In situ fabrication of a microfluidic device for immobilised metal affinity sensing

RESEARCH PAPER

New Biotechnology  Volume 29, Number 4  May 2012

Research Paper

In situ fabrication of a microfluidic device for immobilised metal affinity sensing Abhishek G. Deshpande, Nicholas J. Darton, Kamran Yunus, Adrian C. Fisher and Nigel K.H. Slater Department of Chemical Engineering and Biotechnology, University of Cambridge, New Museums Site, Pembroke Street, Cambridge CB2 3RA, UK

In this work a novel microfluidic device was constructed in situ containing the smallest microscopic copolymeric immobilised metal affinity (IMA) adsorbent yet documented. This device has for the first time allowed the microlitre scale chromatographic assay of histidine-tagged proteins in a biological sample. To enable this approach, rather than using a high capacity commercial packed bed column which requires large sample volumes and would be susceptible to occlusion by cell debris, a microgram capacity co-polymeric chromatographic substrate suitable for analytical applications was fabricated within a microfluidic channel. This porous co-polymeric IMA micro-chromatographic element, only 27 ml in volume, was assessed for the analytical capture of two different histidine-tagged recombinant fusion proteins. The micro-chromatographic adsorber was fabricated in situ by photo-polymerising an iminodiacetic acid (IDA) functionalised polymer matrix around a template of fused 100 mm diameter NH4Cl particles entirely within the microfluidic channel and then etching away the salt with water to form a network of interconnected voids. The surface of the micro-chromatographic adsorber was chemically functionalised with a chelating agent and loaded with Cu2+ ions. FTIR and NMR analysis verified the presence of the chelating agent on the adsorbent surface and its Cu2+ ion binding capacity was determined to be 2.4 mmol Cu2+ (ml of adsorbent)1. Micro-scale equilibrium adsorption studies using the two different histidine-tagged proteins, LacI-His6-GFP and a-Synuclein-His8-YFP, were carried out and the protein binding capacity of the adsorbent was determined to be 0.370 and 0.802 mg (g of adsorbent)1, respectively. The dynamic binding capacity was determined at four different flow rates and found to be comparable to the equilibrium binding capacity at low flow rates. The sensing platform was also used to adsorb LacI-His6-GFP protein from crude cell lysate. During adsorption, laser scanning confocal microscopy identified locations within the adsorbent where protein adsorption and desorption occurred. The findings indicate that minimal channelling, selective product capture and near quantitative elution of the captured (adsorbed) product could be achieved, supporting the application of this new device as a high-throughput process analytical tool (PAT) for the in-process monitoring of histidine-tagged proteins in manufacturing.

Corresponding author: Darton, N.J. ([email protected])

494

www.elsevier.com/locate/nbt

1871-6784/$ - see front matter ß 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.nbt.2012.01.002

New Biotechnology  Volume 29, Number 4  May 2012

The utility of microfluidic devices has been demonstrated in bioanalytical applications. Gao et al. [1] reported the use of a trypsin functionalised PVDF (polyvinylidene fluoride) membrane in a poly(dimethylsiloxane) (PDMS) microfluidic chip that enabled rapid protein digestion for peptide mapping. Extending this approach, Peterson et al. [2] developed a methacrylate based device for rapid tryptic digestion of myoglobin. West and coworkers [3] described the isolation of nucleic acids using an anion exchange adsorbent inside a microchannel and the use of microfluidic devices for proteomic sample preparation was demonstrated by Freire et al. [4]. Microfluidic affinity devices will similarly find applications in proteomics research and in process analytics [4,5]. In situ fabricated adsorbents that avoid microchannel packing will enable small biological samples to be processed rapidly and reliably [6]. Immobilised metal affinity (IMA) adsorption is widely used [7,8] and involves the selective binding of an immobilised metal ion and a polyhistidine-tagged region on a fusion protein [8]. IMA can be used for protein purification, on-column protein refolding, protein surface topography studies and biosensor development [9,10]. Microfluidic IMA devices offer the potential for efficient and rapid protein sensing at the microlitre scale. Slentz et al. [11] described a microfluidic chip packed with 5 mm diameter Cu2+ IMA adsorbent particles for the adsorption of histidine containing peptides such as those resulting from trypsin digested bovine serum albumin. However, being a conventional packed bed this device would be less suitable for the processing of fluids containing cell debris as it would be susceptible to occlusion. In the present work, the performance of a microfluidic device comprising an in situ fabricated IMA absorber is demonstrated for the detection of histidine-tagged recombinant proteins. The adsorbent was a co-polymeric element of HEMA (2-hydroxyethyl methacrylate) and a vinyl pre-functionalised Cu2+ chelating monomer HPIDA (disodium-2,20 -{[2-hydroxy-3-(prop-2-en-1-yloxy)propyl]imino}diacetate) cross-linked with EGDMA (ethyleneglycol dimethacrylate) [12]. Porosity in these adsorbents was created by incorporating NH4Cl particles in the polymer matrix that were subsequently etched out. The objective was to create a porous structure of sufficient channel size to permit the processing of crude cell lysates yet of sufficient adsorptive capacity for analytical use. The application of this novel adsorbent for the affinity capture and detection of two fusion proteins incorporating different length histags and fluorescent marker regions was studied. The first of these fusion proteins, LacI-His6-GFP, contained an affinity tag of six histidine residues and the green fluorescent protein marker. The second fusion protein tested was a-Synuclein-His8-YFP which contained a longer eight histidine residue affinity tag and the yellow fluorescent protein marker.

Materials and methods Reagents and materials Ammonium chloride salt was purchased from Fisher Scientific (UK), Coomassie protein assay from Thermo Scientific (USA), SU8 2100 photo-resist agent from Microchem (UK), allyl glycidyl ether (AGE) from Avocado Research Chemicals (UK), 3-aminopropyltrimethoxysilane (95%) from Acros Chemicals (UK) and sodium chloride and Tris–HCl from Duchefa Biochemie (Netherlands) and Melford Lab

(UK), respectively. Sulphuric acid came from BDH. 2-Hydroxyethyl methacrylate (HEMA), disodium-2,20 -{[2-hydroxy-3-(prop-2-en-1yloxy)propyl]imino}diacetate (HPIDA), ethylene glycol dimethacrylate (EGDMA), sodium carbonate, 2,2-dimethoxy 2-phenyl acetophenone (DMPA), hydrogen peroxide, copper(II) sulphate, imidazole, sodium phosphate, methanol and Tween 20 were purchased from Sigma–Aldrich, UK.

Microfabrication Glass wafers were cut to size and treated with Piranha acid (3:1 ratio of sulphuric acid to hydrogen peroxide) for 5 min to remove any impurities. The glass was rinsed in Milli-Q water and blow dried with nitrogen. Two different moulds, macro moulds (wells) and microchannel moulds, were prepared on glass wafers for polymer synthesis. Macro moulds (length (l) = 5 cm, height (h) = 0.025 cm, width (w) = 5 cm) were prepared using a heat lamination method for batch experiments and microchannel moulds (l = 2.5 cm, h = 0.07 cm, w = 0.155 cm) were prepared using micromachining (ZX45, Warco Milling Equipments) for flow experiments [13]. The moulds were silanised using 3-aminopropyltrimethoxysilane, filled with 100 mm diameter NH4Cl particles and humidified (95% relative humidity) to form a fused NH4Cl particle template. The 100 mm diameter NH4Cl particles were prepared using a pulveriser (Pulverisette 6, Fritsch GmBH) and particle sieve (EVS1, Endecotts Sieve Shaker) and the particle size distribution was obtained using a laser scattering particle sizer (LS230, Beckman Coulter). The porosity obtained by mercury porosimetry was found to be around 38–45% and the BET surface area was approximately 4.4 m2 g1.

Polymer synthesis The co-polymeric adsorbent (Fig. 1) was synthesized by photopolymerisation of HEMA (2-hydroxyethyl methacrylate) and a vinyl pre-functionalised Cu2+ chelating monomer HPIDA (disodium2,20 -{[2-hydroxy-3-(prop-2-en-1-yloxy)propyl]imino}diacetate) cross-linked with EGDMA (ethyleneglycol dimethacrylate). The HPIDA monomer for this synthesis was made by dissolving IDA in 0.2 M sodium carbonate and reacting this with allyl glycidyl ether (AGE) in a 1:1 molar ratio at pH 10 (see Chen et al. [14]). The solution was heated gradually to a temperature of 50–608C in an oil bath and the reaction, catalysed by the sodium carbonate was carried out for over eight hours to ensure complete conversion. A dark yellow viscous aqueous solution containing HPIDA was obtained. Nine different HEMA and HPIDA monomer mixtures were prepared by varying the volume ratio of HEMA to HPIDA (VHEMA/VHPIDA) from 0.5 to 5 whilst keeping the total co-monomer volume constant at 3 ml. The volume of the EGDMA cross-linker added to each of the nine monomeric mixtures was also kept constant at 2 ml. Additionally a 1 volume(%) of 2,2-dimethoxy 2phenyl acetophenone (DMPA) was added as photo-initiator. The monomer mixtures were added to the NH4Cl particle template and exposed to an ultraviolet (UV) light source (Karl Suss mask alignerMJB3, 8 mW cm2) for 10 s for complete polymerisation to give a tertiary-amine metal-ion-chelating poly(HEMA-co-HPIDA-coEGDMA) adsorbent. Exposure time for complete monomer conversion was found to be dependent on the height of the adsorbents as well as composition of the monomer mixture. Following photopolymerisation the adsorbents were treated with deionised water www.elsevier.com/locate/nbt

495

Research Paper

Introduction

RESEARCH PAPER

RESEARCH PAPER

New Biotechnology  Volume 29, Number 4  May 2012

OH

O

OH +

O

O

N

O

+ O

HO

O

HPIDA

EGDMA

HEMA

O

OH

O

O

hv 120 s OH

Research Paper

O

O

OH

*

n

* O

HO

O

O

O O

N OH

O HO O O *

O

Cross-linked by EGDMA

n

*

Poly(HEMA-co-HPIDA-co-EGDMA)

FIGURE 1

Reaction scheme for poly(HEMA-co-HPIDA-co-EGDMA) synthesis from monomers HEMA and HPIDA and EGDMA as a cross-linker.

to dissolve the embedded 100 mm NH4Cl particle template creating a porous structure. The adsorbents were characterised using FTIR and 15N-NMR to confirm the presence of tertiary-amine groups and the amount of nitrogen was quantified using elemental analysis. High resolution images were obtained using Ultra FE-SEM (Field Emission Scanning Electron Microscopy) and XL30-ESEM (Environmental Scanning Electron Microscopy).

Cu2+ loading and quantification 1 g of each of the nine adsorbent samples was rinsed thoroughly with 10 ml of 1 M sodium chloride solution, 10 ml of 2% Tween 20 and 10 ml of methanol to remove any residual monomer. The samples were loaded with 1 ml of 50 mM copper(II) sulphate solution and washed with an aqueous solution of 1 M sodium chloride, 250 mM sodium phosphate and 300 mM imidazole to remove any loosely bound Cu2+. The metal ion capacity (MIC) was determined by eluting the Cu2+ with 1 ml of 0.1 M EDTA and assayed using spectrophotometry at 730 nm. Sample measurements were taken in triplicate and the polymer samples were rinsed in deionised water after each step.

LacI-His6-GFP and a-synuclein-His8-YFP preparation To test the new microfluidic IMA adsorbent two histidine-tagged proteins were produced, LacI-His6-GFP and a-Synuclein-His8-YFP. LacI-His6-GFP protein containing six C-terminal histidine residues was made in transformed BL21 (E. coli) gold competent cells (Stratagene, US), as described by Darby et al. [15]. Cell lysates were prepared from the transformed BL21 cells by centrifuging down the cells at 4000 rpm and resuspending the resulting pellets in phosphate buffer (PBS) for lysis by sonication (4  3 s at maximum 496

www.elsevier.com/locate/nbt

power) followed by incubation on ice for 2 min. This sonication and incubation on ice was repeated four times. The lysed cells were centrifuged at 19,000 rpm and the supernatant was recovered. The LacI-His6-GFP cell lysate was purified using a commercial IMA column, Sepharose HisTrap FF (GE Healthcare), following the manufacturer’s instructions. The second protein, a-SynucleinHis8-YFP containing eight C-terminal histidine residues was kindly gifted by the Department of Chemistry at the University of Cambridge (UK). This had been prepared following expression in recombinant E. coli by affinity chromatography using Ni-NTAAgarose (Invitrogen) according to the supplier’s instructions, followed by size exclusion chromatography using a Superdex S75 column (GE Healthcare). Purity of both proteins was higher than 90% as measured by SDS-PAGE (data not shown).

Protein binding capacity – batch analysis 0.25 g of Cu2+ immobilised poly(HEMA-co-HPIDA-co-EGDMA) adsorbent was equilibrated using 9 ml of wash buffer (50 mM Tris–HCl, 10 mM Imidazole, 300 mM NaCl, pH 7.4) [16]. The adsorbent was then shaken firstly in 3 ml of 500 mg ml1 prepurified LacI-His6-GFP in 50 mM Tris–HCl for 20 min and then 20 min in wash buffer (108 adsorbent volume) followed by 56 adsorbent volumes of elution buffer (50 mM Tris–HCl, 130 mM Imidazole, 300 mM NaCl, pH 7.4). After each step, supernatant samples were collected, filtered (0.45 mm) and LacI-His6-GFP quantified by Bradford assay [17]. Two negative controls were performed in which LacI-His6-GFP was shaken with poly(HEMA-coHPIDA-co-EGDMA) adsorbent with no Cu2+ chelated/immobilised, and copper chelated adsorbent was loaded with a nonhistidine tag containing protein b-Lactoglobulin (500 mg ml1, MW: 18 kDa). To test the comparative binding affinities of the

RESEARCH PAPER

Protein binding capacity – flow analysis A fused NH4Cl template was formed within a glass channel (l = 2.5 cm, w = 0.155 cm, h = 0.07 cm) and the void space filled with a monomeric mixture of HEMA:HPIDA:EGDMA in the volume(%) composition of 41.4:18.6:40. The mixture was then photo-polymerised for 10 s using UV light. The length (L) of poly(HEMA-co-HPIDA-co-EGDMA) adsorbent inside the channel was restricted to 1.5 cm using a photo-mask, giving an adsorbent volume of 16.2 ml. The channel was flushed with deionised water for 12 hours using two syringe pumps (PhD 2000, Harvard apparatus) with flow rates between 0.017 ml s1 and 33 ml s1. The pressure drop during flushing was measured using a differential pressure sensor (Digitron Instrumentation). The connections between the glass channel and the pressure sensor were made using standard plastic tubing. Samples were collected manually for analysis as opposed to a continuous detection to keep consistency with the Bradford assay based measurement technique used during the entire experiments. The adsorbent was loaded with Cu2+, equilibrated using wash buffer (50 mM Tris–HCl, 10 mM Imidazole, 300 mM NaCl, pH 7.4) and then loaded with LacI-His6-GFP (170 mg ml1) at flow rates of 1.5 ml s1 (residence time, t = 2.7 s), 0.66 ml s1 (t = 6.2 s), 0.47 ml s1 (t = 8.77 s) and 0.3 ml s1 (t = 13.7 s). Following protein loading the adsorbent was washed with 5 column volumes of wash buffer (0.135 ml) and eluted using 5 column volumes of elution buffer (50 mM Tris–HCl, 130 mM Imidazole, 300 mM NaCl, pH 7.4). During these operations 20 ml samples were collected every 10 s during the load stage and every 0.5–1 min during the wash and elution stages. Similar experiments were performed with a-Synuclein-His8-YFP (170 mg ml1) at a flow rate of 0.47 ml s1 (t = 8.77 s). The concentration of protein was determined by Bradford assay and analysed by SDS-PAGE with silver staining using Silver Snap kit II protocol (Invitrogen). In a separate experiment, 2 ml of crude cell lysate containing LacI-His6-GFP was fed into the adsorbent at a flow rate of 1.33 ml s1 and then washed and eluted using 5 column volumes (0.135 ml) of wash buffer and 5 adsorbent volumes of elution buffer respectively. Samples of volume 20 ml were collected at an interval of approximately 1 min during the elution stage. The amount of protein loaded (qd) on to the adsorbent at a given time t was calculated using [18]: qd ¼ V f  ½Co  t 1

(1) 1

where Vf (ml s ) is the volumetric flow rate and [C]o (mg ml ) is the feed concentration of protein. Dynamic binding capacity was determined at 10% breakthrough.

Confocal analysis The poly(HEMA-co-HPIDA-co-EGDMA) adsorbent inside the glass channel was imaged at various positions during the loading, washing and elution stages using confocal microscopy. The adsorbent was loaded with LacI-His6-GFP (170 mg ml1) at a flow rate of 0.3 ml s1 (residence time, t = 13.7 s) and confocal images were

3 2.5 2 1.5 1 0.5 0 0

1

2

3

4

5

6

VHEMA /VHPIDA

FIGURE 2

Plot of metal ion capacity (MIC) against volume ratio of HEMA to HPIDA.

taken at various times to show the progression of the adsorbed protein front through a field of view of 0.2 cm length positioned 0.3 cm from the adsorbent entrance. Other confocal microscopy observations were made by moving the microscope along the axis of the adsorbent until breakthrough was observed. LacI-His6-GFP was then kept in the adsorbent overnight to ensure equilibrium binding. After 12 hours, the adsorbent was washed with 5 adsorbent volumes of buffer solution and then eluted using 5 adsorbent volumes of elution buffer.

Results and discussion Optimisation of adsorbent composition Experiments were carried out to study the effect of different adsorbent monomer compositions of HPIDA and HEMA on the metal ion binding capacity of the resulting adsorbents. Figure 2 shows the effect of different volume ratios of HEMA to HPIDA (VHEMA/VHPIDA) on the MIC of poly(HEMA-co-HPIDA-co-EGDMA). At low HEMA concentrations of 10–20 vol% in total monomer mixture the formation of a poly(HEMA-co-HPIDA-co-EGDMA) gel with a low MIC resulted. With these gels Cu2+ was bound weakly and most of it was removed by the wash buffer (1 M NaCl, 250 mM sodium phosphate and 300 mM imidazole). Increase in the proportion of HEMA to 40 vol% in the monomer mixture resulted in the formation of adsorbent structures with structural rigidity. With these adsorbents Cu2+ bound more strongly and the metal ion binding capacity increased to 2.42 mmol/ml of adsorbent. Further increase in the amount of HEMA to 50 vol% in total monomer mixture produced impermeable poly(HEMA-co-HPIDA-coEGDMA). The monomer composition corresponding to maximum metal ion binding capacity was HEMA 41.4 vol%, HPIDA 18.6 vol% and EGDMA 40 vol% and this was selected for the fabrication of adsorbents for the protein binding experiments.

Batch adsorption of protein A polyhistidine-tagged fusion protein, LacI-His6-GFP, was used to demonstrate the protein binding ability of the Cu2+ loaded adsorbents under batch adsorption conditions. Figure 3 shows a logarithmic plot of the total protein amount (MProtein) in the supernatant during loading with LacI-His6-GFP ([C]o = 500 mg ml1), washing the adsorbent with wash buffer and eluting the bound protein using elution buffer. Only LacI-His6-GFP loaded on Cu2+ immobilised www.elsevier.com/locate/nbt

497

Research Paper

LacI-His6-GFP and a-Synuclein-His8-YFP for the adsorbent, two 0.25 g poly(HEMA-co-HPIDA-co-EGDMA) adsorbents were loaded with 3 ml of 9.5 mg ml1 of LacI-His6-GFP (274 kDa) and a-Synuclein-His8-YFP (30 kDa).

MIC/µmol (ml of monolith) -1

New Biotechnology  Volume 29, Number 4  May 2012

RESEARCH PAPER

New Biotechnology  Volume 29, Number 4  May 2012

Research Paper

FIGURE 4

FIGURE 3

Study of amount of protein collected at various stages of polymer-protein Experiment performed capacity analysis for various conditions in batch. with 500 mg ml1 LacI-His6-GFP; Experiment performed with 500 mg ml1 without charging the polymer with copper(II) ions (negative Experiment performed with 500 mg ml1 of non-histidine tag control); b-Lactoglobulin (negative control).

adsorbent was recovered during the elution stage, whereas LacIHis6-GFP did not bind to adsorbents in the absence of Cu2+. Similar experiments were also performed with another protein, b-Lactoglobulin, that did not contain a polyhistidine-tag. Only a very low level

of b-Lactoglobulin ([C]o = 500 mg ml1) was detected in the eluate, consistent with a low degree of non-specific binding (Fig. 3). Batch protein binding studies were carried out with a second polyhistidine-tagged protein to confirm the selectivity of Cu2+ immobilised adsorbents. Figure 4 shows the comparison for LacI-His6-GFP ([C]o = 9.5 mg ml1) and a-Synuclein-His8-YFP ([Co] = 9.5 mg ml1) adsorption on adsorbents. The lower amount of LacI-His6-GFP that bound to the adsorbent as compared to

(c) 180 160 140 120

Total volume flow rate -1 Vf /µl s

100 80

1.50 0.66 0.47 0.30

60 40 20 50

[LacI-His6-GFP]/µg ml-1

[LacI-His6-GFP]/µg ml-1

(a)

Comparative study of amount of histidine-tagged protein collected at various stages of polymer-protein capacity analysis for LacI-His6-GFP and aExperiment performed with 9.5 mg ml1 Synuclein-His8-YFP in batch. LacI-His6-GFP; Experiment performed with 9.5 mg ml1 a-SynucleinHis8-YFP.

30 25

Total volume flow rate -1 Vf /µl s

20 15

1.50 0.66 0.47 0.30

10 5

100 150 200 250 300 350 400

1000

2000

3000

4000

5000

6000

Time/s

(d)

180 160 140

Total volume flow rate -1 Vf /µl s

120 100

1.50 0.66 0.47 0.30

80 60 40 20 200

400

600

Time/s

800

1000

[LacI-His6-GFP]/[LacI-His6-GFP]O

[LacI-His6-GFP]/µg ml-1

(b)

Time/s 1.0 0.8

Total volume flow rate -1 Vf /µl s

0.6

1.50 0.66 0.47 0.30

0.4 0.2

5

10

15

20

Amount of Protein Loaded (qd)/µg

FIGURE 5

(a)–(c) Plot of LacI-His6-GFP concentrations against time at flow rates of 1.5 ml s1, 0.66 ml s1, 0.47 ml s1 and 0.3 ml s1 for (a) loading, (b) wash and (c) elution stages. (d) Plot of normalised concentration of LacI-His6-GFP against amount of protein adsorbed. 498

www.elsevier.com/locate/nbt

New Biotechnology  Volume 29, Number 4  May 2012

RESEARCH PAPER

TABLE 1

Properties of poly(HEMA-co-HPIDA-co-EGDMA) synthesised using photopolymerisation compared with commercial Sepharose HisTrap FF column specifications obtained from GE Healthcare and Bhut et al. [19] HisTrap FF column resin (GE Healthcare)

Poly(HEMA-co-HPIDA-co-EGDMA) monolith

Chelating group Matrix Particle size Surface area Metal ion capacity (Cu2+) Binding capacity LacI-His6-GFP (274 kDa) a-Synuclein-His8-YFP (30 kDa)

Iminodiacetic acid Highly cross-linked agarose 34 mm (extremely porous) 50 m2 g1 12.07 mmol (ml of resin)1 3.580 mg (ml of resin)1 8.849 mg (ml of resin)1

Iminodiacetic acid Highly cross-linked polymer Monolith (porosity introduced using 100 mm NH4Cl particles) 4.4 m2 g1 2.417 mmol (ml of monolith)1 0.370 mg (ml of monolith)1 0.802 mg (ml of monolith)1 Research Paper

Specifications

a-Synuclein-His8-YFP could be due to steric hindrance or a difference in the accessibility of histidine-tags for metal interaction, particularly as LacI-His6-GFP can form tetramers with a total molecular weight of 274 kDa, whereas a-Synuclein-His8-YFP is a smaller molecule with molecular weight of 30 kDa. In comparison to a commercial Sepharose HisTrap FF IMAC adsorbent (GE Healthcare) the metal ion and protein binding capacities obtained for LacI-His6-GFP and a-Synuclein-His8-YFP with poly(HEMA-co-HPIDA-co-EGDMA) adsorbents were lower by up to a factor of 10 (Table 1). This was expected due to the approximately 10 fold lower surface area present in the adsorbents compared to the commercial Sepharose resin [19] (Table 1).

Microfluidic chromatography A metal ion chelating adsorbent was fabricated inside a glass microchannel by photopolymerisation. The pressure drop across the 1.5 cm adsorbent (pore diameter 100 mm) inside the microchannel was always less than 4 kPa cm1, with flow rates in the range 0.3–1.5 ml s1. The adsorbent was fed with LacI-His6-GFP (170 mg ml1) at flow rates of 0.3, 0.47, 0.66 and 1.5 ml s1 and the protein concentration in the flow through measured to determine the dynamic binding capacity as a function of flow rate. A shift in the load, wash and elution curves with decreasing flow rates, indicating increasing residence time, was observed as shown in Fig. 5a–c. Dynamic binding capacity (DBC) values for LacI-His6GFP at 10% of the breakthrough curve (Fig. 5d) were constant at 0.319 mg (ml of adsorbent)1 for low flow rates of 0.3 and 0.47 ml s1 and were of comparable value to the equilibrium binding capacity (EBC) of 0.370 mg (ml of adsorbent)1. At higher flow rates of 0.66 and 1.5 ml s1, DBC values were lower at 0.154 and 0.148 mg (ml of adsorbent)1, respectively. The flow experiments were repeated using a-Synuclein-His8-YFP (170 mg ml1) at a flow rate of 0.47 ml s1. The EBC (0.802 mg (ml of adsorbent)1) and DBC (0.543 mg (ml of adsorbent)1) of aSynuclein-His8-YFP were found to be almost double at a flow rate of 0.47 ml s1, compared to LacI-His6-GFP. SDS-PAGE analysis was performed to confirm the protein adsorption and desorption at various stages of chromatography (Fig. 6a). Tetrameric LacI-His6-GFP of 274 kDa can be detected in SDS-PAGE in its reduced 68.5 kDa monomeric form. Although some of the LacI-His6-GFP can be seen in the first wash step it is evident that there is binding of the His-tagged protein to the IMA microfluidic channel and subsequent elution in the later elution steps. Similarly the SDS-PAGE Fig. 6b demonstrates binding and elution of the 30 kDa a-Synuclein-His8-YFP.

FIGURE 6

(a) SDS-PAGE silver stain gel obtained for fractions of LacI-His6-GFP separation. Samples collected at following stages of operation. L: loading, W1: wash fraction at time 184 s, W-2: wash fraction at time 208 s, W-3: wash fraction at time 237 s, W-4: wash fraction at time 431 s and E-1: elution fraction at time 1040 s, E-2: elution fraction at 1190 s. (b) SDS-PAGE silver stain gel obtained for fractions of a-Synuclein-His8-YFP sample collected at following stages of operation. L: loading, W-1: wash fraction at time 239 s, W2: wash fraction at time 476 s, W-3: wash fraction at time 1132 s, E: elution at 2532 s. www.elsevier.com/locate/nbt

499

RESEARCH PAPER

Figure 7a shows an image of the microfluidic channel used in these analyses. Confocal analysis was carried out to observe the moving protein front in the column as illustrated in Fig. 7b–g. The column was loaded with the protein LacI-His6-GFP (170 mg ml1) at a flow rate of 0.3 ml s1 and confocal images were taken at various times to show the progressive movement of the adsorbed LacI-His6-GFP protein front through the field of view during loading (Fig. 7c–e). Figure 7f shows that the protein

New Biotechnology  Volume 29, Number 4  May 2012

remained bound to the adsorbent in the field of view after washing with buffer but was removed completely following elution. The purification of LacI-His6-GFP from a crude cell lysate was tested next. Figure 8a shows the assayed protein concentration in the elution fractions obtained during this separation. Figure 8b is an SDS-PAGE analysis of the fractions collected. This gel was over-developed to visualize the bound protein resulting in low

Research Paper FIGURE 7

Confocal microscopy images showing LacI-His6-GFP fluorescence superimposed on the optical transmission image of a section of porous poly(HEMA-co-HPIDAco-EGDMA) adsorbent (NH4Cl as porogen) at various stages of meso-chromatography. Images show: (a) the porous monolith inside the micromachined glass microfluidic channel (b) before loading LacI-His6-GFP at z = 0 cm, (c)–(e) during loading when the front of adsorbed protein had progressed to (c) z = 0.05 cm, (d) z = 0.11 cm and (e) z = 0.2 cm, (f) after washing and (g) after elution. 500

www.elsevier.com/locate/nbt

New Biotechnology  Volume 29, Number 4  May 2012

RESEARCH PAPER

channel due to cell fragments in the lysate was occasionally observed. This may be reduced by introducing a pre-filter step.

FIGURE 8

(a) Plot of LacI-His6-GFP lysate concentration against time at a flow rate of 1.33 ml s1 for the elution stage. (b) SDS-PAGE silver stain gel obtained for fractions of sample collected at three main stages of operation, L: loading, W1: first wash fraction, W-2: second wash fraction, W-3: third wash fraction, W-4: fourth wash fraction, E: elution. This gel was over-developed to visualize the bound protein resulting in low definition in the lanes containing highly concentrated cell lysate.

definition in the lanes containing highly concentrated cell lysate. A clean band of 68.5 kDa monomeric LacI-His6-GFP can be seen in the elution fraction. The yield of LacI-His6-GFP from 2 ml of crude cell lysate was 2.67 mg. This demonstrates that Histagged proteins can be successfully purified from crude cell lysate using the microfluidic device. Some clogging of the microfluidic

Fabrication of a micro-adsorbent device and its use for microlitre scale IMA chromatography has been demonstrated. A metal chelating ligand (IDA) containing monomer, HPIDA, was synthesised and used to fabricate the metal chelating poly(HEMA-co-HPIDAco-EGDMA) micro-adsorbent device. Adsorbents were optimised for metal chelation capacity (2.42 mmoles (ml of adsorbent)1) and also for protein binding capacity (0.370 mg (ml of adsorbent)1) for LacI-His6-GFP and (0.802 mg (ml of adsorbent)1) for a-Synuclein-His8-YFP. The resulting metal ion and protein binding capacities were lower than a commercially available resin, by a factor of 10 as was expected from the tenfold lower measured surface area. The ability to fabricate these micro-flow devices in situ and their reduced diffusional mass transfer resistance offers some advantage over existing chromatographic technologies. Batch studies confirmed that polyhistidine-tagged proteins were selectively bound by the poly(HEMA-co-HPIDA-co-EGDMA) adsorbent and remained adsorbed during the wash stage. Dynamic binding capacities of LacI-His6-GFP and a-Synuclein-His8-YFP were calculated at varying flow rates and were found to be constant at low flow rates. Confocal microscopic imaging showed a steady moving protein front in the column without any channelling. The microfluidic device was also shown to separate LacI-His6-GFP from a crude cell lysate, indicating some ability to process feed streams containing solids. Such a microfluidic chip offers attractive potential for future on-line process control in biopharmaceutical cell culture operations.

Acknowledgements AGD would like to thank the Cambridge Commonwealth Trust and the Bombay Cambridge Trust for funding. NJD would like to thank the Biotechnology and Biological Sciences Research Council-Bioprocessing Research Industry Club for funding. The authors would also like to thank Dr Helder Barbosa and Dr Matthew Cheeks of Cambridge University for their assistance with protein expression and adsorbent production, Dr Anna Hine of Aston University (UK) for the supply of the LacI-His6-GFP plasmid and Dr Gabriele Kaminski and Dr Carlos Bertonicini of University of Cambridge (UK) for gifting the a-Synuclein-His8-YFP protein.

References 1 Gao, J. et al. (2001) Anal. Chem. 73, 2648 2 Peterson, D.S. et al. (2002) Anal. Chem. 74, 4081 3 West, J.A.A. and Satterfield, B.C. (2007) Methods in Molecular Biology. Human Press Inc 4 Freire, S.L.S. and Wheeler, A.R. (2006) Lab Chip 6, 1415 5 Lee, A.P. and Lin, G. (2009) Microfluidics for Biological Applications. Springer LLC 6 Liao, J-L. et al. (1996) Anal. Chem. 68, 3468 7 Hellferich, F. (1961) Nature 189, 1001 8 Porath, J. et al. (1975) Nature 258, 598 9 Wilchek, M. and Chaiken, I. (2000) Humana Press

10 11 12 13 14 15 16 17 18 19

Nieba, L. et al. (1997) Anal. Biochem. 252, 217 Slentz, B.E. et al. (2003) J. Chromatogr. A984, 97 Chuda, K. et al. (2006) Rad. Phys. Chem. 75, 26 Becker, H. and Locascio, L.E. (2002) Talanta 56, 267 Chen, F. and Carey, W.S. (1990) US Patent No. 4913880 Darby, R.A.J. and Hine, A.V. (2005) FASEB J. 19, 801 Ren, D. et al. (2004) J. Chromatogr. 1031, 87 Bradford, M.M. (1976) Anal. Biochem. 72, 248 Ghose, S. et al. (2007) Biotechnol. Bioeng. 96, 768 Bhut, B. et al. (2010) J. Chromatogr., A 1217, 4946

www.elsevier.com/locate/nbt

501

Research Paper

Conclusions