Fabrication of Carbohydrate Microarrays on Gold Surfaces: Direct

This paper describes a new and simple microarray platform for presenting multiple ... The biggest hurdles in employing gold-based substrate for fluore...
0 downloads 0 Views 524KB Size
Download PDF
Anal. Chem. 2006, 78, 4786-4793

Fabrication of Carbohydrate Microarrays on Gold Surfaces: Direct Attachment of Nonderivatized Oligosaccharides to Hydrazide-Modified Self-Assembled Monolayers Zheng-liang Zhi,*,† Andrew K. Powell,† and Jeremy E. Turnbull

Molecular Glycobiology Laboratory, School of Biological Sciences, University of Liverpool, Crown Street, Liverpool, L69 7ZB, UK

This paper describes a new and simple microarray platform for presenting multiple nonderivatized oligosaccharides to protein targets, with utility for mapping carbohydrate-protein recognition events. The approach is based on the creation of a hydrazide-derivatized, selfassembled monolayer on a gold surface in a single or twostep procedure, for efficient and selectively oriented anchoring of oligosaccharide probes via their reducing ends, with detection using fluorescence detection of bound proteins. The biggest hurdles in employing goldbased substrate for fluorescence-based microarray detection include fluorescence quenching and nonspecific surface adsorption of proteins. We found that the quenching effect could be minimized by introducing a ω-thiolated fatty acid (C16) self-assembled monolayer between the gold surface and hydrazide groups, followed by detection involving three successive binding protein layers covering the gold surface. In addition, an effective blocking scheme involving poly(ethylene glycol) aldehyde and bovine serum albumin was employed to reduce nonspecific protein adsorption to the chip surface. As proof of principle, we demonstrate here that sulfated oligosaccharide probes from heparin can be effectively and covalently attached without prior derivatization onto the hydrazide-modified, self-assembled monolayer on gold-coated slide surfaces in a microarray format. This platform is used to assess binding of specific heparin-binding protein targets at very high sensitivity, and we also demonstrate that the approach can be extended to nonsulfated sugars. Direct attachment of nonderivatized sugar probes on the chip is advantageous since it avoids the need for laborious prederivatization and cleanup steps. This versatile fluorescence microarray platform provides a facile approach for interrogating multiple carbohydrate-protein interactions in a high-throughput manner and has potential as a common gold surface platform for other diverse interrogations by MALDI-MS, surface plasmon resonance, and quartz crystal microbalances. Carbohydrate-protein interactions are critical in many important biological events, in particular on cell surfaces. Carbohydrate 4786 Analytical Chemistry, Vol. 78, No. 14, July 15, 2006

microarrays, which display many structurally diverse oligosaccharide probes on a single chip substrate, could be used to map out interactions of carbohydrates and proteins in a highthroughput manner.1-3 A key challenge in the establishment of carbohydrate microarrays is the development of reliable and reproducible chemistries for the immobilization of chemically and structurally diverse carbohydrate probes onto a suitable substrate with retention of their functionality. The majority of carbohydrate microarrays so far reported have used derivatized oligosaccharide probes to link covalently to the appropriately modified chip surfaces. For instance, Park et al. fabricated carbohydrate microarrays via chemoselective ligation between maleimide-linked sugars and thiol-derivatized glass surface.4,5 Alternatively, Seeberger et al.6 and Mrksich et al.7 used maleimide-terminated, selfassembled monolayers (SAM) as a platform for the immobilization of thiol-derivatized carbohydrate. In addition, Houseman and Mrksich also reported construction of a carbohydrate microarray on a gold surface using the Diels-Alder-mediated reaction between SAM presenting benzoquinone groups and cyclopentadiene-derivatized saccharides.8 Blixt et al.9 and Xia et al.10 constructed glycan microarrays by either coupling amine-functionalized glycans or glycan-2,6-diaminopyridine conjugates to N-hydroxysuccinimide (NHS) activated glass slides. Feizi and Chai have generated a carbohydrate microarray by immobilizing lipidconjugated oligosaccharides onto nitrocellulose-coated surfaces.11 * Corresponding author. E-mail: [email protected]. † These authors contributed equally to the conception and execution of this work. (1) Wang, D. N. Proteomics 2003, 3, 2167-2175. (2) Feizi, T.; Fazio, F.; Chai, W. C.; Wong, C. H. Curr. Opin. Struct. Biol. 2003, 13, 637-645. (3) Shin, I.; Cho, J. W.; Boo, D. W. Comb. Chem. High Throughput Screening 2004, 7, 565-574. (4) Park, S.; Lee, M. R.; Pyo, S. J.; Shin, I. J. Am. Chem. Soc. 2004, 126, 48124819. (5) Park, S.; Shin, I. Angew. Chem., Int. Ed. 2002, 41, 3180-3182. (6) Ratner, D. M.; Adams, E. W.; Disney, M. D.; Seeberger, P. H. ChemBioChem 2004, 5, 1375. (7) Houseman, B. T.; Gawalt, E. S.; Mrksich, M. Langmuir 2003, 19, 1522. (8) Houseman, B. T.; Mrksich, M. Chem. Biol. 2002, 9, 443-454. (9) Blixt, O.; Head, S.; Mondala, T.; Scanlan, C.; Huflejt, M. E.; Alvarez, R.; Bryan, M. C.; Fazio, F.; Calarese, D.; Stevens, J.; Razi, N.; Stevens, D. J.; Skehel, J. J.; van Die, I.; Burton, D. R.; Wilson, I. A.; Cummings, R.; Bovin, N.; Wong, C. H.; Paulson, J. C. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 17033-17038. (10) Xia, B.; Kawar, Z. S.; Ju, T.; Alvarez, R. A.; Sachdev, G. P.; Cummings, R. D. Nat. Methods 2005, 2, 845-850. 10.1021/ac060084f CCC: $33.50

© 2006 American Chemical Society Published on Web 06/16/2006

Recently, Ko et al. fabricated microarrays based on noncovalent fluorous-based interactions.12 Carbohydrate immobilization on the surfaces has also been realized by noncovalent interaction of biotinylated oligosaccharides and surface-bound streptavidin.13,14 However, all these strategies require multiple chemical steps of derivatization and cleanup of each individual carbohydrate probe that makes microarray fabrication a laborious and complex procedure. Alternative techniques for carbohydrate microarray fabrication have used immobilization of diverse glycan structuresswithout prior modificationsonto a nitrocellulose-coated glass slide15 or black polystyrene-coated glass slides.16 These approaches, while apparently functional, may well have severe limitations in terms of their molecular weight dependence and nonspecificity in the immobilization of carbohydrate probes. Moreover, the molecules deposited on the substrate presumably simply adsorb to the surface and display various orientations, many of which are unlikely to be relevant to the natural presentation of glycans on the cell surface. The randomly deposited glycans thus would not be readily available for protein recognition and binding, and detection sensitivity would also be compromised. The condensation reaction between the aldehyde group of a reducing sugar and an amine group to form an imine linkage is well known and has been exploited for attachment of sugars to aminosilane-derivatized glass surfaces through covalent immobilization of unmodified sugars.17 More recently, alternative approaches have been reported involving the use of more reactive functional groups such as hydrazide and aminooxy-modified glass surfaces to construct carbohydrate microarrays.18,19 These strategies, however, introduced a multistep procedure and thus complications of surface modification of glass slides. With the aim of developing a simpler and generic microarray platform for nonderivatized oligosaccharides, we report here a new oligosaccharide immobilization strategy based on a hydrazide-derivatized SAM on a gold chip surface for efficient and selectively oriented coupling of oligosaccharides. Furthermore, we have also focused on establishing microarrays presenting glycans from a specific family of naturally occurring, sulfated linear oligosaccharides, the heparan sulfates (HS). By binding and modulating various proteins (including growth factors and cytokines, enzymes, protease inhibitors, and extracellular matrix proteins), HS acts as a regulator of many aspects of cell behavior, such as cell growth, migration, and differentiation. There is evidence that specific monosaccharide sequences within HS molecules dictate the specificity of its many activities.20 Since the study of structure and (11) Feizi, T.; Chai, W. Nat. Rev. Mol. Cell. Biol. 2004, 5, 582-588. (12) Ko, K.-S.; Jaipuri, F. A.; Pohl, N. L. J. Am. Chem. Soc. 2005, 127, 1316213163. (13) Leteux, C.; Stoll, M. S.; Childs, R. A.; Chai, W.; Vorozhaikina, M.; Feizi, T. J. Immunol. Methods 1999, 227, 109-119. (14) Leteux, C.; Childs, R. A.; Chai, W. G.; Stoll, M. S.; Kogelberg, H.; Feizi, T. Glycobiology 1998, 8, 227-236. (15) Wang, D.; Liu, S.; Trummer, B. J.; Deng, C.; Wang, A. Nat. Biotechnol. 2002, 20, 275-281. (16) Willats, W. G.; Rasmussen, S. E.; Kristen, T.; Mikkelsen, J. D.; Knox, J. P. Proteomics 2002, 2, 1666-1671. (17) Yates, E. A.; Jones, M. O.; Clarke, C. E.; Powell, A. K.; Johnson, S. R.; Porch, A.; Edwards, P. P.; Turnbull, J. E. J. Mater. Chem. 2003, 13, 2061-2063. (18) Lee, M.-r.; Shin, I. Org. Lett. 2005, 7, 4269-4272. (19) Zhou, X.; Zhou, J. Biosens. Bioelectron. 2006, 21, 1451-1458. (20) Yates, E. A.; Guimond, S. E.; Turnbull, J. E. J. Med. Chem. 2004, 47, 277280.

function of HS is hampered by the fact that their synthesis is not template driven, and by the difficulties involved in chemical or biochemical synthesis of this family of carbohydrate structures, we chose naturally derived oligosaccharides (10-mer fractions) from porcine mucosal heparin (PMH) as initial probes to test the feasibility of the hydrazide-based microarray for protein-binding studies with heparin-binding proteins. The potential for desulfation and low reactivity of heparin-type saccharides makes the application of prederivatizing approaches for microarray generation difficult, and direct immobilization of nonderivatized oligosaccharide probes is obviously preferable for handling large numbers of structures, which can be prepared from natural tissue sources. While this platform was optimized particularly for arraying sulfated oligosaccharides, we have also demonstrated that it can also be applied to other types of glycans, including those that are chemically synthesized. This versatile gold-based carbohydrate microarray platform is compatible not only with fluorescence detection, as demonstrated here, but also has potential for other biophysical detection techniques including matrix-assisted laser desorption/ionization time-of-flight mass spectroscopy,21 surface plasmon resonance,22 and quartz crystal microbalance.23 The techniques described here are therefore driven by the goal of producing a versatile platform that can be used with equivalent immobilization approaches but different readouts for binding. This new methodology establishes a flexible platform for direct chemoselective covalent attachment of nonderivatized oligosaccharides to a modified gold chip surface, facilitating carbohydrate microarray construction for interrogation with multiple target proteins using multiple detection strategies in the study of carbohydrate-protein recognition events. EXPERIMENTAL SECTION Materials and Reagents. All chemical and biochemical products were of analytical grade. Bovine serum albumin (BSA) and betaine were supplied by Sigma. Adipic dihydrazide, 16mercaptohexadecanoic acid (MHDA), 11-mecaptoundecanoic acid (MUA), and ethanolamine were purchased from Aldrich. R-Methoxy-ω-formyl poly(ethylene glycol) (MW 750) (PEG-aldehyde) was obtained from Rapp Polymere GmbH. Alexa Fluor 532 goat antimouse IgG (H+L) conjugate and Alexa Fluor 532 donkey antigoat IgG (H+L) were obtained from Molecular Probes (Eugene, OR). C-myc-tagged phage display anti-heparin scFv antibody HS4C3 (the supernatant from bacteria culture) and an anti-C-myc mouse antibody (9E10) were kindly provided by Prof. van Kuppevelt (University of Nijmegen, The Netherlands).24 Glial cell line-derived neurotrophic factor (GDNF) and its corresponding antibody were obtained from R & D Systems. Fibroblast growth factor receptor 2-IgG1-Fc (FGFR2-Fc) recombinant protein was expressed and purified as described previously.25 Anti-IgG Fc was obtained from Pierce. Concanavalin A (Con A), anti-Con A antibody, snowdrop Galanthus nivalis lectin, and anti-G. nivalis lectin were purchased (21) Su, J.; Mrksich, M. Angew. Chem., Int. Ed. 2002, 41, 4715-4718. (22) Karamanska, R.; Mukhopadhyay, B.; Russell, D. A.; Field, R. A. Chem. Commun. 2005, 26, 3334-3336. (23) Zhang, Y.; Telyatnikov, V.; Sathe, M.; Zeng, X.-Q.; Wang, P. G. J. Am. Chem. Soc. 2003, 125, 9292-9293. (24) ten Dam, G. B.; Hafmans, T.; Veerkamp, J. H.; van Kuppevelt, T. H. J. Histochem. Cytochem. 2003, 51, 727-739. (25) Powell, A. K.; Fernig, D. G.; Turnbull, J. E. J. Biol. Chem. 2002, 277, 2855428563.

Analytical Chemistry, Vol. 78, No. 14, July 15, 2006

4787

from Vector (Burlingame, MA). Phosphate-buffered saline tablets (pH 7.3) (PBS) was obtained form Oxoid. Gold-coated microarray slides were obtained from Nunc. Oligosaccharide Probe Preparation. Heparin decasaccharides, R1-3, R1-6-D-mannopentaose (Man5), and R1-6-mannobiose (Man2) were representative oligosaccharide probes of negatively charged and neutral carbohydrates, respectively. Chemically synthesized Man5 and Man2 were obtained from Dextra Laboratories. Heparin decasaccharides were prepared and purified by partial low pH nitrous acid or partial heparitinase III digestion of PMH according to established procedures.26 The oligosaccharides generated were size-separated by gel filtration chromatography, using a Superdex 30 XK16 × 200 column run at 0.5 mL/ min in 0.5 M ammonium hydrogen carbonate. Peaks were detected by measuring absorbance at 232 nm, and 1-mL fractions were collected. Fractions containing size-defined oligosaccharides were pooled, desalted, and freeze-dried. The size-fractionated decasaccharide mixture was then resolved by strong anionexchange high-performance liquid chromatography (SAX-HPLC), according to established procedures.26 Oligosaccharides were detected by absorbance at 232 nm; fractions (0.5 mL) containing major oligosaccharides were pooled, desalted, and free-dried. For preparation of the reduced oligosaccharides, an aliquot of 0.2 mL of 10 mg/mL oligosaccharide was treated with 8 mg of NaBH4 overnight, the mixture was then desalted and freeze-dried, and the treatment was then repeated four times to give a complete conversion of the reducing end of the sugar. The reduction yield (>99%) was verified by SAX-HPLC. Hydrazide Modification of Gold Surface. The gold-coated glass slides were coated with a SAM of MHDA by soaking the slides for 24-48 h in a 0.1 mM solution of MHDA dissolved in isobutyl alcohol. The SAM-covered slides were washed and sonicated for 5 min in ethanol and dried with a nitrogen stream. Hydrazide derivatization was achieved by dropping 1 mL of a mixture of 2 mg/mL 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide and 1 mg/mL adipic dihydrazide in dimethyl sulfoxide (DMSO) onto the slides for 6 h or overnight; the slides were then washed with ethanol and dried with a nitrogen stream. Microarray Printing. Oligosaccharide printing solutions were prepared in 1 M betaine and were spotted on hydrazide-derivatized gold-coated glass slides using a MicroGrid II compact pin-type contact arrayer (Biorobotics) in 65% relative humidity. Betaine was added in the samples to prevent water evaporation from the droplets. The oligosaccharides were typically arrayed as a 10-fold dilution series with starting pickup solution concentration of 1 mg/ mL for 10 replicate spots (with ∼1 nL/spot delivered by contact of the pins with the surface). The distance between the centers of adjacent spots was 400 µm. The printed slides were incubated overnight at 18 °C in a closed environment (a plastic dish sealed with Parafilm). Oligosaccharides that were not bound after spotting and incubation were removed by washing the slides twice with distilled water. Immobilized oligosaccharides were stable to storage for at least several weeks when stored at 4 °C in a sealed container. Blocking of the Surfaces. The printed slides were treated in 0.5 mL of PEG-aldehyde (20 mg/mL) in water and incubated 1 h (26) Turnbull, J. E. In Proteoglycan Protocols; Iozzo, R. V., Ed.; Humana Press: Totowa, NJ, 2001.

4788

Analytical Chemistry, Vol. 78, No. 14, July 15, 2006

at room temperature. The slides were then rinsed with water, treated with 0.5 mL of 3% BSA blocking solution in PBS buffer with 0.01% Tween-20 for 30 min, and washed again with 0.01% Tween-20 solution for 2 min. Thereafter, the binding proteins and the corresponding detection antibodies (each of 100 µL, diluted with 3% BSA in PBS buffer in 0.01% Tween-20) were added sequentially on the slide and incubated for 40 min for each step. The slide was rinsed three times with 10-fold diluted PBS in 0.01% Tween-20 solution for a total period of 15 min between each binding step and after the final step. After the last step of the fluorescence-labeled antibody binding, the slide was rinsed and dried at room temperature under a stream of gaseous N2. Protein Binding. The oligosaccharide microarrays were probed with an anti-heparin antibody, HS4C3, and other heparinbinding proteins including GDNF and FGFR2-Fc. The binding to microarrays produced with Man5 were tested using lectins Con A and snowdrop G. nivalis. The microarray-bound specific proteins were probed with appropriate cognate antibodies, followed by a fluorescence (Alexa Fluor 532)-labeled secondary antibody. For each binding step, an incubating step of 40 min was used. HybriWell (Grace Biolabs) were used to cover the slides during the incubation in order to prevent evaporation. Proteins bound to the gold chip surfaces, once dried, could no longer be removed by use of typical chip regeneration protocols, e.g., 0.1 M glycerin hydrochloride, pH 2.0. This fact, combined with the ease of generation of multiple slide-based microarrays, indicates that reuse of chips is neither recommended nor required. Scanning and Evaluation. The fluorescence signals of the microarrays were read using a Genepix 4000A laser microarray scanner (Axon Instruments) with PMT voltage set at 800 V and laser power at 100%; signals of the ratio of 635 nm/532 nm were quantified using the GenePix Pro 3.0 image analysis software package. Mean signal intensities of pixels selected from the center of triplicate spots were used for data analysis (data shown are mean ( standard deviation, and < 10% variance in intensity was observed across individual spots). RESULTS AND DISCUSSION Surface Modification and Oriented Oligosaccharide Attachment. The hydrazide-derivatized gold surface was prepared by formation first of a self-assembled monolayer of MHDA on goldcoated glass slides, followed by carbodiimide coupling of a homobifunctional spacer adipic dihydrazide to ω-carboxylic acids to yield aldehyde-reactive terminal hydrazide functional groups, as shown in Scheme 1. High-efficiency coupling was achieved in the last step of modification by allowing the reaction to occur in DMSO, which overcomes the hydrolysis problems of carbodiimide-formed esters.27 The hydrazide functional group displayed on the gold surface could efficiently and chemoselectively react with unmodified carbohydrate in aqueous solutions, forming covalent bonds through the reducing end aldehyde moiety. Hydrazide-modified beads have previously been used effectively for coupling heparin sugars onto solid supports.28,29 The sugarhydrazide linkage chemistry included a ring-opening process at (27) Hermanson, G. T., Bioconjugate Techniques; Academic Press: San Diego, 1996; p 145. (28) Nadkarni, V. D.; Pervin, A.; Linhardt, R. J. Anal. Biochem. 1994, 222, 5967. (29) Nadkarni, V. D.; Linhardt, R. J. BioTechniques 1997, 23, 382-385.

Scheme 1. Schematic Representation of the Surface Modification Process and Fabrication of an Oligosaccharide Microarraya

Figure 1. Effect of SAM fatty acid chain length on fluorescence signals. Fatty acids of different chain lengths (MUA/C11, MHDA/C16) were used to produce the SAM on gold-coated slides. The effect on fluorescence intensity of the resulting oligosaccharide microarray is shown. Fluorescence intensity was in arbitrary units. Inset shows the images of the antibody-stained microspots of the sugar spotted at different pickup solution concentrations. A decasaccharide (10 mer) from HNO2-digested heparin was used. The attached sugar was probed using an HS4C3 anti-heparin antibody and detected using a binding format shown in Scheme 1. Note the green color of the spots represents the ratio of 635 nm/532 nm, not the real fluorescence of the dye (the same for subsequent figures).

a

The approach is based on direct attachment of unmodified oligosaccharides onto hydrazide-derivatized, SAM-covered gold surfaces. The binding of the proteins to the microarray was detected using specific antibodies and visualized by fluorescencelabeled secondary antibodies.

the reducing end residue forming first a hydrazone, followed by conversion of the structure into predominantly the desired native β-pyranose form, leaving the oligosaccharide structure unchanged.30 Thus, the sugar linkage can be made without risk of decomposition of the saccharide structure, and the sugars can be displayed on the surface in a directed orientation analogous to cell surface presentation. However, successful implementation of this platform for the analysis of saccharide-protein interactions based on fluorescence-labeling detection requires the resolution of two fundamental problems: fluorescence quenching and nonspecific (background and noise) protein adsorption to the surface. Overcoming Fluorescence Quenching on Gold Surface. The first problem is the fluorescence-quenching effect on the gold chip surface. It is a commonly recognized phenomenon that a fluorescence-emitting molecule at a metal surface such as gold has reduced fluorescence efficiency because of quenching.31 Hence, we addressed this problem by passivating the gold surface with an ω-thiolated fatty acid SAM. The SAM was used here as an inert “spacer” to hold the fluorescent molecules at a certain distance from the metal interface, thus improving fluorescence (30) Flinn, N. S.; Quibell, M.; Monk, T. P.; Ramjee, M. K.; Urch, C. J. Bioconjugate Chem. 2005, 16, 722-728. (31) Kuhn, H. J. Chem. Phys. 1970, 53, 101-108.

efficiencies. In a previous study, we demonstrated that a long alkyl chain (thiolated fatty acid) was essential for sensitive fluorescence detection in this format.32 The evaluation of the effect of the alkyl chain length of the fatty acid on the fluorescent signal was done in an antibody-binding experiment using a degree of polymerization 10-mer heparin oligosaccharide probe (i.e., consisting of 5 disaccharide repeat units). Figure 1 shows a significant increase of the fluorescence signal intensity with increasing alkyl chain length of the fatty acid used in the SAM, for binding of an HS4C3 anti-heparin antibody to arrayed 10-mer heparin oligosaccharides. Thus, a longer spacer (C16) of MHDA in the monolayer provided improved quantum yield of the fluorescence emission compared with a shorter (C11) analogue, MUA. This observation was consistent with a recent report in which a C18 alkyl chain was used in the monolayer to achieve better detection in a fluorescencebased DNA assay.33 Note that we cannot preclude the possibility that a further increase of the length of the alkyl chain may improve fluorescence detection sensitivity; however, MHDA was the only commercially available product having a requisite alkyl chain length, and it was thus selected for use throughout this study. A further benefit derives from the fact that the carboxylic acid group displayed on the SAM is readily employed for derivatization with hydrazide functional groups via carbodiimide coupling to adipic dihydrazide, which is then amenable for reducing sugar conjugation. We also found that it is possible to synthesize an MHDAadipic dihydrazide adduct (mass 444) in the solution phase (in DMSO) using carbodiimide coupling chemistry and then present (32) Zhi, Z. L.; Morita, Y.; Hasan, Q.; Tamiya, E. Anal. Chem. 2003, 75, 41254131. (33) Chen, Y.; Shortreed, M. R.; Peelen, D.; Lu, M.; Smith, L. M. J. Am. Chem. Soc. 2004, 126, 3016-3017.

Analytical Chemistry, Vol. 78, No. 14, July 15, 2006

4789

Figure 2. Fluorescence quenching on hydrazide-modified SAM gold surface. Fluorescence images of Alexa Fluor 532 mono-NHS ester printed directly onto (a) a hydrazide-modified MHDA monolayer covered gold slide or (b) an aminosilane-derivatized glass slide (positive control). The dye was printed in formamide and incubated overnight.

it onto the gold surface forming a SAM, thus creating a singlestep procedure to make the hydrazide surface (see Supporting Information, Scheme 1). It was found initially that this method produced lower fluorescent signals for the sugar-antibody binding-based assay (data not shown). However, experiments on the optimization of the hydrazide density on the chip surfaces by adding an MHDA-ethanolamine adduct as a “dilutor”, showed that improved detection sensitivity could be obtained (see Supporting Information, Figure 1). Thus, a single-step procedure can be effectively used to simplify the production process for the hydrazide chips. It should be noted that the assay shown in Figure 1 had three protein binding layers (i.e., the target protein and two subsequent antibodies) added consecutively, with the last layer antibody being fluorescence-labeled. Additional assays were performed to compare whether the number of layers, in addition to a SAM, was also important for high-sensitivity detection. As expected, experiments using only one layer of a fluorescence dye attached covalently to the hydrazide-modified SAM-covered gold surface showed virtually no fluorescence signalseven at very high concentrationssas compared to the amino-glass slide (Figure 2). In contrast, both two- and three-layer protein detection approaches allowed detection of good signal levels, with the latter being the most effective (data not shown). It is thus very probable that both an efficient insulating SAM layer and a multilayer protein structure (two or preferably three layers) are essential to provide efficient fluorescence detection on the SAM-modified gold surfaces. The use of antibodies for the detection is not expected to be a significant obstacle for the majority of applications and indeed is likely to be the most straightforward and common platform used for identifying binding partners. In addition, the use of an antibodybased assay can avoid the problems associated with direct labeling of target proteins that may compromise binding or create artifactual protein characteristics. Minimizing Nonspecific Adsorption of Proteins to Gold Chip Surface. As in protein microarray technology, background signal via nonspecific adsorption of interfering proteins on the 4790

Analytical Chemistry, Vol. 78, No. 14, July 15, 2006

SAM-modified gold surface is a potential major problem, and this was encountered in our study. We compared a number of blocking schemes for effectiveness in blocking potential sites for nonspecific binding of the detection proteins to the surface in an antibodybased assay. In particular, BSA, chemical modification of the hydrazide-modified surface using PEG-aldehyde, and a combination of both approaches were assessed (data not shown). It was observed that BSA blocking of the gold surface showed a strong nonspecific signal. In contrast, when PEG aldehyde was applied as the blocking agent, the nonspecific signal was significantly improved, but not eliminated completely. It is thus unlikely that a single blocking approach will work for every possible protein target although a previous study found that PEG blocking could eliminate efficiently nonspecific adsorption for a SPR-based binding detection.34 A combination blocking scheme involving the use of PEG aldehyde and supplementing BSA in binding protein solutions was found to be the best for minimizing the nonspecific background and noise signals on gold surfaces. Note that other commercially available blocking reagents, such as casein and SuperBlock (Pierce), were found to strongly inhibit the binding of some heparin-binding proteins to the immobilized sugars and are thus not suitable as blocking agents (data not shown). Another blocking agent, TopBlock (Sigma) was found to reduce background but was not superior to the simple use of BSA (data not shown). Optimization of Hydrazide Density on SAM Surfaces. A previous report observed that an excessively high terminal biotin density on the SAM surface may have a negative influence on the binding of streptavidin.35 This was explained on the basis that biotin tails need a necessary level of freedom to stack more efficiently with streptavidin. In that work, however, biotin was linked to the alkanethiol prior to the formation of SAM. Thus, a full coverage (100%) of biotin was expected when no “dilutor” was added to the self-assembling loading solution. Similarly, in a separate study, we have observed better binding of Con A lectin to an immobilized mannose monosaccharide on a SAM using a lower density of hydrazide-modified surface (Z.l.Z. and J.E.T., unpublished data). In the present work, we examined whether the surface density of hydrazide groups exerted an influence on the attachment of the oligosaccharides of heparin and subsequently on protein-binding events. We prepared surfaces modified with different coverage ratios of hydrazide groups by introducing a “dilutor”, ethanolamine, during the coupling of adipic dihydrazide. It was found that undiluted hydrazide in fact gave the best specific signals and signal-to-background ratio, as shown in Figure 3. This may be because of the fact that the present study carried out the hydrazide modification on the preordered MHDASAM, and thus, the hydrazide groups were attached to the terminal carboxylic acid groups in a maximal but reduced ratio due to space limitations. An additional factor is that the 10-mer oligosaccharide probes used in this study are likely long enough to overcome space limitations for binding to a target protein. Additionally, since carbohydrate-protein interactions are generally weak and, in some cases, require multivalent interactions for stable binding, a higher density of sugar on the chip surface can (34) Mun ˜oz, E. M.; Yu, H.; Hallock, J.; Edens, R. E.; Linhardt, R. J. Anal. Biochem. 2005, 343, 176-178. (35) Riepl, M.; Enander, K.; Liedberg, B.; Schaferling, M.; Kruschina, M.; Ortigao, F. Langmuir 2002, 18, 7016-7023.

Figure 3. Effect of hydrazide density at SAM surface on the signalto-noise ratio of the oligosaccharidemicroarray signals. The hydrazide density was adjusted by varying the ratio of ethanolamine added to the adipic dihydrazide. Note the difference in the measure of the left and right axes. Detection antibody and the 10-mer heparin oligosaccharide (HNO2 digested, 1 mg/mL pickup spotting solution concentration) were the same as those used in Figure 1.

undoubtedly increase the binding affinity and stability of binding to proteins in some cases. Covalent Coupling of Oligosaccharides to the Surface. One of the major advantages of the present hydrazide-based approach for sugar attachment is that the sugar is covalently attached to the surface via its reducing end. Thus, the sugar moiety is anchored in a directed orientation in such a way that the binding proteins can properly access the molecules for cognate interactions. However, for negatively charged probes such as sulfated oligosaccharides, it is also likely that the probes could be adsorbed nonspecifically onto the positively charged hydrazide surface. To distinguish whether the sugars were attached on the surface by covalent bonds or by nonspecific physical adsorption, we compared the immobilization of two sugar probes with the same length and charge. The only difference between the two probes is that one of the probes was pretreated by reduction with NaBH4, so that its reducing end was blocked and could no longer couple covalently to the hydrazide group. The difference in signal between the untreated and treated probes can be attributed to the contribution of the covalent attachment in the case of the untreated saccharide. Figure 4 shows the signal intensities of these two probes prepared by either partial HNO2 or enzyme digestion of heparin. A significant reduction of the fluorescent signal intensities of the reduced sugars compared with that of nonreduced ones revealed that the sugar probes were attached predominantly by the covalent bond for the nonreduced sugars. In fact, the signal intensity of the nonspecific adsorption part accounts for only ∼1% compared to that of the covalently attached saccharide (note the logarithmic relationship between amount of sugar spotted and the fluorescence signal). This indicates that the majority of the sugar probes were covalently attached rather than nonspecifically adsorbed. Washing of the oligosaccharide microarray chip with 2 M NaCl for 20 min did not result in any significant change in signal compared to washing with water, indicating that the attachment of saccharides to the microarray (either covalently or noncovalently) was stable and resistant to ionic competition (see Supporting Information Figure 2). It is quite

Figure 4. Contribution of covalent attachment to fluorescence signals on the oligosaccharide microarray. Microarrays were prepared of a dilution series of decasaccharides derived from HNO2 and heparitinase III enzymatic digestions of PMH, to compare differences in signal between reduced (reducing end blocked) and untreated oligosaccharides. Rows of 3 spots each of dilution were printed. (a) Decasaccharide (10 mer) prepared by partial heparitinase III enzymatic digestion. (b) 10-Mer saccharide prepared by partial HNO2 digestion. Concentrations of the sugars spotted are shown between panels a and b; R denotes “reduced sugar”, i.e., the reducing end blocked. (c) Shows a comparison of the quantified spot signals obtained from normal and reduced decasaccharide probes. The inset shows the possible orientations of the attached probe of a normal sugar (1) and a probe with blocked reducing end (2). Detection was achieved using the anti-heparin antibody HS4C3, followed by a primary antibody and an Alexa Fluor 532-labeled secondary antibody. Fluorescence intensity is in arbitrary units.

likely, however, that the coupled sugars are adopting a number of conformations on the array surface, and some may well be interacting weakly with the hydrazide surface. However, the strong binding signals achieved for target proteins strongly suggest that, upon protein binding, these saccharides would be sequestered away from the surface in order to intact fully with the protein partners. In addition, we also found that the probe derived from HNO2-digested heparin had generally higher detection sensitivity compared with the enzyme-digested probe. This is likely due to the intrinsic difference in reducing end structures of the sugars after digestion. HNO2 digestion results in generation of an anhydromannose residue with free aldehyde groups at the reducing ends, whereas the bacterial lyase enzyme digestion Analytical Chemistry, Vol. 78, No. 14, July 15, 2006

4791

maintains the natural reducing end in a ring-closed form, which is less reactive to the hydrazide groups.36 Profiling the Interactions of Heparin-Binding Proteins and Lectins with Oligosaccharide Microarrays. Having optimized the conditions for oligosaccharide attachment onto SAMcovered gold surfaces and interrogation with an anti-heparin antibody target, we further extended this approach to profile the binding of the heparin decasaccharide microarray to two types of known heparin-binding proteins, i.e., the growth factor GDNF and the growth factor receptor FGFR2. Different dilutions of the decasaccharide fractions derived from HNO2- and enzyme-digested PMH were arrayed and the binding of target proteins was detected using corresponding primary antibodies and Alexa Fluor 532labeled secondary antibodies. The binding profiles, as shown in Figure 5, showed that all these interactions have been detected successfully, indicating the retention of specific binding activity of the surface-bound oligosaccharides. Good signals were obtained using pickup solution concentrations of sugar in the range 10100 µg/mL; since ∼1 nL of this solution is spotted by the arraying robot, this equates to ∼3-30 fmol of spotted heparin 10-mer and indicates the very high level of detection sensitivity achieved in this microarray format. The concentrations of spotting solutions (10-100 µg/mL or 3-33 µM) required for this study are much lower than that of many other previously described methods e.g., Lee and Shin (∼30 mM),18 Houseman et al. (∼2 mM),7 and Xie et al. (∼0.1-10 mM)10 and are comparable to those described by Zhou and Zhou (∼20-500 µM)19 and Blixt et al. (∼10-100 µM);9 note also that, in all except the Lee and Shin18 and Zhou and Zhou19 methods, prederivatized saccharides are required, so the true amount of starting material required would likely be much higher. As previously, HNO2-digested sugars (Figure 5a) showed better sensitivity than enzyme-digested ones (Figure 5b), similar to that observed in antibody-based screening of these microarrays (Figure 4). Additionally, a non-heparin-binding protein G. nivalis lectin used as a negative control showed no fluorescence signal, reflecting the expected absence of its binding to the microarrayed heparin saccharides. To demonstrate the general applicability of the present microarray platform, we also tested the attachment of mannopentaose and mannobiose, chemically synthesized neutral oligosaccharides with different structural characteristics, to the hydrazidemodified surface and its binding to target lectins (Con A and snowdrop G. nivalis). The bound lectins were detected by an antiCon A primary antibody or an anti-G. nivalis primary antibody, and an Alexa Fluor 532-labeled secondary antibody. The results shown in Figure 5 indicate that the attached Man5 and Man2 bound strongly to Con A and, as expected, more weakly to G. nivalis (probably due to a lower binding affinity for the latter lectin). Note that higher concentrations of sugar in the pickup spotting solution (typically ∼1000 µg/mL) were required for these types of sugars. This probably reflects the lower instrinic affinities of lectin-glycan interactions (typically micromolar compared to nanomolar for heparin-protein interactions) and presumably also the need for multivalent ligand presentation (and thus presumably higher surface attachment densities to achieve optimal saccharide presentation). In a control experiment using a heparin-binding (36) Powell, A. K.; Yates, E. A.; Fernig, D. G.; Turnbull, J. E. Glycobiology 2004, 14, 17R-30R.

4792 Analytical Chemistry, Vol. 78, No. 14, July 15, 2006

Figure 5. Binding of carbohydrate-binding proteins to immobilized sugars in microarray format. Interaction of heparin-binding proteins (GDNF and FGFR, 30 µg/mL) and lectins (Con A and G. nivalis) to heparin decasaccharides and mannose microarrays fabricated on gold. Detection was achieved using primary antibodies specific for each protein, followed by an Alexa Fluor 532-labeled secondary antibody (1 µg/mL). Rows of duplicate spots for each dilution are shown. The concentrations of sugar probes in the pickup solutions are shown at the left side of the panels. Panel a shows the microarray of heparin oligosaccharide (10 mer) produced by partial HNO2 digestion. Panel b shows the microarray of heparin oligosaccharide produced by partial heparitinase III enzymatic digestion. G. nivalis lectin was used as a negative control. Panel c shows the binding of lectins (5 µg/mL Con A and 100 µg/mL G. nivalis) to Man5 and Man2 microarrays. A nonlectin (GDNF) was used as the negative control. Note the smaller size of the spots of Man5 and Man2 compared to that of negatively charged heparin saccharides; this is likely due to the lower surface tensions of the spotted solutions containing the noncharged Man5 or Man2.

protein (GDNF), no fluorescence was detected from any of the Man5 spots, indicating the expected lack of binding to this saccharide. These lectin-binding experiments therefore demonstrated that the surface-immobilized Man5 retains its ability to

specifically recognize and bind to cognate lectins. Overall these data demonstrate that the present microarray platform is suited for arraying oligosaccharides with different chemical and structural characteristics and thus provides potential for high-throughput analysis of specific interactions of carbohydrates with proteins. CONCLUSIONS We have demonstrated a straightforward approach for the covalent attachment of oligosaccharide probes onto a hydrazidemodified, alkyl SAM-covered gold surface, achieving oriented presentation of oligosaccharides on the surface. The fluorescence detection of the microarray was improved by introducing an efficient insulating alkyl SAM layer and use of an additional three layers of detection proteins to minimize quenching effects. Nonspecific adsorption of the target and detection proteins to the gold surface was minimized by implementing a blocking scheme involving a combination of PEG aldehyde surface blocking and BSA in the protein detection solutions. This microarray platform has been applied to sulfated oligosaccharides derived from a natural source, heparin, and also the synthetic oligosaccharides, mannopentaose and mannobiose. In principle, oligosaccharides of any length can be used, as long as they have a free reducing end available for coupling. The expected specific protein-binding specificities for immobilized carbohydrates were retained and binding events to the corresponding specific binding proteins detected with high sensitivity, with as little as a few femtomoles of sugar spotted in the case of the heparin saccharides. The initial approach involved a two-step chip substrate modification prior to robotic spotting of the sugars, but we have also demonstrated that this can be reduced to a single-step procedure using a prederivatized MHDA-adipic dihydrazide adduct and optimized for hydrazide density by diluting with an MHDA and ethanolamine adduct. Most importantly, the approach allows direct saccharide attachment without laborious derivatization or cleanup steps, thus greatly simplifying the carbohydrate microarray fabrication procedure and overcoming the difficulties involved in handling heparin-type complex sugars under stringent prederivatization and cleanup conditions. It should be readily possible to include highly

diverse oligosaccharide structures in the microarray using sugar probes derived either synthetically or purified from tissues. These fluorescent microarrays provide a powerful tool for exploring specificity in carbohydrate-protein interactions, in a highthroughput format suited to “glycomics” approaches to deciphering the informational code of oligosaccharide structures. Direct immobilization of sugars onto arrays will undoubtedly have major advantages over methods requiring prior derivatization as it will allow facile large-scale display of diverse sugar structures. Moreover, this novel glycochip approach provides a great degree of versatility since the gold surfaces provide a common platform for equivalent immobilization but interrogation readouts using detection techniques as diverse as fluorescence, matrix-assisted laser desorption/ionization time-of-flight mass spectroscopy,21 surface plasmon resonance,22 and quartz crystal microbalances.23 ACKNOWLEDGMENT This research was supported by the UK “Glycochips” Consortium funded by the Research Councils UK Basic Technology Program (to J.E.T.), a Biotechnology and Biological Sciences Research Council project grant (to A.K.P. and J.E.T.), a grant to the Liverpool Centre for Bioarray Innovation (Prof. Andy Cossins), and a Medical Research Council Senior Research Fellowship (to J.E.T.). The authors acknowledge the assistance of Dr. Abdel Atrih for help with ES-MS analysis, kind provision of anti-heparin antibodies by Prof. Toin van Kuppevelt and Dr. Gerdy ten Dam (Nijmegen University), and helpful advice from Dr. Ed Yates, Prof. Dave Fernig, and Tim Rudd (Liverpool University), from Prof. Rob Field (University of East Anglia), and from Prof. Jonathan Blackburn (Capetown University). SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review January 12, 2006. Accepted May 10, 2006. AC060084F

Analytical Chemistry, Vol. 78, No. 14, July 15, 2006

4793