Preparation of Aluminum Oxide-Coated Glass Slides for Glycan Microarrays.

In this study, we report the fabrication of aluminum oxide-coated glass (ACG) slides for the preparation of glycan microarrays. Pure aluminum (Al, 300 nm) was coated on glass slides via electron-beam vapor deposition polymerization (VDP), followed by anodization to form a thin layer (50-65 nm) of aluminum oxide (Al-oxide) on the surface. The ACG slides prepared this way provide a smooth surface for arraying sugars covalently via phosphonate formation with controlled density and spatial distance. To evaluate this array system, a mannose derivative of α-5-pentylphosphonic acid was used as a model for the optimization of covalent arraying based on the fluorescence response of the surface mannose interacting with concanavalin A (ConA) tagged with the fluorescence probe A488. The ACG slide was characterized using scanning electron microscopy, atomic force microscopy (AFM), and ellipsometry, and the sugar loading capacity, uniformity, and structural conformation were also characterized using AFM, a GenePix scanner, and a confocal microscope. This study has demonstrated that the glycan array prepared from the ACG slide is more homogeneous with better spatial control compared with the commonly used glycan array prepared from the N-hydroxysuccinimide-activated glass slide.

Introduction

Glycan microarrays have been used as an effective tool for the high-throughput analysis of protein–glycan interactions and are thus useful for disease diagnosis, drug discovery, and vaccine development.[1−28] Numerous surfaces have been made available for glycan arraying, including (a) noncovalent adsorption of sugar derivatives to the surface of a porous nitrocellulose membrane;[1−4,7,9,10,17,22,23,26] metal oxide surface;[5] microtiter plate; fabricated plastics of polystyrene, polypropylene, or polycarbonate; and polyfluorohydrocarbon-linked aluminum oxide-coated glass (ACG) slides[21] and (b) covalent attachment to a gold surface or alkenthiol-activated gold surface,[6,9−11,22] epoxy-activated glass slide,[6,18,19,22,25]N-hydroxysuccinimide (NHS)-activated glass slide,[8−10,14,20,22,24,25] and ACG slide.[15,21] The glycan array on ACG slides developed in our laboratory[15,21] was prepared by spotting glycan–phosphonic acids onto the surface of the ACG slides,[29−31] and the properties of the arrays were characterized using both mass spectrometry and fluorescence scanning microscopy.[15,21] The fluorescence intensity of sugar–protein interaction on the ACG slide was found to be more sensitive and homogeneous with higher glycan density than that on the NHS-activated glass slides.[14,15] We have further used the ACG slides to prepare a mixed-glycan array for the study of broadly neutralizing monoclonal antibodies isolated from HIV patients and found that some of the antibodies recognized two different glycans simultaneously; a new observation that was not detected previously by the use of NHS-activated glass slides.[28] In addition, the ACG slide surface can be used for noncovalent hydrophobic adherence of glycans with a fluorohydrocarbon tail for the identification and study of enzyme activity analyzed using MALDI mass spectrometry.[16,21]

Despite all of these studies, the fabrication method for the preparation of ACG slides has not been clearly defined. In our previous study, the anodized aluminum oxide (AAO) surface had properties and functions similar to those of the native aluminum oxide (NAO) surface.[16] Different vapor deposition polymerization (VDP) techniques have resulted in various degrees of surface roughness. Even the surface morphology of the ACG slides appeared different from that of aluminum objects that have been widely used in industry.[29,32−38] Previous surface treatment of aluminum objects involved mostly in-depth anodization for the control of pore sizes but not for the preparation of glycan microarrays. On the basis of our experience, the AAO surface to be used in glycan microarrays should have the following properties: (1) The aluminum oxide (Al-oxide) layer should be smooth enough for the covalent coupling reaction with sugar derivatives of phosphonic acid.[29,30] (2) The thickness of the Al-oxide layer on the surface should be adjusted to provide optimal fluorescence intensity for the detection of the protein–sugar interaction. In this study, we use the glass slide with electron-beam (E-beam)-coated aluminum to produce ACG slides with various thicknesses under various anodization conditions and then study their effect on the fluorescence intensity in sugar–protein interactions. Mannose with an α-5-pentylphosphonic acid tail was used as a model compound for covalent attachment to the Al-oxide layer, and the fluorescence intensity of Alexa Fluor-488-tagged ConA (ConA-A488) that interacts with the covalently bound sugar was used to evaluate and optimize the system. Computer software was used to design experiments to fabricate the surface with various thicknesses and roughness of the AAO layer for the study. Data analysis of the result was conducted mathematically and used for the development of optimized conditions for the preparation of ACG slides.

To characterize the ACG slide to be used in a covalent glycan microarray, sugar loading capacity, uniformity, and spatial distribution of sugar derivatives on the surface were compared with those on the commercially available NHS-activated glass slides using atomic force microscopy (AFM), a GenePix scanner, and confocal microscopes.

Results and Discussion

Glycan microarrays are mostly used to study glycan–protein interactions with the goal of understanding the nature of the multivalent interactions between the cell-surface glycans and the proteins they interact with. Therefore, an ideal glycan array is one that best mimics the glycans that are found on living cell surfaces. For example, the density and spatial distribution of Globo-H on malignant cells and their stem cells of breast cancer patients progressively increase. To identify the patients who would benefit from vaccination with the Globo-H vaccine,[39] to study the disease progression, and to monitor the antibody response to the vaccine in vitro, it would be useful if the glycan arrays were able to mimic the changes on the cell surface at different stages of the disease. However, there is no such glycan array available to date as the cell surface is a dynamic and heterogeneous system with variations in substrates on the surface, and as such the development of a glycan array to mimic the real cell surface remains a major challenge. Nevertheless, the preparation of a surface for glycan array in a controllable and reproducible manner with regard to the distribution, density, and spatial orientation of glycans is important to address some fundamental questions such as the avidity of weak binding and multivalent heteroligand interaction. The glycan array on ACG slides has been shown to meet these needs.

Surface anodization of aluminum has been used in numerous industrial applications mostly in the electrical oxidation of objects for pore size creation, color-filling, pore-sealing and pore-polishing. The surface anodization requires the use of strong acid such as sulfuric acid and long reaction time from hours to days. There are many reports describing the preparation of anodized Al-oxide surface with pore structures varying from nanometers to micrometers,[29,32−38,40] but little is known about the preparation of Al-oxide with smooth uniform thickness,[36,41] as the ACG slide used for a glycan microarray requires a smooth surface with low surface roughness.

Before a thorough study of surface anodization, we screened several VDP techniques to produce Al-oxide on metal-coated glass slides. These VDP screening studies included (1) Al-oxide of 50 nm thickness produced using atomic layer deposition onto the Al-coated (with e-beam evaporation) glass slide; (2) Al-oxide of 2 nm thickness obtained from direct oxygen plasma treatment on a 300 nm thick aluminum (with e-beam evaporation) coated on glass slides; and (3) Al-oxide of 36 nm thickness produced using magnetron sputtering on the Al-coated (with magnetron sputtering), the nickel-coated (with e-beam evaporation), or the chromium-coated (with e-beam evaporation) glass slides. All of these substrates were examined using the GenePix fluorescent scanner. The original purpose of screening diversified methods was to create different Al-oxide microstructures on the surface to identify an optimal way to enhance the signal. However, the best fluorescence intensity produced from these slides was the AAO grown on glass slides. Without further investigation of the VDP techniques, we then focused on the use of E-beam evaporation with argon plasma to coat a layer of pure aluminum on the glass slide followed by acid anodization of the aluminum surface. Our pervious study[15] showed that the slide surface roughness with rms (root mean square) < 18 nm can be used for microarray printing. Schott’s glass slides provided a smooth surface (rms 0.33–0.42 nm scanned using AFM over 10 × 10 μm2 area) for aluminum coating. The Al-coated glass slides produced in this way have a smooth surface with a rms of 0.75 nm (for 100 nm) and 2.33 nm (for 300 nm) of coated aluminum. An optimized AAO surface gave a rms of 2.31 nm, with surface roughness similar to that of the Al-coated glass slide. The same results were obtained when glass slides with higher rms were used for Al-oxide preparation. We believe that during the electrochemical surface anodization, the electropolishing process[42−44] occurred simultaneously to remove the convex, sharp, and rough materials from the surface. Therefore, a thin layer of AAO with surface roughness similar to glass can be prepared under the optimized surface anodization reaction condition.

Figure shows a schematic drawing of the fabrication of an ACG slide to be used in a glycan microarray. The clean 1 mm thick glass slide was coated with pure aluminum (300 nm) in an argon plasma-assisted E-beam VDP coating chamber. Surface anodization was conducted via wet electrochemical reaction. The final optimized ACG slide contains a layer of anodized Al-oxide (approx. 50–65 nm) as measured using ellipsometry.

Figure 1

Schematic drawing of ACG slide preparation.

A top view of the scanning electron microscope (SEM) image (150 000×) of an AAO surface is given in Figure A in which the granular aluminum crystals underneath could be seen and no pores were created. Unlike those surfaces with in-depth anodized alumina reported in the literature, the surface morphology shown in Figure A looks similar to a surface with uniform distribution of a thin layer of transparent Al-oxide. Figure B shows the cross section SEM image (80 000×) of the sandwich-like ACG slide. Its surface roughness, as shown in Figure C, scanned over 10 × 10 μm2 using AFM was similar to the ordinary glass (<3 nm). In glycan microarray preparation, dispensing one droplet (0.6 nL per spot) of sugar solution wetted out an area of approximately 150 μm in diameter, which covers approximately 750 times the surface area, as shown in Figure A.

Figure 2

(A) Top view (150 000×) and (B) cross section SEM images (80 000×) of a typical sandwich-like ACG slide. (C) Roughness analysis with image statistics (rms: 2.3 nm) of the optimized AAO surface over 10 × 10 μm2 area analyzed using AFM.

To produce a smooth AAO surface, the anodization was conducted via electrochemical reaction, by placing the Al-coated glass slide in an oxalic acid aqueous solution in a 4 °C incubator at controlled voltage and reaction time. A 10 L oxalic acid solution (0.3 M) was prepared in stock to complete the entire study. As shown in Figure , the beaker (600 mL) was filled up with the electrolyte, and the acid solution was collected separately and reused.

We found that it is difficult to make a nontransparent Al-oxide layer with an Al-coating thickness of just 100 nm. Visually, we can see that the substrate fabricated from the 100 nm Al-coated glass slide was semitransparent. Therefore, Al-coated glass slides with 300 nm of aluminum have been used in all surface anodization experiments. As can be seen in the cross section image in Figure B, part of the aluminum has been converted into Al-oxide. The final Al thickness was greater than 200 nm, and the substrate was nontransparent.

Figure shows a molecular model of mannose/ConA-A488 binding. Mannose with an α-5-pentylphosphonic acid tail was covalently bound to the surface of the ACG slide. The ConA-A488 (c11252 from Invitrogen) was bound specifically to the immobilized mannose. Con A is a lectin tetramer with a subunit dimension of 42 × 40 × 39 Å3, and each subunit has a mannose binding site.[45,46] Geometrically, only two binding sites per tetramer are available for mannose binding. The most effective mannose/ConA-A488 binding produces the strongest fluorescence intensity, and we have been using the fluorescence intensity of mannose/ConA-A488 binding to tailor-make the Al-oxide surface for the glycan microarray.[15] Designed experiments were used to optimize the preparation of ACG slides for glycan microarray.

Figure 3

Molecular model of mannose/ConA-A488 binding. Mannose with α-5-pentylphosphonic acid was covalently bound to the ACG slide surface. Fluorescence-tagged concanavalin A (ConA-A488) was then bound to mannose specifically. Con A is a lectin tetramer with subunit dimension of 42 × 40 × 39 Å3. Each subunit has a mannose binding site. Geometrically, only two binding sites per molecule are available for mannose binding. The most effective mannose/ConA-A488 binding should give the strongest fluorescence intensity, and this model system has been used to optimize the AAO surface for the glycan microarray.

An optimized ACG slide should possess a smooth glasslike AAO layer, thus mild conditions for making the AAO surface have been selected for investigation. By keeping the reaction temperature consistently low (at 4 °C) and constant acid solution concentration (at 0.3 M), the voltage and the reaction time were the only two reaction variables. To screen the reaction parameters, two sets of experiments were conducted using the computer software program Design Expert 8.0 (Supporting Information).

With the information obtained from the first two sets of designed experiments (as shown in the Supporting Information), we are now able to define the ranges of voltage (9.8–45.2 volts) and reaction time (71.4–198.6 s) for the RSM (response surface measurement) experiment. The slides fabricated under these conditions were used to array mannose solution to form a 10 × 12 array matrix with sugar concentration ranging from 100 mM to 1 pM. These experiments have essentially established the response in fluorescence intensity resulting from sugar/protein binding. For comparison, an NHS-activated glass slide was arrayed simultaneously using the in-house synthesized mannose derivative of α-pentoxylamine[21] and was used to interact with the same ConA-A488 (c11252 Invitrogen) protein solution (3 μg/mL) for the binding intensity study.

The results of response measurements are shown in Table . A total of 13 Al-coated glass slides were used in this set of experiments. The second and third columns are variable ranges of voltage and reaction time. The fourth to seventh columns are the measured data of AAO thickness, electrical current, fluorescence intensity of arrayed mannose (100 μM) solution with ConA-A488 binding, and maximum binding intensity (Bmax) for each glass slide prepared. The Bmax in column 7 of Table was estimated using the fluorescence intensity obtained from 1 μM to 100 mM with GraphPad Prism 5.0. All mathematical models for each response measurement are given in Table S5.

Table 1

Optimization Experiment—Factors, Voltage (volt), Reaction Time (s) and Responses of AAO Layer Thickness (nm), Electrical Current (mA), Fluorescence Intensity of 100 μM Mannose Solution Arrayed on Each Slide Surface, and Bmax Derived from Michaelis–Menten Equation Using GraphPad Prism7.0

run no.	factor 1 voltage (volt)	factor 2 Rx. time (s)	response 1 AAO thickness (nm)	response 2 current (mA)	response 3 fluorescence intensity (of 100 μM sugar conc.)	response 4 Bmax (fluorescence intensity)
1	27.5	198.64	83.2	11.5	3.87 × 106	3.93 × 107
2	40	90	87.9	20	5.21 × 105	7.36 × 106
3	27.5	135	69.3	12.9	6.27 × 105	9.65 × 106
4	27.5	71.36	49.9	9.4	3.19 × 106	4.76 × 107
5	27.5	135	69.9	11.8	6.47 × 106	4.86 × 107
6	15	180	36.2	5.6	2.74 × 105	4.11 × 106
7	45.18	135	140.5	31.8	1.42 × 106	2.12 × 107
8	27.5	135	72.7	13.3	4.58 × 106	4.89 × 107
9	27.5	135	70.2	12.6	4.59 × 106	4.36 × 107
10	9.82	135	15.5	2.1	3.55 × 105	5.74 × 106
11	15	90	23.1	4	1.28 × 105	1.80 × 106
12	40	180	124.9	22.4	1.10 × 105	2.83 × 106
13	27.5	135	62.5	10.8	4.44 × 106	4.85 × 107

Figure A shows the changes in electrical current and voltage versus reaction for ACG slide no. 5 (Exp. number 5). The results of other slides are given in the Figure S3A. Figure B shows the images of a 10 × 12 matrix with 10 repeated spots, with each column of the mannose solution concentration varied from 100 mM to 1 pM across the row (also shown as ACG slide no. 5 in Figure S3B).

Figure 4

(A) Changes in electrical current for the fabrication of ACG slide no. 5. (B) Image of fluorescence intensities that resulted from the mannose/ConA-A488 binding of ACG slide no. 5. This image shows a 10 × 12 matrix with 10 repeated arrays, with each column of the mannose solution concentration varied (in consecutively 10 times dilution) from 100 mM to 1 pM. (C) Response surface of the modified quadratic model for AAO thickness transformed into a function of voltage and reaction time, YAAO thickness = a × (V)1/2 + b × (RT)1/2; intercept ≠ 0 (P < 0.0001). (D) Response surface fluorescence intensity with respect to 100 μM sugar concentration arrayed on the ACG slide surfaces. (E) Response surface of Bmax (1 μM to 100 mM) derived from model fitting, Y = a × V + b × V2 (with significant P value) given in Table S5. Both (D) and (E) show an optimal curvature of high fluorescence intensity within the range of this study.

Figure C shows the response surface with respect to voltage and reaction. The AAO thickness is transformed into a modified quadratic function of voltage and reaction time with YAAO thickness = a × (V)1/2 + b × (RT)1/2; intercept ≠ 0. The modified quadratic mathematical functions of “bioactivity” (i.e., the fluorescence intensity and Bmax) with respect to the fabrication factors (voltage and reaction time) are shown in Figure D,E.

Both Figure D (derived from experimental data) and 4E (derived from theoretical calculation) have shown similar curvatures, indicating that an optimal curvature of high fluorescence intensity within the range of this study, there indeed existed reaction condition(s) for making ACG slides, which produced the highest fluorescence intensity. Additional data of the ACG slide arrayed with various sugar concentrations are given in Figure S4.

Figure D shows an example of the fluorescence intensity with 100 μM sugar concentration arrayed on the ACG slide surfaces. The response surfaces of all sugar solution concentration (as shown in Figure S4) were derived from the modified quadratic models (given in Table S5), with math equations either Yintensity = a × V + b × V2; intercept = 0, or Yintensity = a × (V)1/2 + b × (V2)1/2; intercept = 0, suggesting that voltage is a critical variable in making “good” ACG slides. From the fluorescence intensity data, the voltages and reaction times used for making the optimized ACG slides are thus defined.

Figure E shows the response surface of Bmax with respect to reaction conditions. To identify the optimal ACG slide surface for the highest fluorescence intensity, Bmax was derived from the Michaelis–Menten’s equation,[47]Y = (Bmax × X)/(Kd + X), where Y is the specific ligand/protein binding [expressed as the fluorescence intensity of mannose solution concentration (varying from 1 μM to 100 mM) arrayed on each ACG slide surface used in this system] and x is the concentration of the specific ligand (mannose) that binds to ConA-A488. Bmax and Kd for each slide were obtained from GraphPad Prism. Model fitting of Bmax (using Design Expert) turned out to be Y = a × V + b × V2 (as given in Table S5). Both Figure D (derived from experimental data) and 4E (derived from theoretical calculation) have shown similar curvatures, indicating that an optimal curvature of high fluorescence intensity within the range of this study, there indeed existed reaction condition(s) for making the ACG slides, which gave the highest fluorescence intensity of the same sugar/protein binding system.

As shown in Figure S3C (using the model fitting shown in Table S5), the electrical current is also derived as a function of voltage, Ycurrent = a × V + b × V2; intercept ≠ 0. This derived mathematical equation has been deviated from the theoretical prediction of linearity. As the ACG slide made at high voltages, its electrical current did not reach equilibrium and continually drifted upward; the model deviated from the theoretical prediction (Ycurrent = a × V) is an indication of the heat-sink issue before large-scale anodization.

The predicted AAO thickness for optimal intensity is given in Table . As indicated, the thickness for optimal fluorescence turned out to be 55 nm (±11 nm), and the predicted value using Bmax was 52.4 nm with the standard deviation of ±0.3 nm. Nonetheless, the curvatures of the optimal regions in Figure D,E have been quite flat, suggesting that the best AAO surfaces can be made within the ranges of the reaction conditions (voltages and reaction times).

Table 2

Optimized Reaction Condition for Making the ACG Slide from an Al-Coated Glass Slide Based on the Fluorescence Intensity Resulting from Mannose/ConA-A488 Binding and Bmax Analysis

optimized condition	voltage	STD	Rx. time	STD	AAO thickness	STD	current	STD
A488-tagged Con A	25.8	0.7	135.3	21	55	11	10.3	0.5
Bmax/A488-tagged Con A	25.8	0.02	137	19	52.4	0.3	10.3	1.7

As shown in Table , an optimized ACG slide contains a layer of AAO with 50+ nm on the surface. It also showed that the optimized reaction conditions for making ACG slides are voltage between 25.8 ± 0.7 volts and reaction time between 135 ± 21 s. The reaction temperature was set constantly at 4 °C using freshly prepared and up to the third repeated use of 0.3 M oxalic acid aqueous solution. Figure shows the thickness variations of the AAO layer versus the electrical current under the suggested optimized reaction conditions (25.8 volts, 121 s). With the glass slides obtained from different suppliers (Schott and Arrayit), 30 slides were fabricated (one slide at a time repeatedly) under this optimized reaction condition using either freshly prepared or up to the sixth repeated use of the oxalic acid solution. The surface roughness of the starting glass base material (with rms 0.33–0.42 nm and 0.47–3.89 nm for the glasses obtained from Schott and Arrayit, respectively) affected the final rms (2.31 vs 4.56 nm) but not the thickness of the AAO layer. A few outliers in Figure were obtained, where the ACG slides were fabricated on 2 consecutive hot days, where the reaction temperature was probably not well maintained at 4 °C. The in-line voltage control (by generator) was recorded, and thus the electrical current was measured during the fabrication of ACG slide. The thickness of the AAO layer was analyzed using ellipsometry. As can be seen in Figure , the electrical current is linearly proportional to the thickness of the AAO layer of the substrate. Alternatively, instead of using ellipsometry, the AAO thickness can be estimated from the in-line electrical current measurement during the ACG slide fabrication.

Figure 5

Thickness of the AAO layer vs electrical current measured during AAO fabrication. Repeated surface anodization experiments (30 times) using one of the optimized reaction conditions (25.8 volts and 121 s) with the acid solutions either freshly prepared or reused up to the sixth repeated use of the oxalic acid solution. The thickness of the AAO layer was measured using ellipsometry. Spots within the red square are the slides with optimized AAO thickness.

In summary, the thickness of the AAO layer grown from this system ranged from 23.1 to 185.7 nm (at 15 volts/90 s and 50 volts/120 s, respectively, as shown in the Supporting Information). There were no pores created during surface anodization for the formation of Al-oxide. Because the thickness of the coated aluminum layer was only 300 nm, in 15 min of reaction time, all Al-oxide/aluminum layers were dissolved and only the clear transparent glass slide remained. Surface roughness of the AAO layer fabricated within the experimental conditions ranged from 2.85 to 9.07 nm, as indicated in Figure S1A,C(b), all within the “capable region” (<18 nm) for glycan microarrays. At constant low reaction temperature (4 °C), the reaction time had a minor effect on the thickness growth of the AAO layer, as indicated by the response measurement shown in Figure C. However, an optimized surface of the ACG slide (with rms at around 2 nm) was fabricated under the combined conditions of voltage and reaction time at 4 °C. Therefore, we proposed the AAO growth on the Al-coated glass slide into two regions: (1) the “major AAO growth” (fast thickness increasing) region and (2) the “electropolishing (surface smoothing) growth” region, as presented in Figure A).

Figure 6

(A) On-time measurement of the AAO layer formation under various voltages and reaction times. The starting pure Al-coated glass slide was as smooth as glass. The AAO growth (thickness increasing) depends on the voltage at the beginning, and electropolishing (surface smoothing) and extended AAO growth occur later. (B) Fluorescence intensity differences in ACG slide no.5 of RSM (response surface measurement) study vs NHS glass slide at various sugar solution concentrations arrayed on the surface.

Figure B shows the comparison of fluorescence intensity of mannose/ConA-A488 binding on ACG slides no. 5 from the RSM (response surface measurement) experiments vs the conventional NHS-activated glass slide. The ACG slide gave higher fluorescence intensity than the NHS glass slide resulting from the fluorescence tagged ConA binding to mannose on the surface. The differences in these two types of slides (ACG slide vs NHS glass slide) have been investigated further.

As shown in Figure , the particle counts and particle height of mannose derivatives on the ACG slide (rms 4.00 nm) and on the NHS glass slide (rms 1.02 nm) were analyzed using AFM scanned over the matrix of 10 × 10 μm2 area. Particle counts were obtained by counting the number of particles above the height of one half width of the particle height distribution. Mannose derivatives can be covalently bound to the surface only where the activated (either Al-oxide or NHS) functional groups are available. The ACG slide provides a surface of more uniformly distributed reactive sites for covalent reaction than that of the NHS glass slide. Particle height changes on the slide surfaces are summarized and given in Table S6. Higher rms indicated higher surface roughness of the ACG slide than that of the NHS-activated slide. The mean height varied before and after sugar grafting and protein binding is an indication of conformational change. Along this concept, as mannose was covalently bound on the slide surface, the mean, minimum, and maximum particle height on the ACG slide were higher than those on the NHS slide, suggesting that the mannose derivative existed in a more extended structural conformation perpendicular to the slide surfaces, as shown in Scheme .

Figure 7

Particle counts and particle height of mannose derivatives covalently bound to ACG slide (A) and NHS slide surfaces (B) (analyzed using AFM) suggested more uniformly distributed sugar molecules on the AAO surface than that on the NHS glass slide.

Scheme 1

Conformation of Mannose on ACG Slide vs NHS-Activated Glass Slide

Figure A,B shows the GenePix scanning images (at PMT 380) of ConA-A488 bound to mannose (1 mM) arrayed on the ACG slide versus the NHS-activated glass slide. As can be seen, the Al-oxide surface has a higher sugar loading capacity such that the array spots are reaching saturation even at the very low value of photomultiplier tubes (PMT) of 380. The fluorescence intensities of the averaged 20 spots for the ACG slide and NHS slide are given in Figure C. The GenePix scanner has a resolution up to 5 μm (25 μm2 per 1 pixel), and the dimension of the arrayed spots ranges from 60 to 250 μm in diameter with corresponding 110–1960 pixels, respectively. This has made it possible for us to access the uniformity/distribution of covalently bound sugars within an array spot. Figure D shows the analysis of spot 1–5 (the first row and the fifth spot counting from the right) of ACG versus NHS slides as shown in Figure A,B. The result from the GenePix scanner recorded the spot dimension (Dia.), average fluorescence intensity per pixel (F488 Mean) with standard deviation (F488 SD), total fluorescence intensity of the spot (F488 total intensity), percentage saturation (F488% sat.), and coefficient of variation within the spot (F488 CV) are tabulated in Figure D. Even though the individual pixel intensity cannot be resolved visually with the naked eye, the coefficient of variation of pixel intensities (F488 CV) of 17 versus 65 indicates that the ConA-A488/mannose distribution within a specific array spot on the ACG slide is more uniform than that on the NHS glass slide.

Figure 8

(A) GenePix scanning images (at PMT 380) of ConA-A488 bound to mannose (1 mM) on ACG slide (A) vs that on NHS glass slide (B). (C) ConA-A488/mannose (1 mM) binding on ACG slide vs NHS-activated glass slide. The fluorescence intensities of the averaged 20 spots for ACG slide vs NHS glass slide. (D) Spot analysis of ACG slides vs NHS-activated glass slides. (E) Confocal microscope (Leica SP8) images of ConA-A488/mannose binding on ACG slide (left) vs NHS-activated glass slide (right).

The GenePix scanner has a maximum resolution of 5 μm, whereas confocal microscopes have better resolution up to several hundred nanometers (1/2λ). Figure E shows the confocal microscope images of ConA-A488/mannose binding (Lica SP8) on ACG slide versus NHS-activated glass slides of approximately 30 × 30 μm2 within the spots (using Leica SP8). Consistent with the results obtained from GenePix fluorescence intensity and AFM scanning, the amorphous Al-oxide on the ACG slide provides not only higher sugar loading capacity but also more uniformly distributed glycan molecules on the surface. The ACG slide should serve as a better surface for glycan array to facilitate our understanding of glycan–protein interaction.

Conclusions

A smooth surface anodization of aluminum-coated glass slide has been developed to fabricate the AAO layer with thickness optimized to 50–65 nm from the 300 nm aluminum-coated glass slide for the glycan microarray. The fabrication of AAO layer has been considered in two regions: (1) the “major fast AAO growth” region and (2) the “electropolishing growth” region. At constant temperature, the combined effects on voltage and reaction time are related to the thickness of AAO growth. Using the Con A/mannose binding system, the sugar loading capacity and uniformity/distribution of the reactive site for sugar attachment to the ACG slide have been compared with those of the NHS-activated glass slide. The ACG slide has a surface that is more stable, higher loading capacity, and more uniformly distributed reactive site for sugar arraying through covalent phosphonate formation. This array system is more convenient to prepare and should be useful for the study of multivalent interaction and heteroligand binding in addition to the traditional use in the study of protein–sugar interaction.

Experimental Section

Fabrication of the Aluminum-Coated Glass Slide

The aluminum (99.999% purity)-coated glass slides were fabricated by the Vacuum and Thin Film Technology Division, Instrument Technology Research Center at the National Applied Research Laboratories, Hsinchu Science Park, Hsinchu, Taiwan. The VDP equipment of the ion beam-assisted (Veeco 16 cm RF ion source) E/B gun coater was assembled by the domestic F.S.E. Corporation. The glass slides were purchased either from Schott Nexterion (Glass B cleanroom cleaned) or from Arrayit Corporation (SuperClean 2 Microarray Substrate SMC2) with nominal dimensions of 75.6 × 25 × 1 mm3. The thickness of the coated aluminum layer was fixed at 100 and 300 nm. After being coated with pure aluminum, the slides were packed immediately (one slide per container under the nitrogen atmosphere), vacuum-sealed with an air-tight laminated foil, free from exposure to oxygen to prevent the formation of NAO, and kept sealed until the electrochemical reaction for surface anodization. The thickness, surface roughness, and particle counts of the AAO layer were analyzed using nondestructive ellipsometry (SOPRA ES4G) and AFM (Veeco di Dimension 3100 SPM), and the surface morphology and cross section view were examined using SEM (FE-4300).

Surface Anodization of the Aluminum-Coated Glass Slide

The electrochemical reaction was conducted in 0.3 M oxalic acid aqueous solution in a 4 °C temperature-controlled incubator. Inside of the incubator, a thermocouple was dipped into the water-bath acid solution to ascertain the reaction temperature. A plastic plate was fabricated in-house to hold the platinum electrode and a fixture of the positive electrode such that the aluminum-coated glass slide could be clamped or removed easily from the power supply (Keithley 2400). Surface anodization was controlled by voltage and reaction time varying from 9 to 54 volts and from 48 to 200 s, respectively, depending on the designed experiment (using Design Expert 8.0). After surface anodization, the slide was washed thoroughly with deionized water, purge dried with nitrogen gas, and then annealed in a 100 °C oven for 10 min. After being cooled to room temperature, this ACG slide was stored overnight in a 30% relative humidity chamber at room temperature, ready for the following experiments.

AAO coating thickness analyzed using ellipsometry (SOPRA ES4G).

AAO surface roughness analyzed using AFM, Veeco di Dimension 3100 SPM.

Microarray (BioDot AD3200G) and covalent formation of mannose-α-5-pentylphosphonic acid with Al-oxide on the slide surface. The mannose derivative was dissolved in a 30:70 ratio of water/ethylene glycol mixture with solution concentration (depending on the experiments) varying from 100 mM to 1 pM (i.e., 12 samples with consecutive serial dilutions), microarrayed, and kept in a 80% humidity chamber for 2 h, then stored in a 30% HR chamber, ready for protein binding analysis the next day.

Binding of ConA-A488 to mannose. ACG slides were loaded on the FAST frame (maximum of four slides per frame) with each slide divided into 16 wells. The mannose derivative was arrayed in 10 × 10 or 10 × 12 matrices per well (depending on the experiments). The ConA-A488 (100 μL, 33 μg/mL) buffer solution was filled into each well for sugar/protein binding, which took about 30 min to 1 h at room temperature. Following the incubation for sugar/protein binding, standard washing (three times bovine serum albumin/phosphate binding buffer, three times phosphate-buffered saline/Tween buffer, and three times D.I. water) was conducted to remove the noncovalently bound sugar, mobile sugar bound protein, and any excess protein. The slide was dried carefully and subjected to a fluorescence intensity reading using a GenePix 4300A microarray scanner.

Computer analysis to obtain the optimized AAO fabrication condition. The reaction variables and response measurements were analyzed (using Design Expert 8.0). Data analysis has revealed the surface of AAO fabrication conditions with respect to the fluorescence intensity of sugar/protein binding on the surface.

Characterization of Sugar Loading Capacity, Uniformity, and Structural Conformation of Sugar Derivatives Covalently Bound to the Slide Surfaces

(1) Slide surfaces with covalently bound sugars were examined under AFM. The outcome essentially gave information on the sugar loading capacity and the uniformity of sugars on the ACG slide compared with that on the NHS glass slide. (2) The conformational differences in the sugar derivatives covalently bound to the slide surface were also examined via the particle height analysis using AFM. (3) The microarray sugar/protein binding data of individual spots were examined using a GenePix scanner; within an arrayed spot, the “average pixel intensity” and its coefficient of variation (% CV) were reported. (4) The same sugar/protein binding slides were also examined under a confocal microscope.