Portable tumor biosensing of serum by plasmonic biochips in combination with nanoimprint and microfluidics

2.1 Materials and characterization methods

Soft transparent polymer stamps and TU-170 UV nanoimprint resist are from Obducat AB (Lund, Sweden). 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC), N-hydroxysuccinimide (NHS), and phosphate buffer saline (PBS) are from LvYin Reagent (Xiamen, China). Polydimethylsiloxane (PDMS) is from Dow Corning (Midland, MI, USA). 11-Mercaptoundecanoic acid (MUA) is from Cool Chemistry (Beijing, China). Bovine serum albumin (BSA) is from Sangon Biotech (Shanghai, China). CEA, anti-CEA, α-fetoprotein (AFP), ferritin (Fer), carbohydrate antigen 19-9 (CA 19-9), and anti-CA 19-9 are from Siemens Healthcare Diagnostics (New York, NY, USA). The carbohydrate antigen 242 (CA 242) and anti-CA 242 are from Shenzhen New Industries Biomedical Engineering (Shenzhen, China). Scanning electron microscope (SEM; Hitachi S-4800) and atomic force microscopy (AFM; SPA400) systems are used to characterize the morphology of PNA surface.

In order to authenticate the changes on gold surface by functionalization of nanoimprinted chips, electrochemical measurements are performed by a CHI660 electrochemical workstation from CH Instruments (Austin, TX, USA) with a three-electrode system, which includes an auxiliary platinum electrode, a reference saturated calomel electrode, and a gold electrode as the working electrode. We have carried out measurements for the gold electrode surface without functionalization, with MUA, and with MUA/anti-CEA, respectively. For each measurement, the gold electrode is washed by deionized water thoroughly and immersed in a solution of 5 mm potassium ferricyanide. Voltammetry measurements have been performed for a potential range from −0.2 to 0.7 V at a sweep rate of 0.1 V/s.

2.2 Fabrication of immunochips

As shown in Figure 1B, in the soft nanoimprint process, a polymer stamp is used to duplicate the nanocave arrays on a 2 in nickel mold by thermal nanoimprint, and a 2-inch silicon wafer spin-coated by the UV resist of about 210 nm is covered by the polymer stamp and processed at 65°C with a pressure of 30 bar for 1 min. After UV exposure, the wafer is etched by oxygen plasma and evaporated by 5 nm chromium and 250 nm gold. The PNA wafer is cut into small pieces (5 mm in diameter) by a laser cutting machine in order to be integrated with the microfluidic chip.

In the fabrication process of microfluidic channels shown in Figure 1B, copper microwires with a diameter of 150 μm are fixed on a rectangle mold, and PDMS is then cast on the mold and kept at 60°C for 3 h. After that, the PDMS is dip in ethanol for 10 min, and the microwires are pulled out. Next, the PDMS is heated in an oven at 60°C for 30 min, and microfluidic inlets, outlets, and detection chambers are perforated by a mechanical mold. A glass substrate, PNA pieces, and a glass cover slice are sequentially bonded and integrated into an entire packaged chip. Redundant openings of microfluidic channels are treated and blocked by a small quantity of unsolidified PDMS at 60°C for 3 h.

The PNA pieces in the detection chambers are antibody-functionalized by covalently immobilizing anti-CEA (or anti-CA 19-9/anti-CA 242) onto the thiolated modified surfaces of PNA, as shown in Figure 2. In this process, 1 mm MUA solution of ethanol is injected into the detection chambers through the inlets of microfluidic channels and maintained for 3 h to form a self-assembled monolayer on the surface of PNA. After that, the MUA solution is exhausted from the chip, and ethanol and deionized water are injected into the microfluidic channels and chambers for thorough cleaning. Next, a solution with 400 mm EDC and 100 mm NHS is injected into the detection chambers and maintained for 30 min in order to activate the MUA monolayer, which is followed by extracting the solution and cleaning with deionized water. Finally, 40 μg/ml of anti-CEA (or anti-CA 19-9/anti-CA 242) solution is injected and maintained for 1 h in order to immobilize the antibody on the PNA surface, BSA (1 wt%) is injected to block the remaining activated carboxylic groups and to prevent nonspecific adsorption from human serums on the sensing surface, and the chip is injected and washed with deionized water and ready for use. In order to ensure the consistency of experiments, all the solutions injected to the chambers are overflowing each time.

Figure 2:

Schematic drawing of functionalization, detection, and recycling for the PNA surface.

The SEM image of the PNA surface is provided in the left corner, with the red scale bar of 1 μm. (The schematic drawing does not represent the precise sizes of the adsorbed materials compared with the dimensions of nanocaves.)

2.3 Optical measurement and simulation

A reflection optical fiber probe in combination with a portable UV-visible spectrometer from Avantes BV (Apeldoom, Netherlands) is used to measure the optical spectra of all the PNA surfaces in the detection chambers. The fiber probe consists of six lighting fibers and one reading fiber (200 μm core) with an operational range in the visible region. In optical measurements, the fiber probe is perpendicularly located on top of the chamber with a distance of ~2 mm to the glass cover of the chip, and the reflectance from a silver mirror is used as a spectral reference. At each measurement, the detection chamber is washed thoroughly and filled with deionized water. A full-wave finite element numerical method is used to simulate the optical reflectance of PNA under normal incidence by adopting infinite unit cells with Floquet periodic boundary conditions. We simply assume that the gold nanocaves have a circular shape for a theoretical approximation. The optical constants of gold are obtained from the literature [30]. Adaptive inhomogeneous tetrahedral mesh is adopted to discretize the unit cell, and the minimum edge length of the mesh element is as small as 0.5 nm to ensure the convergence and reproducibility of numerical simulations [31].

2.4 Detection of CEA

The detection of CEA is based on its specific binding on the anti-CEA (40 μg/ml in PBS). In order to indicate the specificity of CEA for the immunochip, five kinds of PBS solutions of 20 ng/ml CEA, 30 ng/ml AFP, 100 ng/ml Fer, 20 ng/ml CEA in combination with 30 ng/ml AFP, and 20 ng/ml CEA in combination with 100 ng/ml Fer are injected into five separate microfluidic chambers, respectively. The chambers are filled with the solutions for 30 min and followed by the optical measurements with the reflection optical probe. Five groups of CEA solutions with various concentrations of 5, 10, 20, 30, and 40 ng/ml are injected into five separate chambers, incubated for 30 min, and followed by the optical measurements with the reflection fiber probe. For each CEA concentration, we adopt three groups of solution for spectral measurements, which are further used for s.d. calculations. The chips used for antigen detection can be recycled by removing proteins adsorbed on the PNA surface. As shown in Figure 2, after the antigen detection, 0.5 mol/l hydrochloric acid solution is injected into the chambers, maintained for 20 min, and washed thoroughly by deionized water. Spectral measurements are performed to verify the result of the recycling process.

2.5 Diagnosis of tumor markers in human serums

All the human serum samples are collected by the First Affiliated Hospital of Xiamen University (Xiamen, China) with the consent from some individuals and are analyzed in advance by ADVIA Centaur XP Immunoassay System (Erlangen, Germany), which is based on CLIA. The serum samples are stored at −20°C and used for the immunochip diagnosis at room temperature after thawing. Six serum samples from different individuals are injected into their respective detection chambers through microfluidic channels on the chip and allowed for the immunological reaction of CEA for 30 min. After that, the serums are extracted, and the detection chambers filled with deionized water are measured by the reflection optical probe. The concentration of CEA is determined on the basis of the optical sensing curve obtained in Section 2.4. Additionally, a serum sample from one individual is injected into three separate microfluidic chambers immobilized by anti-CEA, anti-CA 19-9, and anti-CA 242, respectively. After incubating for no more than 30 min, multiple tumor markers from the same serum is detected simultaneously in the immunochip.

3.1 Analysis of plasmonic nanostructures

The 2 in PNA wafers are fabricated by soft UV nanoimprint and gold deposition [32]. The plasmonic sensing components integrated in the detection chambers are cut out from the PNA wafers. The soft UV nanoimprint is a versatile and inexpensive fabrication method for mass production of nanostructures [33], [34]. As shown in the SEM image of Figure 2, the gold nanocaves are uniform with the periodic spacing of ~500 nm and the diameter of ~250 nm, and the depth of ~200 nm for the nanocaves is also confirmed by an AFM system (Figure S1). The structure uniformity in a large area benefits the specification of standard and unified PNA sensors for plasmonic immunochips. The gold nanocaves act as two-dimensional arrays with orthogonal reciprocal lattice vectors, which allows for the grating-coupled plasmon modes and traps the spatial light around the resonant wavelength. The resonant wavelengths of these modes under normal incidence can be estimated by the following equation [35], [36]:

where P, integers (m, n), ε_Au, and ε represent the periodicity of gold nanocaves, the order of plasmonic mode, the permittivity of gold, and the environmental permittivity surrounding the gold surface, respectively. On the basis of equation (1), the period around 500 nm is selected by considering the plasmonic light confinement in the visible range (especially for the sensing in the solution), which facilitates the use of the low-cost photodetector of visible light. The diameter of 250 nm and the depth of 200 nm are large enough to contain and adsorb MUA, antibodies, and antigens, which might have dimensions of up to tens of nanometers. The selection for the geometry, dimensions, and periodicity of the PNA are also determined by the related references [15], [19]. As shown in the reflectance spectrum of Figure 3A, the resonant wavelength of the (1, 0) order plasmonic mode is indicated by the spectral dip around λ=569 nm, which is due to the light confinement and resonant dissipation in gold under normal incidence from the air. Figure 3A also demonstrates that the optical simulation has good consistency with the spectral measurement, which further confirms our theoretical analysis. The electric field distribution of the resonant mode is illustrated in Figure 3B, and it indicates that the electric fields are tightly confined and enhanced surrounding the periodic nanocaves, which enables the sensitive detection of materials captured on the PNA surface. The estimated electromagnetic field decay length on top of the PNA surface is within the scale of around 100 nm. On this condition, the length scale of the detected protein molecules in our experiments would remain within the enhanced electromagnetic field decay length. Such a decay length is larger than that of biosensors based on localized surface plasmon resonance (10–30 nm) and smaller than that of conventional biosensors based on surface plasmon resonance (hundreds of nanometers) [19]. This scale design is well fit for the accumulated lengths of MUA, antibodies, and antigens in our detection.

Figure 3:

Optical performance of PNA.

(A) Measured and simulated reflectance spectra of bare PNA under normal incidence in the air. (B) Electric field distribution for the (1, 0) order plasmonic resonance mode at λ=569 nm. (C) Reflectance spectra of the (1, 0) order resonance in the PNA for various solvents. (D) Reflectance dip shift of the (1, 0) order resonance as a function of refractive index.

In order to evaluate the detection sensitivity of the PNA surface, bulk refractive index measurements are performed by successively immersing the same PNA sensor in four different solvents, including deionized water, ethanol, isopropanol, and glycol. As shown in Figure 3C, the reflectance dip of the (1, 0) order resonance shifts to longer wavelengths because of the increase of the refractive index (or permittivity) surrounding the gold nanocaves, which is consistent with the estimation by equation (1). The shift of the resonant dip as a function of refractive index is plotted in Figure 3D, and the error bars of measured data are too small to be observed. Figure 3D shows that the sensitivity of bulk refractive index is as high as 490.2 nm/RIU in the visible range, which is advantageous and sufficient for high-sensitivity biosensing compared with other plasmonic surfaces by nanoimprint lithography [37], [38]. In view of the adsorbed materials, including MUA, anti-CEA, and CEA, their sizes are much smaller than the diameter of nanocaves (~250 nm), and they can be equally adsorbed on the top surface and the interior of the nanocaves. On the basis of equation (1), ε can also be considered as an effective permittivity combining the adsorbed materials and the surrounding solvent, and the resonant wavelength increases as the effective permittivity (or effective refractive index) becomes larger. Specific binding of CEA with various concentrations to the PNA sensing surface changes the local refractive index differently, which can be monitored by measuring wavelength shifts of the resonant dip. This fundamental physics provides a convincing guide for the study of plasmonic sensors on the basis of gold nanocaves.

In addition, the use of gold nanocave array has several advantages in view of practical applications. First, the entire sensing surface is encapsulated by gold, and this benefits the process of antibody functionalization, in which the superficial gold needs to be thiolated by MUA in advance. Second, gold is chemically stable, and only the material exposure of gold will maintain the steadiness of the plasmonic sensors, especially for the recycling procedure by treatment of HCl. Third, instead of plasmonic angular coupling on a prism or transmission measurement by a microscope, the PNA allows the reflection scheme of normal incidence with the illumination source and the photodetector at the same side, which facilitates the integration for a compact portable detection system. Fourth, the operational wavelengths of the nanostructure are controlled within the visible range, which enables the use of low-cost charge-coupled digital camera devices for portable applications.

3.2 Immobilization of PNA with anti-CEA

Sufficient immobilization of active antibody on the gold surface is a critical procedure for biomarker detection based on plasmonics [39], [40]. In our study, a self-assembled monolayer of MUA is adopted as an intermediate material layer to immobilize anti-CEA on the gold surface by the scheme of covalent-bonding. In order to confirm that MUA and anti-CEA are immobilized on the surface of PNAs, optical reflectance measurements and electrochemical measurements are performed for further analysis, as shown in Figure 4A and B. As observed in Figure 4A, compared with the spectrum of PNA without immobilization, the spectra redshifted to 3.46 and 7.47 nm for the PNA with MUA and the PNA with MUA/anti-CEA, respectively, because of the increase of local refractive index. The spectral characterization for the sensing surface in the air is also performed by nitrogen drying and reflectance measurements after each immobilization step (Figure S2), and the resonant dip shift further illuminates the steady bonding of MUA and anti-CEA on the PNA surface.

Figure 4:

Characterizations of bare gold, gold with MUA, and gold with MUA/anti-CEA.

(A) Reflectance spectra with filled deionized water in detection chambers. (B) Electrochemical voltammetry measurement of a gold electrode.

The antibody functionalization is also evaluated by electrochemical measurements using a gold electrode, as shown in Figure 4B. The voltammetric curve for each immobilization step implies the conductive behavior of the gold electrode surface, and [Fe(CN)₆]³⁻ ions in the solution act as an electrochemical probe for the surface change. The peak current due to the redox process decreases significantly when MUA is immobilized on the gold surface, because the long chain MUA has no conductivity and its negative charge obstructed [Fe(CN)₆]³⁻ ions diffusing to the gold electrode surface. As anti-CEA is bound on Au/MUA, the electrode surface tends to electro-neutrality, and there is no obvious current peak observed. On the basis of the above results, both the measurements of optical spectra and electrochemical voltammetry demonstrate that MUA and anti-CEA are immobilized on the PNA surface subsequently as anticipated.

3.3 Immunologic specificity

The developed plasmonic immunochips are intended for ultimate CEA detection in human serums. As the composition of human serum is complex, including a variety of antibodies and antigens, the anti-CEA functionalized on the surface of PNA should have the capability to capture CEA only for diagnosis detection. In order to confirm this, the solutions of AFP (~70 kDa), CEA (~200 kDa), and Fer (~474 kDa) are applied to test the selectivity of the immunochip for CEA detection. These three antigens have various molecular sizes, which are much smaller than the diameter and depth of the nanocaves, and they can be contained on the top surface of the array and in the interior of nanocaves when the solutions are applied on the plasmonic sensor to test the immunologic specificity of anti-CEA. As shown in Figure 5A, there is no shift of resonance wavelength for the solutions of 100 ng/ml Fer and 30 ng/ml AFP, whereas there is the same red shift of ~2.3 nm for the other three solutions with the same CEA concentration of 20 ng/ml. This is because the surface of functionalized PNA only captured CEA specifically from the solutions without CEA, and there is no specific binding of Fer or AFP on the surface. Additional evidences by AFM and energy dispersive spectrometer are also provided in Figures S3 and S4 to confirm the capture of CEA. These results indicate the immunologic specificity of our chips for CEA detection.

Figure 5:

Immunologic specificity and sensing capability.

(A) Reflectance spectra of specific tests. (B) Resonant wavelength dip shift as a function of CEA concentration. The fitting line is based on measurements from 5 to 40 ng/ml, and the error bars are based on s.d. calculation for three data points measured at each concentration. The linear fitting function y and correlation coefficient R are denoted in the plot.

In view of waste recycling and environmental protection, we have recycled used plasmonic immunochips. The 0.5 mol/l hydrochloric acid solution is injected to treat the surface of PNA immobilized with the antibody and antigen, and the performance for five runs of recycling is verified by comparing its optical spectra before and after the treatment. As observed in Figure S5, the wavelength of plasmonic resonance after the acid treatment is the same as that of the original plasmonic surface with MUA, which demonstrates the reproducibility and stability of recycled chips for CEA detection.

3.4 Sensing capability of immunochips

In order to test the sensitivity of plasmonic immunochips, we incubate five detection chambers with different concentrations of CEA ranging from 5 to 40 ng/ml. The largest detected concentration of 40 ng/ml is controlled within the range of antigen saturation on the PNA surface. All measurements have the same incubation time and the same liquid volume in the microfluidic channels and detection chambers, with the aim of ensuring the measurement precision. On the basis of the measured spectra (Figure S6), the concentration-dependent shift of plasmonic resonant dip is summarized in Figure 5B, in which the dip shift increased proportionally to the CEA concentration. The linear fitting has a correlation coefficient R of 0.9985, which is very close to the value 1 and demonstrates very good linear relationship between concentration of CEA and resonant wavelength dip shift. In our system, the detection of 5 ng/ml CEA solvent can be clearly designated in experiments, which is much lower than the threshold concentration of 20 ng/ml for early cancer prediction. It is worth mentioning that, in all the measurements, the microfluidic channels and detection chambers are filled with corresponding operational solvents. This not only keeps the activity of adsorbed anti-CEA and CEA in detection but also avoids inconvenient drying of the PNA in the microfluidic chips. Additionally, the detection with filled solvents is more versatile for the application of microfluidic chips and ensures quantitative analysis and detection precision. Therefore, the plasmonic biochips functionalized by anti-CEA provide an exceptional sensing capability for label-free, convenient, sensitive, and reliable detection of CEA.

3.5 Rapid detection of human serums with microfluidic channels

The performance of quantitative CEA detection by the chip for medicine has been validated by testing the serum samples from six individuals, and the results are compared with measurements from CLIA in The First Affiliated Hospital of Xiamen University (Xiamen, China). As shown in Figure 6A, the results detected by the plasmonic immunochips demonstrate good agreement with the results from CLIA. When the detected concentration of CEA in serum is higher than 10 ng/ml, there is a deviation of less than |±5.9%| in comparison with the conventional clinical examination. The chip detections for six individuals have been performed concurrently and completed in 30 min in room temperature (Figure S7), which demonstrates the potential advantage of combining nanoimprint and microfluidics for high-throughput quantitative analysis of CEA. On the basis of the tested results from Serum 1 to Serum 6 in Figure 6A, Individual 1, Individual 2, and Individual 3 have a CEA concentration lower than 20 ng/ml, which is a normal level and might imply the healthy status; the CEA concentrations for Individual 4, Individual 5, and Individual 6 are higher than the threshold of 20 ng/ml, which might indicate the disease of tumor.

Figure 6:

Detection of human serums.

(A) CEA detection results of serum samples from six individuals using the plasmonic immunochip, compared with using CLIA. (B) Concurrent sensing of three tumor markers in separate microfluidic channels.

Besides, the clinical detection of CEA for tumor diagnosis has been adopted for many years, but only CEA detection might be incomplete for accurate prognosis of the type of tumor. A type of tumor usually caused higher levels of many kinds of biomarker in the serum [41]. For instance, the combined detections of CEA, CA 19-9, and CA 242 provide significant evidences for the precise diagnosis and prognosis of pancreatic carcinoma [42]. In view of the detection importance of multi-tumor markers, we have performed a proof-of-concept experiment for the simultaneous sensing of CEA, CA 19-9, and CA 242 in the serum of one individual. As shown in Figure 6B, by using three separated microfluidic channels, the three kinds of tumor markers are sensed by the shift of plasmonic resonant wavelength within 30 min. This implies the promising potential of the chips for comprehensive and accurate detection of various kinds of cancer.