Nanopatterned submicron pores as a shield for nonspecific binding in surface plasmon resonance-based sensing

Abstract

We present a novel approach to tackle the most common drawback of using surface plasmon resonance for analyte screening in complex biological matrices – the nonspecific binding to the sensor chip surface. By using a perforated membrane supported by a polymeric gel structure at the evanescent wave penetration depth, we have fabricated a non-fouling sieve above the sensing region. The sieve shields the evanescent wave from nonspecific interactions which interfere with SPR sensing by minimizing the fouled area of the polymeric gel and preventing the translocation of large particles, e.g. micelles or aggregates. The nanopatterned macropores were fabricated by means of colloidal lithography and plasma enhanced chemical vapor deposition of a polyethylene oxide-like film on top of a polymeric gel matrix commonly used in surface plasmon resonance analysis. The sieve was characterized using surface plasmon resonance imaging, contact angle, atomic force microscopy and scanning electron microscopy. The performance of the sieve was studied using an immunoassay for detection of antibiotic residues in full fat milk and porcine serum. The non-fouling membrane presented pores in the 92–138 nm range organized in a hexagonal crystal lattice with a clearance of about 5% of the total surface. Functionally, the membrane with the nanopatterned macropores showed significant improvements in immunoassay robustness and sensitivity in untreated complex samples. The utilization of the sensor built-in sieve for measurements in complex matrices offers reduction in pre-analytical sample preparation steps and thus shortens the total analysis time.

---

Introduction

Surface Plasmon Resonance (SPR) sensing offers a powerful platform for molecular interaction monitoring, whose capability has been abundantly demonstrated for a wide variety of applications.^1–5 In the analytical field, SPR-based biosensors are positioned as rapid, quantitative, real-time and often automated techniques. However, many SPR-based sensing analytical applications are hampered by the reduced performance in complex samples, which has been recently recognized as a severe limitation in the further development of the technology. Even though a separate sample pre-treatment is an option for some applications, the possibility of avoiding it is one of the major advantages of SPR over traditional analytical methods and therefore has an important added value that should be preserved. The matrix interference with the bioassay performance is mainly attributed to nonspecific binding (NSB) of the sample components to the sensor chip surface.^6,7 The inability to limit NSB of several sample components, including lipids, lipoproteins and whole micelles, causes unstable, high background signals and poor repeatability of measurements in complex samples presenting a major technical challenge in bioanalytical assay development. The difficulties in bioassay transfer from optimal buffer conditions to real-life samples are experienced by many researchers, compromising application possibilities of many bioassays and naturally leading to few publications that describe the problem. For example, at low analyte concentrations, the specific ligand–analyte interaction can be masked by the nonspecific protein adsorption up to the point where the analyte is no longer detectable. This is the major difficulty in the development of SPR biosensors aimed towards medical monitoring, where the measurements are performed in undiluted serum.⁸

NSB essentially is non-covalent binding, which is driven by hydrophobic, electrostatic and exchange interactions between molecules, usually attributed to protein adsorption.^9,10 Since SPR is sensitive to a total refractive index change in close proximity to the sensor chip surface, there is no difference between the specific interaction that takes place between the ligand on the chip surface and the analyte in the sample, and nonspecific interaction between the sample components and the sensor chip surface.^7,11,12 Most of the research has been directed towards reducing the nonspecific binding either via introduction of new polymers as alternatives for commonly used carboxymethyldextran (CM-dextran) or via sample treatment with NSB blocking additives.^6–8,13 For example, Masson et al. looked at bioassay performance in full bovine serum using several biocompatible polymers on the SPR sensor chip. Even though some of them were found to be less prone to nonspecific binding, the CM-dextran still offered the largest signal for antigen detection.⁸ In a different study by Elliot et al., a buffering system, supplemented with soluble CM-dextran, was reported to reduce the nonspecific interactions of bovine serum components with the sensor chip. The most substantial reduction achieved with this approach still required 10-fold dilution of the sample. Moreover, in that study, even under the optimized conditions, the total binding signal in the matrix was lower and the IC₅₀ values were higher in 10-fold diluted bovine serum than in the buffer.⁶ Very recently Bolduc et al. reported reduced non-specific adsorption of proteins on short polar and ionic peptide monolayers in comparison to other monolayers.¹⁴ However when high ligand loads are needed for a working assay, using monolayers presents a disadvantage. Additionally, nonspecific adsorption of other sample matrix components, such as lipids, was not addressed in that particular study.

To limit the NSB, we propose a novel approach, to place a submicron porous filtering layer on top of the CM-dextran polymeric matrix above the ≈350 nm penetration depth of the Surface Plasmon Wave (SPW), using dextran with a nominal thickness of 500 nm. At this SPW penetration depth, up to 76% of the signal caused by adsorption of supramolecular structures, e.g. large aggregates and micelles, to the filtering layer can be reduced. The use of higher dextran, to distance the filtering layer even further from the SPW probing area, would compromise assay performance due to diffusion limitations and concentrated ligand immobilization away from the sensing region. To set up the assay, the ligand can be covalently immobilized in CM-dextran through the filtering layer, allowing the specific interaction with the analyte. The filtering layer would reduce the nonspecific adsorption by limiting the diffusion of small and preventing large matrix components from reaching the SPW probing region where the biointeraction occurs.

This non-fouling filtering layer bearing the nanopatterned macropores was fabricated by means of colloidal lithography and plasma enhanced chemical vapor deposition of polyethylene oxide on top of CM-dextran (PEO-pores).¹⁵ Many nanoporous materials are used in biological applications; however submicron nanopatterned polyethylene oxide-based membranes have not been reported before.¹⁶ The influence of PEO-pore sensor chip coverage on biomolecular interactions in buffers and in complex samples was studied using imaging SPR in direct comparison to commonly used CM-dextran. As a model for analytical application, a previously described SPR-based immunoassay for enrofloxacin detection was used.¹⁷ For complex samples, with target analyte pertinence, full fat milk and porcine serum were chosen. Full fat milk contains high amounts of protein as well as fat and casein micelles which seem to interfere with the SPR measurements.¹⁸ Even though the specific factors causing NSB in porcine serum are presently unidentified, it is also considered to be a difficult matrix for analysis with many analytical methods including SPR biosensors.

Experimental

Chemicals and materials

Round sensor chips, coated with 500 nm or 200 nm high carboxymethyldextran with a medium degree of cross-linkage, were purchased from Xantec bioanalytics GmbH (Duesseldorf, Germany). The Biacore amine coupling kit (containing 0.1 M N-hydroxysuccinimide (NHS), 0.4 M N-ethyl-N-(3-dimethylaminopropyl) carbodiimide hydrochloride (EDC) and 1 M ethanolamine hydrochloride pH 8.5 HBS-EP buffer) were purchased from GE Healthcare (Uppsala, Sweden). The norfloxacin–NH₂ derivative (NorF) was kindly supplied by Dr Sheryl Tittlemier (Health Canada, Ottawa). Rabbit polyclonal antiserum (MAR06101), raised against a norfloxacin–COOH derivative was kindly supplied by Laboratoire d'Hormonologie Animale (Marloie, Belgium). The monoclonal anti-κ-casein (Mab 4G10) antibody was previously described elsewhere.¹⁹ Porcine serum was kindly provided by Central Veterinary Institute (Lelystad, Netherlands). The full-fat goat milk powder (Mekkermelk from Henri Willig, Katwoude, The Netherlands) was purchased locally. The iSPR instrument, round sensor chip holder, refractive index matching oil (n = 1.518), hemispheric prism (BK7) and a 4 μL flow cell were purchased from IBIS Technologies B.V. (Hengelo, The Netherlands). Diethylene glycol dimethyl ether (DEGDME) (CH₃OCH₂CH₂)₂O, polystyrene beads (500 nm diameter, 10% monodispersity) and the rest of the chemicals were purchased from Sigma-Aldrich (Zwijndrecht, The Netherlands).

Colloidal lithography

The PEO-pores were produced by nanosphere lithography. Briefly, a polystyrene bead monolayer nano-mask was deposited as described elsewhere,²⁰ by spin-coating a bead solution at 350 rpm for 60 seconds on a 500 nm thick dry carboxymethylated dextran layer covering gold sensor chips.

Plasma deposition of the polyethylene oxide (PEO) layer

The deposition of a film of plasma polymerized polyethylene oxide was performed in a homemade stainless-steel (300 mm × 300 mm × 150 mm) reactor with two symmetrical internal parallel-plate electrodes (diameter of electrodes = 140 mm, distance between the two electrodes = 50 mm). The plasma was generated by a radio frequency generator (13.56 MHz) connected to the upper electrode whereas the other electrode was grounded and used as a sample holder. The plasma polymerization was carried out by using a pulsed plasma discharge (time on = 10 ms, time off = 100 ms, nominal power = 5 W) of DEGDME vapors. The thickness of the PEO on a flat surface was aimed at 10 nm; however it was expected that the thickness would vary due to the sinking of the colloidal mask into the soft hydrophilic substrate. Further description of the reactor can be found elsewhere.²¹ For the nanomask lift-off, after the plasma PEO deposition, the sensor chips were submerged in distilled water and sonicated for five minutes.

Atomic force microscopy

Nanofabricated and control sensor chip surfaces were imaged using an Agilent AFM 5500 microscope (Agilent Tecnologies Inc. – Santa Clara, California, U.S.A.) under a standard setup, in Magnetic AC tapping mode, using MAClever type V silicon nitride tips (spring constant = 0.5 N m⁻¹).

Scanning electron microscopy and focused ion beam (SEM-FIB) milling

The PEO-pores on the dextran chips were imaged using a SEM-FIB FEI Nova 600I Nanolab (FEI, Eindhoven, The Netherlands) at relatively low acceleration voltages (5 kV) that allow the use of an immersion lens detector without significant loss of resolution. Samples were grounded contacting the conductive gold layer to the ground of the microscope to avoid charging effects on the images.

In order to access and measure the porous membrane region, the sensor chips were first dehydrated using an ascending ethanol series, and then critical point dried using an EMITECH K850 reactor. Prior to the milling process, we deposited inside the microscope a thin strip of platinum on top of the membrane to avoid it from collapsing and to protect it from scattered ions. For this purpose we used a Pt gas precursor activated by the electron column at high acceleration voltages (30 keV). Then ion milling was done using the minimum current possible and highest acceleration voltage (1.5 pA and 30 keV respectively). To minimize the cutting time and the damage produced by scattered ions without losing the sight of the cut edge, we designed a 52 degree trapezoidal cut in the plane perpendicular to the membrane instead of a rectangular one.

Ligand microarraying

The sensor chip, mostly covered by PEO-pores, was activated offline with the EDC–NHS (1 : 1) mixture, rinsed with 5 mM acetic acid, dried under a nitrogen stream and immediately spotted with ligands using the continuous flow microfluidic (CFM) spotter (Wasatch Microfluidics, Salt Lake City, USA). The ligands were prepared beforehand as follows. NorF was dissolved in 20 mM carbonate buffer pH 8.5 with 30% (v/v) dimethylformamide (DMF) and immobilized in 10 mM carbonate buffer pH 9.6 at a final concentration of 0.3 mM. The anti-κ-casein antibody was protein G purified, desalted into 20 mM acetate buffer pH 4.5 and immobilized at a final concentration of 0.05 μg mL⁻¹. During the spotting, 6 immobilization cycles of 5 minutes each were applied. 20 mM acetate buffer pH 4.5 was used as a priming buffer during the spotting. Unreacted groups in the CM-dextran were blocked with 0.5 M ethanolamine pH 8.5 for 10 minutes. If not used directly, the sensor chip was washed with RO water, dried under a nitrogen stream and stored at 4 °C.

iSPR measurements

iSPR measurements were conducted using the IBIS iSPR instrument with a light source of 840 nm and evanescent wave penetration depth of about 350 nm, approximated according to the equation Dp = (λ₀/2π)*((ξ′m + ξ′d)/ξ′d²)^1/2, where, λ₀ is the wavelength, ξ′m the real part of the permittivity of the metal and ξ′d the real part of the permittivity of the dielectric. Assuming ξ′m = −29.88 of gold and ξ′d = 2.017 of the hydrogel in the matrix. The sensor chip was assembled with the prism using refractive index matching oil in a round chip holder and the flow cell was fixed on top of the sensor chip surface. The sample was delivered to the sensor chip surface through tubing and was pumped back and forth at 10 μL s⁻¹ during the interaction (7 min). The surface was equilibrated with the HBS-EP buffer (containing 10 mM 4-(2-hydroxyethyl) piperazine-1-ethanesulfonic acid pH 7.4, 150 mM sodium chloride, 3 mM EDTA, 0.005% and (v/v) surfactant polysorbate (P20)) and regions of interest (ROIs) of size 300 μm × 300 μm were defined on the spots. The SPR angle was scanned on each pre-defined ROI in the range between −3° and +1° in steps of 200 m° (one data point every 5 seconds). SPR curves were fitted automatically by IBIS software ver. 4.5 while curve parameters were limited to 20 points before and after the dip. All the measurements were performed in the “analysis mode”, recording the SPR angle shift (m°) as a function of time (seconds). Subsequently, SPR data were analyzed using SPR inspection tool software ver. 1.6.0.0 (IBIS Technologies B.V., Hengelo, The Netherlands). Post-measurement data sampling for each angle shift was done by averaging at least five data points collected around the desired time. Raw sensorgrams were first zeroed to the angle before the injection and then referenced to the angle of the blank spot (REF), either on PEO-pores or on CM-dextran. The maximum binding responses on NorF or αCas spots and non-specific binding on REF spots were calculated from the angle shift during the dissociation phase (around 600 seconds). Bulk responses and baseline build-up on REF spots, were calculated from the angle shift before the end of the sample injection (around 550 seconds) and the angle shift after the regeneration (around 750 seconds), respectively.

Immobilization check and concentration measurements

A freshly prepared sensor chip, partly covered with PEO-pores and microarrayed with NorF and αCas, was first tested for the ligand immobilization efficiency. Since the regeneration solution for the interaction of norfloxacin and anti-enrofloxacin antibody causes loss of anti-κ-casein antibody activity, the immobilization of αCas was checked first. The surface was conditioned with at least three serial injections (two minutes contact time each) of a regeneration solution containing 10 mM HCl until the baselines of αCas spots were stable in HBS-EP buffer. κ-Casein was injected at a final concentration of 5 μg mL⁻¹ in HBS-EP buffer and the maximal binding responses on αCas were measured. The same was repeated with anti-enrofloxacin (1 : 200) in HBS-EP on NorF spots, only under harsher regeneration conditions: 20% (v/v) Acetonitrile in 10 mM NaOH. Surface saturation rate was tested by a longer injection of anti-enrofloxacin (1 : 400) in HBS-EP on NorF spots. From these steps, a final anti-enrofloxacin dilution (1 : 440) was selected according to the obtained responses, considering sufficiently high response without baseline build-up. Next, enrofloxacin concentration measurements were performed in HBS-EP buffer, full fat milk and porcine serum. Enrofloxacin standard solutions were prepared in HBS-EP buffer at concentrations ranging from 0.005 to10 000 ng mL⁻¹ and each was mixed with anti-enrofloxacin (1 : 220). These mixtures were injected over the sensor chip in duplicate, starting with blank solution containing only the antibody. The measurement cycle included sample injection (7 minutes contact time) and one injection of regeneration solution (1 minute contact time). For measurements in complex matrices, full fat goat milk and porcine serum were diluted ten times in HBS-EP buffer to prevent high bulk responses which would shift the SPR angle outside of the scanning range of the instrument. 1 g of full fat goat milk powder was dissolved in 9 mL of HBS-EP buffer, stirred for 0.5 h at RT and diluted ten times in HBS-EP buffer and porcine serum was directly diluted ten times in HBS-EP buffer, no filtering or centrifugation steps were used. Enrofloxacin standard solutions in diluted milk and serum were prepared and measured in the same way as in buffer on the same sensor chip. Relative binding (B/B₀) was calculated by dividing the response of the enrofloxacin containing solution (B) by the response of the blank solution (B₀). To generate calibration curves, B/B₀ values were plotted against enrofloxacin concentrations. The calibration curves were fitted with a non-linear 4-parameters (4P) model using GraphPad Prism software ver. 5.02 (GraphPad Software Inc, San Diego, USA) and half maximal inhibitory concentration (IC₅₀) was interpolated. Additionally, immunoassays were characterized by the limits of detection (LODs) which were calculated by subtracting three standard deviations from the average maximum response of the blank solution and by the dynamic measurement ranges, which were set between 0.2 and 0.8 B/B₀.

Results and discussion

Nanofabrication and characterization of PEO-pores

The diagram in Fig. 1A shows the nanofabrication process of the pores in PEO membrane on top of the CM-dextran sensor chip. The sensor chip surface was spin coated with a polystyrene nano-sphere mask achieving approximately 65% coverage. One half of the chip was masked during the PEO plasma deposition, to include a control area without the sieve (CM-dextran). On the other half, a PEO layer was plasma deposited on the bead nanomask, with a controlled deposition rate to yield 10 nm thickness on a flat substrate. However, the colloidal mask sunk into the soft porous hydrophilic substrate and the obtained PEO thickness was not uniform, but continuous and perforated upon lift-off as can be seen in the AFM images (Fig. 1B).

Fig. 1 Sensor chip nanofabrication process and surface characterization. (A) The nanomask of 500 nm polystyrene beads was spin-coated on the carboxymethylated dextran (CM-dextran) – covered sensorchip. Half of the sensor chip surface was masked and PEO was plasma deposited, producing a PEO layer on top of the CM-dextran. The liftoff was done by sonication. (B) The surface topography, pore distribution and the height of the polymer were characterized by means of atomic force microscopy (AFM), scanning electron microscopy (SEM) and ion milling. A wedge was milled on the surface through all the layers down to the glass substrate using a focused ion beam (FIB). The thickness was estimated by measuring the distance from the gold layer to the base of the protective platinum strip on PEO layer on a 52 degrees tilted image.

The macropores had a hexagonal lattice and the diameter of each macropore was estimated topographically to be 92 ± 15 nm with atomic force microscopy (AFM) and about 138 ± 21 nm with scanning electron microscopy (SEM) (Fig. 1B). The AFM image shows the footprint of each bead on the dextran hydrogel which differs from the flat perforated membrane displayed by the SEM image where only the electron contrast of the surface is visible. A wedge was milled on the surface using a focused ion beam (FIB) and thickness was estimated from the gold to the PEO layer on a 52 degree tilted image (Fig. 1B). The pores in this image are faded due to the charging of the surface at the edge of the cut and the damage introduced by the scattered ions during the milling. The CM-dextran height was measured with the SEM after critical drying of the chip to prevent collapsing of the polymer. The thickness with the sieve on top was estimated to be 350 nm, using the distance from the gold surface where the milling rate was decreased to the top of the PEO layer. The difference in height obtained for these sensor chips, which are commercialized as 500 nm dextran, is significant and probably results from the partial collapse of the polymer during the critical drying process. Placing the sieve at a minimum height of 350 nm, was sufficient for the purpose of shielding the binding sites from nonspecific binding and a minimum of 63% and up to 76% of the evanescent wave with a ≈ 350 nm penetration depth. The area of the CM-dextran, which was masked to prevent PEO deposition, showed superficial random roughness in AFM (Fig. 1B) and a significantly higher hydrophilicity (contact angle 25.5 ± 0.9°), when compared to PEO-pores (contact angle 52 ± 1.8°).

Biomolecular interactions on PEO-pores

To study PEO-pore compatibility with applications in biomolecular interaction analysis, two main points were addressed: surface capacity to ligand immobilization and analyte binding to the ligand. High and low molecular weight compounds, the anti-κ-casein antibody (αCas) and norfloxacin antibiotic (NorF), were covalently immobilized on both halves of the sensor chip (CM-dextran and PEO-pores) using a continuous flow microfluidics spotter (CFM) (Fig. 2A and B). Spots 4, 8 and 6 appear brighter than the rest, corresponding to higher angle shift due to higher molecular weight ligand (αCas) immobilization. As expected, on the rest of the spots the immobilization is not evident from the SPR image, due to the low molecular weight of NorF (see Fig. 2C). Brighter areas in the iSPR image in Fig. 2C correspond to higher PEO density coverage due to the absence of a colloidal mask during the spin-coating. Even though the mask was placed where the horizontal scratch mark appears, the PEO coating overspills partially to the untreated side of the surface, as can be seen from the SPR image. Only the spots which were further away from the mask border were used for the comparison.

Fig. 2 Sensor chip lay out. (A) Schematic representation (out of scale) of immobilized ligands on the sensor chip. Both high molecular weight (antibody) and low (antibiotics) molecular weight ligands were covalently immobilized, via primary amines, on the PEO-pores and on CM-dextran simultaneously using continuous flow spotter (CFM). The immobilization was tested with injections of anti-norfloxacin antibody and κ-casein. (B) Spotting lay out for CM-dextran and for PEO-pores, including norfloxacin–NH₂ spots (NorF), anti-casein antibody spots (αCas) and blank spots which were used for signal referencing in imaging surface plasmon resonance measurements. The octagonals show the spotted regions. On spots number 4, 6 and 8 anti-casein antibody was immobilized and on the rest norfloxacin–NH₂. The double line in the middle shows the applied mask border during the PEO-pores production. (C) iSPR image of the spotted sensor chip. Arrow indicates brighter regions with isotropic PEO deposition.

To be able to monitor the binding reaction on NorF spots, the grid of the regions of interest (ROIs) was placed according to the αCas spots. To test the immobilization efficiency, κ-casein and anti-enrofloxacin antibody were serially injected over the spotted surfaces. Maximal binding responses measured on αCas spots were comparable for PEO-pores and CM-dextran, both showing good ability for high molecular weight ligand immobilization by diffusion through the sieve (Fig. 3A). Maximal binding responses measured on NorF spots were 30% lower on PEO-pores than on the CM-dextran surface (Fig. 3A) and showed 1.5 times faster surface saturation rate on PEO-pores than on the CM-dextran surface (Fig. 3B) indicating a slightly lower immobilization efficiency of norfloxacin on PEO-pores or lower pore coverage in the NorF spots than the average. The differences between PEO-pore capacity for the anti-casein antibody and norfloxacin antibiotic can be attributed to the different hydrophobicities of the two ligands. Norfloxacin, despite being a small molecule and expected to diffuse easily through the PEO-pores, is quite hydrophobic and might pre-concentrate on the less hydrophilic PEO membrane, instead of diffusing further to the hydrophilic CM-dextran, where the immobilization occurs. Overall, although the PEO-pores reduce in about 95% the exposed CM-dextran area, they show no substantial interference either with the immobilization of large and small molecules on the sensor chip surface, or with ligand–analyte binding, thus suggesting a good compatibility with biomolecular interaction analysis.

Fig. 3 Biomolecular interactions on PEO-pores. (A) Maximal binding responses measured on anti-casein and NorF spots, both on PEO-pores (hollow blue) and on CM-dextran (black) surfaces, after serial injections (7 minutes each) of κ-casein (5 μg mL⁻¹) and anti-enrofloxacin antibody (1 : 200). Error bars show standard deviations between the spots during triplicate injections. (B) Representative blanked and referenced sensorgrams measured on NorF/PEO-pores (blue) and NorF/CM-dextran (black) spots during prolonged injection of anti-enrofloxacin antibody (1 : 400), showing surface saturation rates.

Analytical performance of PEO-pores in complex matrices

To study the performance of PEO-pores in complex matrices, an immunoassay for enrofloxacin detection was performed in full fat milk and untreated porcine serum. This assay was formerly established for the generic detection of fluoroquinolones using SPR biosensing (Huet et al. 2008). In that study, norfloxacin was used as the immunizing hapten to produce a generic antibody against fluoroquinolones which showed 100% cross-reactivity with enrofloxacin.

Usually, milk and serum samples undergo a pre-treatment, e.g. extraction, filtering and centrifugation, prior to analysis with the SPR biosensor in order to remove interfering matrix components.^18,22 Here, full fat milk and crude serum samples were only subjected to dilution in the running buffer. The immunoassay was performed in the competitive format, based on inhibition of antibody – binding to the antibiotic immobilized on the surface by the antibiotic in solution. Enrofloxacin standard solutions were prepared in buffer, milk and serum at different concentrations, mixed with the anti-enrofloxacin antibody and injected over the norfloxacin spotted sensor chip, covered with PEO-pores and with CM-dextran (Fig. 2B and C). Each measurement cycle included surface equilibration with the buffer, sample injection, dissociation in the buffer, and regeneration and re-equilibration with the buffer as shown in Fig. 4A. The same figure shows the raw sensorgrams obtained on the NorF/PEO-pores (blue) and on the NorF/CM-dextran (black) with anti-enrofloxacin containing porcine serum. Sensorgrams on PEO-pores showed lower binding responses than on CM-dextran, as observed earlier for measurements in the buffer (Fig. 3A).

Fig. 4 Concentration measurements in complex matrices on PEO-pores. (A) Raw sensorgram measured on NorF/PEO-pores (blue) and NorF/CM-dextran (black) spots during duplicate injections of anti-enrofloxacin antibody in diluted porcine serum. Each measurement cycle lasted 13 minutes, including 7 minutes incubation with the sample and 2 minutes regeneration with 20% acetonitrile in 10 mM NaOH. Maximum binding responses were calculated from the angle shift during the dissociation phase (around 600 seconds), after zeroing and referencing to a blank spot. (B) Zeroed and referenced sensorgrams measured on NorF/PEO-pores (blue) and NorF/CM-dextran (black) spots during injections of anti-enrofloxacin antibody mixed with enrofloxacin spiked into a diluted porcine serum at 0, 0.005, 0.06, 0.675, 7.5, 82.6, 909 and 10 000 ng mL⁻¹ final concentrations. (C–E) Enrofloxacin calibration curves measured in buffer, diluted full fat milk and diluted porcine serum respectively, On NorF/PEO-pores (blue) and NorF/CM-dextran (black) and on separate 200 nm CM-dextran sensor chip (grey). B/B₀ stands for relative binding, calculated from the maximum binding response of the sample containing antibiotics (B) and the maximum binding response of the sample containing only the antibody (B₀). Insets indicate the limits of detection calculated for PEO-pores and control CM-Dextran in each matrix. The error bars represent standard deviations between three measurements performed in duplicate on different days. The lines show curves fitted with non-linear 4-parameters variable slope model.

Further on, the porcine serum was spiked with increasing enrofloxacin concentrations, mixed with anti-enrofloxacin and serially injected over the chip surface (Fig. 4B). The raw sensorgrams measured on NorF/PEO-pores and on the NorF/CM-dextran were zeroed and referenced to REF/PEO and REF/CM spots respectively. On both surfaces a decrease in signal was observed with increasing enrofloxacin concentration in the serum. Also, the binding of anti-enrofloxacin to the enrofloxacin spot was not completely inhibited by high enrofloxacin concentrations on CM-dextran, indicating non-specific binding occurrence on CM-dextran but not on the PEO-pores. For each surface, calibration curves in buffer, in milk and serum were measured and plotted using relative binding values (B/B₀) as a function of enrofloxacin concentration (ng mL⁻¹) (Fig. 4C–E). The curves were fitted with the 4-parameter non-linear model and the limit of detection (LOD) was calculated (Fig. 4C–E insets). For comparison, the same calibration curves were measured on a standard 200 nm CM-dextran sensor chip, which is commonly used for most SPR-based assays (grey triangle in Fig. 4C–E). The calibration curves measured in the buffer on PEO-pores and CM-dextran completely coincided, and consequently displayed a very similar LOD, IC₅₀ and dynamic range showing better performance when compared with the separate 200 nm CM-dextran sensor chip commonly used for these applications. In milk and serum (Fig. 4D and E), the calibration curves measured on CM-dextran reached an 80% inhibition plateau around 80 pg mL⁻¹ enrofloxacin, showing no response to further increasing concentrations. On the contrary, the calibration curves measured on PEO-pores showed full signal inhibition in milk at 2000 pg mL⁻¹ enrofloxacin, and 95% inhibition at 10 000 pg mL⁻¹ enrofloxacin (Fig. 4C–E). Calibration curves that do not reach full inhibition at the highest analyte concentrations usually suggest nonspecific binding, since the analyte is not able to fully suppress the signal. This phenomenon narrows the detection range of the assay and is commonly observed in complex matrices with various degrees of interference. When the nonspecific binding causes the plateau at less than 70% inhibition, the assays are usually considered to be not applicable to measurements in real life samples.

Assay robustness also presents a challenge for SPR-based measurements in complex matrices. Often, the variation between independent measurements is too high to produce reliable results. Here the PEO-pores showed lower standard deviations, both in milk (10% on PEO-pores versus 15% on CM-dextran) and in serum (2% on PEO-pores versus 10% on CM-dextran) than the CM-dextran. The same trend was observed in comparison to 200 nm CM-dextran. These results suggest that PEO-pores provide better controlled conditions for bioassay performance in complex matrices and are capable of improving assay robustness and consequently assay limit of detection (LOD) (Fig. 4D and E).

Reduction of NSB by PEO-pores is also evident from the sensorgrams obtained on blank, REF/CM-dextran and REF/PEO-pores, spots during milk and serum injections (Fig. 5). Three parameters were compared: the bulk shift caused by the differences in refractive index between the running buffer and the sample, NSB and a baseline build-up (Fig. 5A). The bulk responses measured in buffer and serum did not differ between CM-dextran and the PEO-surfaces indicating a complete diffusion of the sample components through the filter (Fig. 5B). Alternatively, in milk, the bulk responses measured were consistently lower in the PEO-pores, possibly indicating the exclusion of micelles or large aggregates from the sensing region by the sieve (Fig. 5C). The NSB was evidently lower on PEO-pores: 50% reduction in milk and 25% reduction in serum. The highest NSB observed for both surfaces was in porcine serum, which agrees with the inhibition plateau of the enrofloxacin assay in serum (Fig. 5D).

Fig. 5 iSPR measurements on blank PEO-pores in complex matrices. (A) Zeroed sensorgrams measured on blank PEO-pores (blue) and blank CM-dextran (black) during buffer, diluted full fat milk and diluted porcine serum injections. (B–D) Bulk responses, non-specific binding and baseline build up measured during enrofloxacin calibration curve injections in buffer, diluted full fat milk and diluted porcine serum, respectively, on blank PEO-pores (hollow blue) and blank CM-dextran (black). For bulk response calculation, the angle shift before the end of sample injection (1) was taken (around 550 seconds). For non-specific binding calculation, the angle shift during the dissociation phase (2) was taken (around 600 seconds). For baseline build up calculation, the angle shift after the regeneration (3) was taken (around 750 seconds). Bulk shift responses were plotted on the left y-axis, whereas NSB and baseline build-up response were plotted on the right y-axis for comparison.

Furthermore, an interesting trend in baseline build-up was observed. The PEO-pores showed a rather constant baseline in each measurement cycle, whereas CM-dextran showed a baseline build-up which was somewhat random in serum and milk (between 0 and 10 m⁰). This baseline build-up hinders analyte binding in the subsequent injection causing high variability of repeats. PEO-pores greatly reduced the build-up in both complex matrices, explaining the improvement in reproducibility and increase of sensitivity of the assay. In milk, the PEO-pores probably sieved out the lipid and casein micelles outside of the sensing region. Caseins, comprising 80% of milk proteins, form aggregates of several thousand individual protein molecules with average diameters of 150 to 200 nm, which are stable through drying and rehydration processes and thus are present not only in native starting material but in many byproducts such as milk powders as well.²³ Notwithstanding, most of the total milk lipid is in the form of globules ranging in size from 0.1 to 15 μm in diameter, depending on the product.^24,25 Both the size of the casein micelles and the fat globules in milk exceed the diameter of the PEO-pores, preventing them from reaching the sensing surface and interfering with the specific analyte–ligand interaction. Additional mechanism for PEO-pores action could be the reduction in the CM-dextran surface area available for NSB. Despite the advantages of the CM-dextran over plain gold surfaces it is recognized that the CM-dextran itself can be a diffusion limiting factor for large molecules.^26–29 Therefore the fat globules and micelles would not completely diffuse inside the dextran layer, meaning that most of the NSB most likely occurs in the top layers of dextran. However, since the extent of penetration of large matrix components to the dextran layer is unknown, the mechanism of action for PEO pores could be a combination of both: reduction of area available for NSB and sieving out of the particles larger than 100 nm.

Overall the PEO-pores displayed a promising capability to selectively shield the SPW from negative matrix effects from both low and high molecular weight matrix components at all analyte concentrations tested. The PEO-pore layer improves the SPR assay applicability and sensitivity by broadening of the assay's dynamic range and through reducing the sample to sample variability. The current study was aimed at testing whether PEO-pores would be beneficial for immunoassay performance in complex matrixes. Since positive indication was found, the mechanism behind the particular mode of action of the PEO-pores will be investigated in further research.

Conclusions

Nonspecific binding of sample components in complex matrices to the sensor surface is a common bottleneck in the development of analytical bioassays on SPR platform, and are usually tackled through extensive sample pre-treatment. The source of NSB is matrix dependent, and its source is often not identified. In this study we proposed a simultaneous sample pre-treatment and analysis by engineering the sensor chip surface. By selectively shielding the SPW probed volume with a nanopatterned PEO membrane (PEO-pores), on top of the carboxymethylated dextran layer, we lowered the NSB from matrix components in the SPW probed volume of the sensor chip. The overall baseline build-up of the probed region was reduced due to lowered exposure to nonspecific binding of molecules from the complex matrix. The use of PEO-pores for antibiotic detection in full fat milk and porcine serum resulted in improvement of the LOD, through a reduction of NSB and sample to sample variability. Following this proof of principle, future work will focus on further sensitivity enhancement by the expected higher tolerance of the PEO-pores to higher concentrations of complex biological sample matrices and the exploration of different sieve height and pore size. In our view, the advances described for on-chip sample preparation show a great potential to improve not only SPR-based assays performance in complex matrices but also other optical evanescent wave-based sensor devices suffering from similar sample preparation drawbacks. Thus, the proposed method widens the application range of these technologies in the bioanalytical field.

Acknowledgements

We thank Dr Philippe Delahaut for providing anti-norfloxacin antibody and Dutch Technology Foundation (STW) for funding this study (project TMF 6635: “Multi-analyte screening with μfluidic biochips”).

References

1. J. N. Anker, W. P. Hall, O. Lyandres, N. C. Shah, J. Zhao and R. P. Van Duyne, Nat. Mater., 2008, 7, 442–453.
2. N. F. C. Visser and A. J. R. Heck, Expert Rev. Proteomics, 2008, 5, 425–433.
3. K. Phillips and Q. Cheng, Anal. Bioanal. Chem., 2007, 387, 1831–1840.
4. C. T. Campbell and G. Kim, Biomaterials, 2007, 28, 2380–2392.
5. R. L. Rich and D. G. Myszka, Drug Discovery Today: Technol., 2004, 1, 301–308.
6. C. Situ, A. R. G. Wylie, A. Douglas and C. T. Elliott, Talanta, 2008, 76, 832–836.
7. J.-F. Masson, T. Battaglia, J. Cramer, S. Beaudoin, M. Sierks and K. Booksh, Anal. Bioanal. Chem., 2006, 386, 1951–1959.
8. J.-F. Masson, T. M. Battaglia, M. J. Davidson, Y.-C. Kim, A. M. C. Prakash, S. Beaudoin and K. S. Booksh, Talanta, 2005, 67, 918–925.
9. J. Andrade, Surface and Interfacial Aspects of Biomedical Polymers, Plenum, New York, 1985.
10. Z. Teng, J. L. Moore, A. R. Craig, G. Hickey, C. Schelp and T. Q. Wei, US Pat., US 2007/0122850 A1, 2007.
11. H. Raether, Surface plasmons on smooth and rough surfaces and on gratings, Springer Verlag, 1998.
12. E. Kretschmann and H. Raether, Z. Naturforsch., A: Astrophys., Phys. Phys. Chem., 1968, 23, 2135.
13. H. Vaisocherová, Z. Zhang, W. Yang, Z. Cao, G. Cheng, A. D. Taylor, M. Piliarik, J. Homola and S. Jiang, Biosens. Bioelectron., 2009, 24, 1924–1930.
14. O. R. Bolduc, C. M. Clouthier, J. N. Pelletier and J.-F. Masson, Anal. Chem., 2009, 81, 6779–6788.
15. S.-M. Yang, S. G. Jang, D.-G. Choi, S. Kim and H. K. Yu, Small, 2006, 2, 458–475.
16. P. A. Shashishekar, J. Chunmin, A. C. Larry, A. M.-R. Nancy and J. N. Roger, Wiley Interdiscip. Rev.: Nanomed. Nanobiotechnol., 2009, 1, 568–581.
17. W. Haasnoot, H. Gerçek, G. Cazemier and M. W. F. Nielen, Anal. Chim. Acta, 2007, 586, 312–318.
18. W. Haasnoot, E. E. M. G. Loomans, G. Cazemier, R. Dietrich, R. Verheijen, A. A. Bergwerff and R. W. Stephany, Food Agric. Immunol., 2002, 14, 15–27.
19. W. Haasnoot, N. G. E. Smits, A. E. M. Kemmers-Voncken and M. G. E. G. Bremer, J. Dairy Res., 2004, 71, 322–329.
20. M. Lejeune, L. M. Lacroix, F. Brétagnol, A. Valsesia, P. Colpo and F. Rossi, Langmuir, 2006, 22, 3057–3061.
21. F. Brétagnol, M. Lejeune, A. Papadopoulou-Bouraoui, M. Hasiwa, H. Rauscher, G. Ceccone, P. Colpo and F. Rossi, Acta Biomater., 2006, 2, 165–172.
22. H. E. Indyk, Int. Dairy J., 2010, 20, 106–112.
23. D. G. Dalgleish and M. Corredig, Annu. Rev. Food Sci. Technol., 2012, 3, 449–467.
24. O. Ménard, S. Ahmad, F. Rousseau, V. Briard-Bion, F. Gaucheron and C. Lopez, Food Chem., 2010, 120, 544–551.
25. B. Coulter, Milk Fat Size Distribution by Laser Diffraction, Report BCPCS002.LE, 6 February 2006.
26. S. Lofas and B. Johnsson, J. Chem. Soc., Chem. Commun., 1990, 1526–1528.
27. C. Rodriguez Emmenegger, E. Brynda, T. Riedel, Z. Sedlakova, M. Houska and A. B. Alles, Langmuir, 2009, 25, 6328–6333.
28. M. W. Oli, W. P. McArthur and L. J. Brady, J. Microbiol. Methods, 2006, 65, 503–511.
29. Q. Yang, C. Kaul and M. Ulbricht, Langmuir, 2010, 26, 5746–5752.

---