper will further be developed for new applications in life sciences. Broadening the options of parallelization and assay implementation, including sample treatment on-a-chip will certainly contribute to an increased number of publications in the near future. Definitely a total analysis system approach instead of a device-based approach will lead to a growth in the field of m-FFE. We would like to acknowledge the funding of this research by the Dutch Technology Foundation STW. Furthermore, Dr. Gerhard Weber and Professor Petr Boček are acknowledged for help with theoretical aspects. The authors have declared no conflict of interest. 7 References [1] Hannig, K. Z., Anal. Chem. 1961, 181, 244–254. [2] Bauer, J. (Ed.), Special Issue on Free-Flow Electrophoresis, Electrophoresis 1998, 19, 1057–1235. Figure 18. Illustration of the FFE microchip with an integrated gold area for SPR detection. (a) The chip is placed inside the SPR imaging instrument with a monochromatic light beam coupled into the chip device. (b) Cross-section of the chip detection region: biomolecular interaction and molecule immobilization cause refractive index changes which can be monitored real time by local light intensity changes [53]. lyzed offline by MS for identification purpose. However, until now only preliminary results were presented and more work is currently performed. [3] Hannig, K., Heidrich, H. G., Free-Flow Electrophoresis, GIT Verlag, Darmstadt 1990. [4] Weber, G., Boček, P., Electrophoresis 1998, 19, 1649–1653. [5] Weber, P. J. A., Weber, G., Eckerskorn, C., in: Simpson, R. J. (Ed.), Purifying Proteins for Proteomics: A Laboratory Manual, Cold Spring Harbor Laboratory Press, New York 2003. [6] Mazereeuw, M., de Best, C. M., Tjaden, U. R., Irth, H., van der Greef, J., Anal. Chem. 2000, 72, 3881–3886. [7] Chartogne, A., Tjaden, U. R., van der Greef, J., Rapid Commun. Mass Spectrom. 2000, 14, 1269–1274. [8] Tudos, A. J., Besselink, G. A. J., Schasfoort, R. B. M., Lab Chip 2001, 1, 83–95. [9] Dittrich, P. S., Tachikawa, K., Manz, A., Anal. Chem. 2006, 78, 3887–3908. 6 Conclusions [10] Manz, A., Eijkel, J. C. T., Pure Appl. Chem. 2001, 73, 1555– 1561. [11] Das, C., Fan, Z. H., Electrophoresis 2006, 27, 3619–3626. In FFE, an electrical field is applied perpendicular to the flow direction and charged molecules are deflected from the carrier electrolyte flow direction, in a way that is controlled by the electrophoretic mobility, the flow velocity, and the electrical field strength. The research field of m-FFE methods is rather small, as evident from the number of papers reviewed. However, it carries great promise due to the continuous flow of separated sample enabling integration and continuous monitoring in miniaturized analysis systems. In this paper, we showed that steady innovation is occurring in this research field. The results shown so far indicated that with m-FFE systems one can achieve high separation resolutions which are comparable to or even better than those reached in their larger counterparts. The microfluidic devices furthermore accomplish this separation in a fraction of the time the larger devices need, and consuming a fraction of the sample volume. A trend can be foreseen in which the integration of LOC devices using one of the FFE principles  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim [12] Pamme, N., Lab Chip 2007, 7, 1644–1659. [13] Baker, D. R., Capillary Electrophoresis, John Wiley and Sons Inc., New York 1995. [14] Křivánková, L., Boček, P., Electrophoresis 1998, 19, 1064– 1074. [15] Roman, M. C., Brown, P. R., Anal. Chem. 1994, 66, 86A–94A. [16] Kašička, V., Prusík, Z., Pospísek, J., J. Chromatogr. 1992, 608, 13–22. [17] Kašička, V., Prusík, Z., Sazelova, P., Jiracek, J., Barth, T., J. Chromatogr. A 1998, 796, 211–220. [18] Raymond, D. E., Manz, A., Widmer, H. M., Anal. Chem. 1994, 66, 2858–2865. [19] Gebauer, P., Boček, P., Electrophoresis 2005, 26, 453–462. [20] Raymond, D. E., Manz, A., Widmer, H. M., Anal. Chem. 1996, 68, 2515–2522. [21] Kohlheyer, D., Besselink, G. A. J., Schlautmann, S., Schasfoort, R. B. M., Lab Chip 2006, 6, 374–380. www.electrophoresis-journal.com Electrophoresis 2008, 29, 977–993 Microfluidics and Miniaturization 993 [22] Fonslow, B. R., Bowser, M. T., Anal. Chem. 2006, 78, 8236– 8244. [38] Fonslow, B. R., Barocas, V. H., Bowser, M. T., Anal. Chem. 2006, 78, 5369–5374. [23] Kohlheyer, D., Eijkel, J. C. T., Schlautmann, S., van den Berg, A., Schasfoort, R. B. M., Anal. Chem. 2007, 79, 8190–8198. [39] Kobayashi, H., Shimamura, K., Akaida, T., Sakano, K. et al., J. Chromatogr. A 2003, 990, 169. [24] Albrecht, J. W., Jensen, K. F., Electrophoresis 2006, 27, 4960– 4969. [40] Lu, H., Gaudet, S., Schmidt, M. A., Jensen, K. F., Anal. Chem. 2004, 76, 5705–5712. [25] Wang, Q., Tolley, H. D., LaFebre, D. A., Lee, M. L., Anal. Bioanal. Chem. 2002, 373, 125–135. [41] Song, Y. A., Hsu, S., Stevens, A. L., Han, J., Anal. Chem. 2006, 78, 3528–3536. [26] Righetti, P. G., Isoelectric Focusing: Theory, Methodology and Applications, Elsevier Biomedical, Amsterdam 1983. [42] Janasek, D., Schilling, M., Manz, A., Franzke, J., Lab Chip 2006, 6, 710–713. [27] Boček, P., Deml, M., Gebauer, P., Dolník, V., Analytical Isotachophoresis, Verlag Chemie, Weinheim 1988. [43] Schasfoort, R. B. M., Schlautmann, S., Hendrikse, J., van den Berg, A., Science 1999, 286, 942–945. [28] Everaerts, F. M., Beckers, K. L., Verheggen, T. P. E. M., Isotachophoresis – Theory, Instrumentation and Applications, Elsevier, Amsterdam 1976. [44] Matsumoto, H., Komatsubara, N., Kuroda, C., Tajima, N. et al., Chem. Eng. J. 2004, 101, 347–356. [29] Gebauer, P., Boček, P., Electrophoresis 1995, 16, 1999–2007. [45] Malmstrom, J., Lee, H., Nesvizhskii, A. I., Shteynberg, D. et al., J. Proteome Res. 2006, 5, 2241–2249. [30] Kuhn, R., Hoffstetter, S., Wagner, K. H., Electrophoresis 1990, 11, 942–947. [46] Fu, R., Xu, B., Li, D., Int. J. Therm. Sci. 2006, 45, 841–847. [31] Stone, V. N., Baldock, S. J., Croasdell, L. A., Dillon, L. A. et al., J. Chromatogr. A 2007, 1155, 199–205. [32] Zhang, C.-X., Manz, A., Anal. Chem. 2003, 75, 5759–5766. [33] Xu, Y., Zhang, C. X., Janasek, D., Manz, A., Lab Chip 2003, 3, 224–227. [34] Janasek, D., Schilling, M., Franzke, J., Manz, A., Anal. Chem. 2006, 78, 3815–3819. [35] Takamura, Y., Onoda, H., Indokuchi, H., Oki, A., Horiike, Y., Electrophoresis 2003, 24, 185–192. [36] Fonslow, B. R., Bowser, M. T., Anal. Chem. 2005, 77, 5706– 5710. [37] de Jesus, D. P., Blanes, L., do Lago, C. L., Electrophoresis 2006, 27, 4935–4942.  2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim [47] Albrecht, J. W., El-Ali, J., Jensen, K. F., Anal. Chem. 2007, 79, 9364–9371. [48] Schasfoort, R. B. M., Expert Rev. Proteomics 2004, 1, 123– 132. [49] Janasek, D., Franzke, J., Manz, A., Nature 2006, 442, 374–380. [50] Hoffmann, P., Häusig, U., Schulze, P., Belder, D., Angew. Chem. Int. Ed. 2007, 46, 4913–4916. [51] Koster, S., Verpoorte, E., Lab Chip 2007, 7, 1394–1412. [52] Belder, D., Ludwig, M., Wang, L.-W., Reetz, M. T., Angew. Chem. Int. Ed. 2006, 45, 2463–2466. [53] Kohlheyer, D., Besselink, G. A. J., Schlautmann, S., Schasfoort, R. B. M., Nanotech 2006, Montreux, Switzerland 2006. [54] Homola, J., Anal. Bioanal. Chem. 2003, 377, 528–539. www.electrophoresis-journal.com