Anal. Chem. 2004, 76, 6941-6947
Integrated Lectin Affinity Microfluidic Chip for Glycoform Separation Xiuli Mao,†,§ Yong Luo,† Zhongpeng Dai,† Keyi Wang,‡ Yuguang Du,† and Bingcheng Lin*,†
Dalian Institute of Chemical Physics, Chinese Academy of Sciences, China, Institute of Biochemistry and Cell Biology, Shanghai Academy of Life Sciences, Chinese Academy of Sciences, China, and Graduate School of the Chinese Academy of Sciences
Lectin affinity chromatography was miniaturized into a microfluidic format, which results in improvement of performance, as compared to the conventional method. A lectin affinity monolith column was prepared in the microchannel of a microfluidic chip. The porous monolith was fabricated by UV-initiated polymerization of ethylene dimethacrylate (EDMA) and glycidyl methacrylate (GMA) in the presence of porogeneities, followed by immobilization of pisum sativum agglutinin (PSA) on the monolith matrix. Using electroosmosis as the driven force, lectin affinity chromatographies of three kinds of glycoprotein, turkey ovalbumin (TO), chicken ovalbumin (CO), and ovomucoid (OM), were carried out on the microfluidic system. All the glycoproteins were successfully separated into several fractions with different affinities toward the immobilized PSA. The integrated system reduces the time required for the lectin affinity chromatography reaction to ∼3%, thus, the overall analysis time from 4 h to 400 s. Only 300 pg of glycoprotein is required for the whole separation process. Moreover, troublesome operations for lectin affinity chromatography are simplified. The carbohydrate moiety of a glycoprotein may affect the immunogeneity, half-life, bioactivity, and stability of a potential therapeutic product.1 Protein glycosylation can occur at two or more positions in the amino acid sequence, and the glycans at even a single position may be heterogeneous or may be missing from some molecules. This leads to the populations of glycosylated variants of a single protein, usually referred to as glycoforms, whose relative proportions are found to be reproducible and not random. However, the glycoforms may be affected by several factors, including the environment in which the protein is glycosylated, the manufacturing process, and the isolation procedures.2 Therefore, glycoform separation and resolution are critical to understand the structure, function, and heterogeneity of the glycosyl groups on proteins.3 * To whom correspondence should be addressed. Dalian Institute of Chemical Physics, Chinese Academy of Sciences, 457 Zhongshan Road, Dalian, 116023, China. Fax +86-411-8437-9065. E-mail:
[email protected]. † Dalian Institute of Chemical Physics. ‡ Shanghai Academy of Life Sciences. § Graduate School of the Chinese Academy of Sciences. (1) Hooker, A. D.; James, D. C. Mol. Biotechol. 2000, 14, 241-249. (2) Rassi, Z. E.; Mechref, Y. Electrophoresis 1996, 17, 275-301. 10.1021/ac049270g CCC: $27.50 Published on Web 10/30/2004
© 2004 American Chemical Society
Lectins by definition are multivalent proteins of nonimmune origin that bind to sugars rather specifically and agglutinate cells.4 The ability of lectins to probe variations in carbohydrate structures on cell surface glycoproteins and glycolipids has made them a paradigm for protein/carbohydrate recognition. Immobilized lectins have been used to select glycoprotein from biological extracts on the basis of the glycosylation found in the protein.5 Lectin column has also been particularly useful in characterization of glycoforms and oligosaccharides from glycoproteins.4,6-9 Highmannose-type, complex-type, and hybrid-type sugar chains have been fractionated by using immobilized lectin columns in series, each with a different binding affinity for the various glycoforms.10 Fractionation within a class can be accomplished by gradient elution of a specific column with increasing concentrations of a displacing agent.11 Lectin affinity chromatography (LAC)5,6 is one of the most useful techniques. LAC can separate the glycoforms into fractions with different affinity abilities toward the lectin immobilized on the columns, which is useful in the determination of glycans’ alteration of the glycoprotein correlated to disease states, but the long-time and complex operation process limits its applications. Other techniques, such as affinity-reversed micellar extraction12 and capillary zone electrophoresis,13 have been developed to separate the glycoproteins; however, information of the glycan contained in the glycoforms was lost.14 An integrated microfluidic chip seems to be an effective platform to take advantage of these techniques and give more information. Owing to the short analysis time and the feasibility for multichannel or high-throughput analysis, the microfluidic chip (3) Kishino, S.; Di, Z.; Sugawara, M.; Iseki, K.; Migazaki, K. J. Chromatogr. 1992, 582, 246-248. (4) Satish, P. R.; Surolia, A. J. Biochem. Biophys. Methods 2001, 49, 625-640. (5) Alperin, D. M.; Latter, H.; Lis, H.; Sharon, N. Biochem. J. 1992, 285, 1-4. (6) Yoshida, K.; Morguchi, H.; Suni, S.; Horimi, H.; Kitahara, S.; Umeda, H. J. Chromatogr., B 1999, 723, 75-80. (7) Yoshida, K.; Sumi, S.; Honda, M.; Hosoya, Y.; Yano, M.; Arai, K.; Ueda, Y. J. Chromatogr., B 1995, 672, 45-51. (8) Yoshikawa, K.; Umetsu, K.; Shinzawa, H.; Yuasa, I.; Yamashita, K.; Suzuki, T. FEBS Lett. 1999, 458, 112-116. (9) Sumi, S.; Arai, K.; Kitahara, S.; Yoshida, K. J. Chromatogr., B 1999, 727, 9-14. (10) Yamamoto, K.; Tsuji, T.; Osawa, T. Mol. Biotechol. 1995, 3, 25-36. (11) Iturbe, S.; Narasimhan, S.; Merrick, J. M.; Falk, J. A.; Letarte, M. J. Immunol. 1986, 136, 4588-4595. (12) Choe, J.; Zhang, F.; Wolff, M. W.; Murhammer, D. W.; Linhardt, R. J.; Dordick, J. S. Biotechnol. Bioeng. 2000, 70 (5) 484-490. (13) Suzuki, S.; Honda, S. Electrophoresis 1998, 19, 2539-2560. (14) Che, F. Y.; Song, J. F.; Shao, X. X.; Wang, K. Y.; Xia, Q. C. J. Chromatogr., A 1999, 749, 599-608.
Analytical Chemistry, Vol. 76, No. 23, December 1, 2004 6941
has been becoming one of the most powerful analytical tools in recent years.15 Protein immobilization techniques have been recently extended into microfluidic format for various bioanalytical procedures, such as immunoabsorption,16-17 enzymatic reactions,18-20 and so on. Various strategies of protein immobilization have also been demonstrated, including chemical modification of the microchannel surface,16-17,21 packing of biochemical-coated beads,22-25 and packing with a biochemical-bearing porous monolithic slab.18 To our knowledge, microfluidic chip-based glycoform analysis by lectin affinity chromatography has not been reported. In the present paper, an electrokinetically controlled microfluidic system with a lectin affinity monolith column fabricated inside the microchannel is demonstrated as a potential useful platform for glycoform separation and analysis. Experimental parameters, including composition of the monolith material and buffer systems were optimized, and the performance of the pisum sativum agglutinin affinity microfluidic chip (PSA-AMC) was demonstrated by separation of chicken ovalbumin (CO), ovomucoid (OM), and turkey ovalbumin (TO) into different fractions of glycoforms. EXPERIMENTAL SECTION Materials. All chemicals used in this study were of analytical grade. Turkey ovalbumin (TO, A7269), chicken ovalbumin (CO, A2512), ovomucoid (OM, T9253), and human serum albumin (HSA, A1887) were purchased from Sigma (St. Louis, MO). Glycidyl methacrylate (GMA), ethylene dimethacrylate (EDMA), cylcohexanol, decyl alcohol, and dodecanol were purchased from Acros Organics (Geel, Belgium), Methyl-R-D-mannopyranoside (MM) was from ICN Biochemicals (Cleveland, OH), and 5-(6)carboxytetramethylrhodamine succinimidyl ester (TAMRA SE) was from Molecular Probe (Eugene, OR). Pisum sativum agglutinin (PSA) was purchased from Shanghai Dongfeng Biotechnology Company (Shanghai, China). Instrumentation. An Oriel UV illumination serial 68810 ARC fitted with a 200-500-W Hg lamp was used for UV initiation of the polymerization reaction. Microfluidic chip-based separations were performed on a Microfluide Tool Kit (Micralyne, Edmonton, AB, Canada). The separations were monitored by a laser-induced fluorescence system with a 5-mW laser with an excitation wavelength of 532 nm. Scanning electron microscope images were collected using a JSM-6360LV high-resolution analytical scanning electron microscope (Japan). The glass chips used in the experi(15) Lin, C. C.; Lee, G. B.; Chen, S. H. Electrophoresis 2002, 23, 3550-3557. (16) Dodge, A.; Fluri, K.; Verpoorte, E.; de Rooij, N. F. Anal. Chem. 2001, 73, 3400-3409. (17) Linder, V.; Verpoote, E.; Thormann, W.; de Rooij, N. F.; Sigrist, H. Anal. Chem. 2001, 73, 4181-4189. (18) Peterson, D. S.; Rohr, T.; Svec, F.; Frechet, J. M. J. Anal. Chem. 2002, 74, 4801-4808. (19) Mao, H.; Yang, T.; Cremer, P. S. Anal. Chem. 2002, 74, 379-385. (20) Lazar, I. M.; Ramsey, R. S.; J. Ramsey, M. Anal. Chem. 2001, 73, 17331739. (21) Lahann, J.; Choi, I. S.; Lee, J.; Jensen, K. F.; Langer, R. Angew. Chem., Int. Ed. 2001, 40, 3166-3169. (22) Huang, T. B.; Peterz-Luna, V. H.; Schreyer, B.; Sklar, L. A.; Lopez, G. P. Anal. Chem. 2002, 74, 1149-1156. (23) Hostis, E. L.; Michel, Ph. E.; Fiaccabrino, G. C.; Strike, D. J.; de Rooij, N. F.; Koudelka-Hep, M. K. Sens, Actuators, B 2000, 64, 156-162. (24) Andersson, H.; Van der Wijngaart, W.; Enoksson, P.; Stemme, G. Sens. Actuators, B 2000, 67, 203-208. (25) Malmstadt, N.; Yager, P.; Hoffman, A. S.; Stayton, P. S. Anal. Chem. 2003, 75, 2943-2949.
6942 Analytical Chemistry, Vol. 76, No. 23, December 1, 2004
Figure 1. Layout of the microfluidic chip: 1, running buffer reservoir (R1); 2, eluent buffer reservoir (R2); 3, sample reservoir (R3); 4, sample waste reservoir (R4); 5,6, washing reservoir; 7, waste reservoir (R7). Table 1. Compositions of the Polymerization Mixtures for the Monolith Matrix
glycidyl methacrylate (v/v %) ethylene dimethacrylate (v/v %) cylcohexanol (v/v %) decyl alcohol (v/v %) dodecanol (v/v %) 2,2-dimethoxy-2-phenylacetophenone (w/v %)
A
B
C
24 15 54 6 0 1
24 15 0 0 60 1
24 15 0 0 60 1
ments were homemade by standard photolithography and wet chemical etching techniques. The cross section of the microchannel was 70 µm × 20 µm. The layout of the chip is shown in Figure 1. In addition to the primary reservoirs required, reservoirs 5 and 6 were added on our chip for the microchannel washing to get stable and reproducible EOF (Figure 1). PSA Monolith Preparation. A two-step procedure was adopted to fabricate the PSA affinity column. A porous monolith matrix was synthesized by UV-initiated polymerization, and then PSA was immobilized on the monolith matrix. Prior to the fabrication of the monolith, the wall of the microchannels was pretreated with 3-(trimethoxysilyl) proxyl methacrylate (binding silane) following the procedures described by Ericson et al.26 Briefly, the microchannels were flushed with 0.2 mol/L NaOH for 30 min; washed with water, followed by 0.2 mol/L HCl for 30 min; then washed with water and acetone and dried at 150 °C for 12 h. Then the microchannels were filled with a 30% (v/v) 3-(trimethoxysilyl) proxyl methacrylate solution in acetone. The device was sealed and kept overnight at room temperature and then rinsed with acetone and dried. The microchannels were filled with the monolith material mixture (A, B, or C) (Table 1). The sections of the chip that should not contain a monolith were covered with an opaque mask. The length of the unmasked area was kept at 500 µm. The polymerization was initiated by exposing the masked chip to UV light (500 W) for 6 min. After the completion of the reaction, methanol was passed (26) Ericson, C.; Liao, J. L.; Nakazato, S.; Hjerten, S. J. Chromatogr., A 1997, 767, 33-41.
Figure 2. Operation sequence of PSA-AMC. The arrows indicate the movement of solutions.
through the monolith column to remove the porogenic solvents and other unreacted soluble compounds by a vacuum device. The washing process was continued for 2 h. Monoliths were also prepared in a 50-µm-i.d. capillary using the same conditions as above. These monolith columns were subjected to a scanning electron microscope (SEM). Immobilization of lectins onto the monolith matrix took place via the epoxy groups found naturally in the polymer structure of the monolith material and the -amino groups of the lectins.27 The microchannel containing monolith was washed with increasing polar solvents (methanol, methanol/water, 1:1 (v/v)), then filled with a 100 mM carbonate buffer (pH 9.5) and equilibrated for 2 h before a 5.0 mg/mL solution of PSA in the same carbonate buffer was introduced. The immobilization reaction was allowed to proceed for 16 h at 30 °C. The microchannels were washed with the same carbonate buffer for 1 h. The extra epoxy groups were subsequently blocked by reacting with 1 M ethanolamine for 1 h at room temperature. The microchannels were thoroughly washed with running buffer. The microchannels beside the PSA monolith column were washed with 1 M NaOH. Glycoprotein Labeling and Purification. A 100-µL portion of 20 mg/mL glycoprotein (or HSA) in 0.1 M carbonate buffer (pH 8.3) was mixted with 10 µL of 10 mg/mL TAMRA SE (in DMF), and the reaction was allowed to continue for 8 h in darkness. Labeled glycoprotein was separated from the excess dye with a separation column containing Sephadex G-50 using 50 mM amine bicarbonate (pH 8.0) as eluent buffer. Glycoprotein (or HSA) concentration was determined by the absorption OD value at 280 nm with a 6505 UV/vis spectrophotometer (JENWAY). Operation Procedures. Once the monolith was synthesized and PSA was immobilized on it, the chip was ready to be tested using the procedure illustrated in Figure 2. Arrows indicate the movement of solution during each step, with voltages being applied on the reservoirs at the end of each flow path. Microfludic chip-based lectin affinity chromatography was similar to other lectin affinity techniques. In this study, a mixture of heterogeneous glycoprotein (TO, CO, or OM) was driven through the monolith matrix with PSA immobilized on it. The glycoforms with terminal mannosyl residues or N-acetylchitobiose fucose structure were (27) Berruex, L. G.; Freitag, R.; Tennikova, T. B. J. Pharm. Biomed. Anal. 2000, 24, 95-104.
Table 2. Voltages Applied on the Chip in Each Step voltage (V) step
time (s)
R1
R2
R3
R4
R7
sample injection 1 injection 2 + wash elution 1 elution 2
15 100 100 100
310 2000 float float
float float 2000 2000
400 1400 1000 1000
ground 1400 1000 1000
750 ground ground ground
adsorbed onto the PSA monolith because of their interaction with immobilized PSA, whereas other glycoforms were washed off by the running buffer and detected. Running buffer washing was continued until no more glycoform was detected. The adsorbed glycoforms were eluted from the matrix using a displacing sugar with a competent structure (methyl-R-D-mannopyranoside in this study). The whole procedure for PSA-AMC experiments contained five steps shown in Figure 2, that is, sample injection 1, sample injection 2, washing, elution 1, and elution 2. The procedure was carried out four times for each glycoprotein. The processes were continuous, except for the switch of applied voltage from the running buffer reservoir to the eluent buffer reservoir between the washing and elution steps. Table 2 presents the voltages applied in each step for the operation of PSA-AMC in detail. The detection point was 3 mm below the monolith column. The compositions of running buffer and eluent buffer are shown as follows. Running buffer: 20 mM Hepes buffer (pH 7.49) + 0.1% Triton X-100 + 1 mM MnCl2 + 1 mM CaCl2; Eluent buffer 1: 20 mM Hepes buffer (pH 7.49) + 0.1% Triton X-100 + 1 mM MnCl2 + 1 mM CaCl2 + 10 mM MM; Eluent buffer 2: 20 mM Hepes buffer (pH 7.49) + 0.1% Triton X-100 + 1 mM MnCl2 + 1 mM CaCl2 + 500 mM MM. Chip-based capillary zone electrophoresis of glycoprotein was also carried out for comparison purposes. RESULTS AND DISCUSSION Preparation of Monolith Matrix. The compositions of the polymerization mixtures used in this study are presented in Table 1. UV-initiated polymerization was performed to get a microscale and a localized monolith column. The polymerization process was completed within 6 min. Figure 3 shows a monolith matrix of 500µm length in the microchannel and the clear brim of the monolith column. Analytical Chemistry, Vol. 76, No. 23, December 1, 2004
6943
Table 3. Flow Rates of the Buffer Systems
Figure 3. 500-µm monolith matrix in the microchannel.
Figure 4. SEM micrographs of the internal structure of monoliths A, B, and C.
Figure 4 shows the SEM micrographies of the internal structure of monoliths A, B, and C, corresponding to the monolith material mixtures A, B, and C in Table 1. It was observed that the composition of the polymerization mixture has a great impact on the performance of the monolith columns. Three porogeneities were selected to get different pore diameters. As can be seen, the pore structure of monolith C is larger than those of monoliths A and B. Experimental results showed that desorption of glycoprotein from either monolith A or monolith B was more difficult than that from monolith C; hence, the broadened peaks of components were obtained using monolith A and B. The porogeneities of monoliths A and B are cylcohexanol and decyl alcohol, which form smaller pores and result in a larger surface area. Thus, it is predictable that more lectin was immobilized on the monolith matrixes A and B. When desorption steps were carried out, more time was needed to displace the bound glycoprotein with MM. In the following experiments, polymerization mixture C was selected. Buffer Systems. According to the operation procedure of PSAAMC, three buffer systems, running buffer, eluent buffer 1, and eluent buffer 2, were prepared. Electrokinetic driving was used in this study, since it is simple and does not need an extra device. 6944 Analytical Chemistry, Vol. 76, No. 23, December 1, 2004
buffer system
νeof (cm/s)
t (s)
running eluent buffer 1 f running buffer eluent buffer 2 f eluent buffer1
0.1162-0.3046 0.0969-0.1632 0.0839-0.1414
5-13 9-15 10-17
To get enough delivery rates of sample and buffers, high voltages were applied on the chip. Lower ionic strength buffer, 20 mM Hepes was used to decrease the Joule heat. Most lectins have a hydrophobic binding site, and thus, hydrophobic interaction will occur between lectin and glycoproteins to be separated, which often results in the nonspecific adsorption of separated glycoproteins. So the nonspecific interaction between lectin and polypeptides of glycoproteins should be eliminated. In general, nonionic surfactants, such as Triton X-100, Tween 20, and NP-40 have no effects on the activity of PSA and can eliminate the nonspecific bindings between lectin and polypeptides. In this study, Triton X-100 was selected and added into the running and eluent buffers.21 The buffers provide the affinity binding environment of carbohydrate and immobilized PSA. Metal ions, Ca2+ and Mn2+, should be added into the buffers, since they are essential for PSA activity. AS a result, the composition of the running buffer was 20 mM Hepes buffer (pH 7.49) + 0.1% Triton X-100 + 1 mM MnCl2 + 1 mM CaCl2. MM was used as the displacing reagent for desorption of glycoprotein. To get the fractions of glycoprotein with different affinity abilities toward PSA, two solutions with different concentrations of MM were used. The eluent buffer 1, with 10 mM MM added in the running buffer, was used to displace the fraction that has weak binding affinity toward immobilized PSA, and the eluent buffer 2, with 500 mM MM added in the running buffer, was used to displace the fraction that has strong binding affinity toward immobilized PSA.21 Characterization of Chip. The vinylization of the microchannel wall with 3-(trimethoxysilyl) proxyl methacrylate reduces the available charge sites for generation of EOF and consequently represses EOF. NaOH was used to wash the microchannels in addition to the monolith column to get enough EOF for the reagent to deliver. Measurement of EOF with running buffer using the current monitoring method described by Huang28 reveals that the EOF of the microchannel is near that of the untreated microchannel (Table 3). When MM was added to the running buffer and supplied for the elution, the current increased until the buffer filled the entire flow path, at which point the current stabilized, so the apparent EOF of the system in which the eluent buffer 1 is displacing the running buffer was lower than that of the running buffer and higher than that of eluent buffer 1 alone. The same result was obtained for the system in which eluent buffer 2 displaced eluent buffer 1. Table 3 presents the apparent EOF of the three systems used in the experiment. The EOF is from 2.6 × 10-4 to 3.6 × 10-4 cm2/V‚s. The electric field applied on reservoirs is from 323 to 544 V/cm in washing and elution processes. Flow rates in these buffer systems are from 0.0839 to 0.3046 cm/s. So the components can be detected within 20 s, and the range of the flow rate provides efficient reagent delivery. (28) Huang, X.; Gordon, M. J.; Zare, R. N. Anal. Chem. 1988, 60, 1837-1838.
Figure 5. Electropherograms of HSA detected above and below the PSA monolith column on PSA-AMC: (a) electropherogram of HSA by PSA-AMC detected above the PSA monolith column; (b) electropherogram of HSA by PSA-AMC detected below the PSA monolith column. The x axis is the time used for operation. The steps of washing and elution were continuous except for the interruption for a voltage switch or change of eluent buffers. The operation conditions are shown in the text.
Nonspecific Adsorption of PSA-AMC. The nonglycoprotein, human serum albumin (HSA), was used as a negative control to evaluate the nonspecific adsorption of PSA-AMC. Figure 5 presents the electropherograms of HSA detected above and below PSAmonolith column. The recycle rate of HSA passed through PSAAMC is 97% according to the peak areas. No protein was detected when eluted with both eluent buffer 1 and eluent buffer 2 (Figure 5 b). It shows that TAMRA SE-labeled HSA had no adsorption on the PSA monolith matrix and that the adsorption of glycoform on the PSA-AMC is the specific interaction between glycan and PSA. Reproducibility and Glycoform Separation. The injection volume was estimated to be ∼98 pL according to the dimensions of the microchannel, and the concentrations of the CO, OM, and TO were 3.0, 3.0, and 1.5 mg/mL, repectively, so the whole analytical process for glycoform separation needs
